xref: /openbmc/linux/fs/xfs/xfs_ioctl.c (revision 160b8e75)
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 #include "scrub/xfs_scrub.h"
48 #include "xfs_sb.h"
49 
50 #include <linux/capability.h>
51 #include <linux/cred.h>
52 #include <linux/dcache.h>
53 #include <linux/mount.h>
54 #include <linux/namei.h>
55 #include <linux/pagemap.h>
56 #include <linux/slab.h>
57 #include <linux/exportfs.h>
58 
59 /*
60  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
61  * a file or fs handle.
62  *
63  * XFS_IOC_PATH_TO_FSHANDLE
64  *    returns fs handle for a mount point or path within that mount point
65  * XFS_IOC_FD_TO_HANDLE
66  *    returns full handle for a FD opened in user space
67  * XFS_IOC_PATH_TO_HANDLE
68  *    returns full handle for a path
69  */
70 int
71 xfs_find_handle(
72 	unsigned int		cmd,
73 	xfs_fsop_handlereq_t	*hreq)
74 {
75 	int			hsize;
76 	xfs_handle_t		handle;
77 	struct inode		*inode;
78 	struct fd		f = {NULL};
79 	struct path		path;
80 	int			error;
81 	struct xfs_inode	*ip;
82 
83 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
84 		f = fdget(hreq->fd);
85 		if (!f.file)
86 			return -EBADF;
87 		inode = file_inode(f.file);
88 	} else {
89 		error = user_lpath((const char __user *)hreq->path, &path);
90 		if (error)
91 			return error;
92 		inode = d_inode(path.dentry);
93 	}
94 	ip = XFS_I(inode);
95 
96 	/*
97 	 * We can only generate handles for inodes residing on a XFS filesystem,
98 	 * and only for regular files, directories or symbolic links.
99 	 */
100 	error = -EINVAL;
101 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
102 		goto out_put;
103 
104 	error = -EBADF;
105 	if (!S_ISREG(inode->i_mode) &&
106 	    !S_ISDIR(inode->i_mode) &&
107 	    !S_ISLNK(inode->i_mode))
108 		goto out_put;
109 
110 
111 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
112 
113 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
114 		/*
115 		 * This handle only contains an fsid, zero the rest.
116 		 */
117 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
118 		hsize = sizeof(xfs_fsid_t);
119 	} else {
120 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
121 					sizeof(handle.ha_fid.fid_len);
122 		handle.ha_fid.fid_pad = 0;
123 		handle.ha_fid.fid_gen = inode->i_generation;
124 		handle.ha_fid.fid_ino = ip->i_ino;
125 		hsize = sizeof(xfs_handle_t);
126 	}
127 
128 	error = -EFAULT;
129 	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
130 	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
131 		goto out_put;
132 
133 	error = 0;
134 
135  out_put:
136 	if (cmd == XFS_IOC_FD_TO_HANDLE)
137 		fdput(f);
138 	else
139 		path_put(&path);
140 	return error;
141 }
142 
143 /*
144  * No need to do permission checks on the various pathname components
145  * as the handle operations are privileged.
146  */
147 STATIC int
148 xfs_handle_acceptable(
149 	void			*context,
150 	struct dentry		*dentry)
151 {
152 	return 1;
153 }
154 
155 /*
156  * Convert userspace handle data into a dentry.
157  */
158 struct dentry *
159 xfs_handle_to_dentry(
160 	struct file		*parfilp,
161 	void __user		*uhandle,
162 	u32			hlen)
163 {
164 	xfs_handle_t		handle;
165 	struct xfs_fid64	fid;
166 
167 	/*
168 	 * Only allow handle opens under a directory.
169 	 */
170 	if (!S_ISDIR(file_inode(parfilp)->i_mode))
171 		return ERR_PTR(-ENOTDIR);
172 
173 	if (hlen != sizeof(xfs_handle_t))
174 		return ERR_PTR(-EINVAL);
175 	if (copy_from_user(&handle, uhandle, hlen))
176 		return ERR_PTR(-EFAULT);
177 	if (handle.ha_fid.fid_len !=
178 	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
179 		return ERR_PTR(-EINVAL);
180 
181 	memset(&fid, 0, sizeof(struct fid));
182 	fid.ino = handle.ha_fid.fid_ino;
183 	fid.gen = handle.ha_fid.fid_gen;
184 
185 	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
186 			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
187 			xfs_handle_acceptable, NULL);
188 }
189 
190 STATIC struct dentry *
191 xfs_handlereq_to_dentry(
192 	struct file		*parfilp,
193 	xfs_fsop_handlereq_t	*hreq)
194 {
195 	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
196 }
197 
198 int
199 xfs_open_by_handle(
200 	struct file		*parfilp,
201 	xfs_fsop_handlereq_t	*hreq)
202 {
203 	const struct cred	*cred = current_cred();
204 	int			error;
205 	int			fd;
206 	int			permflag;
207 	struct file		*filp;
208 	struct inode		*inode;
209 	struct dentry		*dentry;
210 	fmode_t			fmode;
211 	struct path		path;
212 
213 	if (!capable(CAP_SYS_ADMIN))
214 		return -EPERM;
215 
216 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
217 	if (IS_ERR(dentry))
218 		return PTR_ERR(dentry);
219 	inode = d_inode(dentry);
220 
221 	/* Restrict xfs_open_by_handle to directories & regular files. */
222 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
223 		error = -EPERM;
224 		goto out_dput;
225 	}
226 
227 #if BITS_PER_LONG != 32
228 	hreq->oflags |= O_LARGEFILE;
229 #endif
230 
231 	permflag = hreq->oflags;
232 	fmode = OPEN_FMODE(permflag);
233 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
234 	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
235 		error = -EPERM;
236 		goto out_dput;
237 	}
238 
239 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
240 		error = -EPERM;
241 		goto out_dput;
242 	}
243 
244 	/* Can't write directories. */
245 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
246 		error = -EISDIR;
247 		goto out_dput;
248 	}
249 
250 	fd = get_unused_fd_flags(0);
251 	if (fd < 0) {
252 		error = fd;
253 		goto out_dput;
254 	}
255 
256 	path.mnt = parfilp->f_path.mnt;
257 	path.dentry = dentry;
258 	filp = dentry_open(&path, hreq->oflags, cred);
259 	dput(dentry);
260 	if (IS_ERR(filp)) {
261 		put_unused_fd(fd);
262 		return PTR_ERR(filp);
263 	}
264 
265 	if (S_ISREG(inode->i_mode)) {
266 		filp->f_flags |= O_NOATIME;
267 		filp->f_mode |= FMODE_NOCMTIME;
268 	}
269 
270 	fd_install(fd, filp);
271 	return fd;
272 
273  out_dput:
274 	dput(dentry);
275 	return error;
276 }
277 
278 int
279 xfs_readlink_by_handle(
280 	struct file		*parfilp,
281 	xfs_fsop_handlereq_t	*hreq)
282 {
283 	struct dentry		*dentry;
284 	__u32			olen;
285 	int			error;
286 
287 	if (!capable(CAP_SYS_ADMIN))
288 		return -EPERM;
289 
290 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
291 	if (IS_ERR(dentry))
292 		return PTR_ERR(dentry);
293 
294 	/* Restrict this handle operation to symlinks only. */
295 	if (!d_is_symlink(dentry)) {
296 		error = -EINVAL;
297 		goto out_dput;
298 	}
299 
300 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
301 		error = -EFAULT;
302 		goto out_dput;
303 	}
304 
305 	error = vfs_readlink(dentry, hreq->ohandle, olen);
306 
307  out_dput:
308 	dput(dentry);
309 	return error;
310 }
311 
312 int
313 xfs_set_dmattrs(
314 	xfs_inode_t     *ip,
315 	uint		evmask,
316 	uint16_t	state)
317 {
318 	xfs_mount_t	*mp = ip->i_mount;
319 	xfs_trans_t	*tp;
320 	int		error;
321 
322 	if (!capable(CAP_SYS_ADMIN))
323 		return -EPERM;
324 
325 	if (XFS_FORCED_SHUTDOWN(mp))
326 		return -EIO;
327 
328 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
329 	if (error)
330 		return error;
331 
332 	xfs_ilock(ip, XFS_ILOCK_EXCL);
333 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
334 
335 	ip->i_d.di_dmevmask = evmask;
336 	ip->i_d.di_dmstate  = state;
337 
338 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
339 	error = xfs_trans_commit(tp);
340 
341 	return error;
342 }
343 
344 STATIC int
345 xfs_fssetdm_by_handle(
346 	struct file		*parfilp,
347 	void			__user *arg)
348 {
349 	int			error;
350 	struct fsdmidata	fsd;
351 	xfs_fsop_setdm_handlereq_t dmhreq;
352 	struct dentry		*dentry;
353 
354 	if (!capable(CAP_MKNOD))
355 		return -EPERM;
356 	if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
357 		return -EFAULT;
358 
359 	error = mnt_want_write_file(parfilp);
360 	if (error)
361 		return error;
362 
363 	dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
364 	if (IS_ERR(dentry)) {
365 		mnt_drop_write_file(parfilp);
366 		return PTR_ERR(dentry);
367 	}
368 
369 	if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
370 		error = -EPERM;
371 		goto out;
372 	}
373 
374 	if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
375 		error = -EFAULT;
376 		goto out;
377 	}
378 
379 	error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
380 				 fsd.fsd_dmstate);
381 
382  out:
383 	mnt_drop_write_file(parfilp);
384 	dput(dentry);
385 	return error;
386 }
387 
388 STATIC int
389 xfs_attrlist_by_handle(
390 	struct file		*parfilp,
391 	void			__user *arg)
392 {
393 	int			error = -ENOMEM;
394 	attrlist_cursor_kern_t	*cursor;
395 	struct xfs_fsop_attrlist_handlereq __user	*p = arg;
396 	xfs_fsop_attrlist_handlereq_t al_hreq;
397 	struct dentry		*dentry;
398 	char			*kbuf;
399 
400 	if (!capable(CAP_SYS_ADMIN))
401 		return -EPERM;
402 	if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
403 		return -EFAULT;
404 	if (al_hreq.buflen < sizeof(struct attrlist) ||
405 	    al_hreq.buflen > XFS_XATTR_LIST_MAX)
406 		return -EINVAL;
407 
408 	/*
409 	 * Reject flags, only allow namespaces.
410 	 */
411 	if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
412 		return -EINVAL;
413 
414 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
415 	if (IS_ERR(dentry))
416 		return PTR_ERR(dentry);
417 
418 	kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
419 	if (!kbuf)
420 		goto out_dput;
421 
422 	cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
423 	error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
424 					al_hreq.flags, cursor);
425 	if (error)
426 		goto out_kfree;
427 
428 	if (copy_to_user(&p->pos, cursor, sizeof(attrlist_cursor_kern_t))) {
429 		error = -EFAULT;
430 		goto out_kfree;
431 	}
432 
433 	if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
434 		error = -EFAULT;
435 
436 out_kfree:
437 	kmem_free(kbuf);
438 out_dput:
439 	dput(dentry);
440 	return error;
441 }
442 
443 int
444 xfs_attrmulti_attr_get(
445 	struct inode		*inode,
446 	unsigned char		*name,
447 	unsigned char		__user *ubuf,
448 	uint32_t		*len,
449 	uint32_t		flags)
450 {
451 	unsigned char		*kbuf;
452 	int			error = -EFAULT;
453 
454 	if (*len > XFS_XATTR_SIZE_MAX)
455 		return -EINVAL;
456 	kbuf = kmem_zalloc_large(*len, KM_SLEEP);
457 	if (!kbuf)
458 		return -ENOMEM;
459 
460 	error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
461 	if (error)
462 		goto out_kfree;
463 
464 	if (copy_to_user(ubuf, kbuf, *len))
465 		error = -EFAULT;
466 
467 out_kfree:
468 	kmem_free(kbuf);
469 	return error;
470 }
471 
472 int
473 xfs_attrmulti_attr_set(
474 	struct inode		*inode,
475 	unsigned char		*name,
476 	const unsigned char	__user *ubuf,
477 	uint32_t		len,
478 	uint32_t		flags)
479 {
480 	unsigned char		*kbuf;
481 	int			error;
482 
483 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
484 		return -EPERM;
485 	if (len > XFS_XATTR_SIZE_MAX)
486 		return -EINVAL;
487 
488 	kbuf = memdup_user(ubuf, len);
489 	if (IS_ERR(kbuf))
490 		return PTR_ERR(kbuf);
491 
492 	error = xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
493 	if (!error)
494 		xfs_forget_acl(inode, name, flags);
495 	kfree(kbuf);
496 	return error;
497 }
498 
499 int
500 xfs_attrmulti_attr_remove(
501 	struct inode		*inode,
502 	unsigned char		*name,
503 	uint32_t		flags)
504 {
505 	int			error;
506 
507 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
508 		return -EPERM;
509 	error = xfs_attr_remove(XFS_I(inode), name, flags);
510 	if (!error)
511 		xfs_forget_acl(inode, name, flags);
512 	return error;
513 }
514 
515 STATIC int
516 xfs_attrmulti_by_handle(
517 	struct file		*parfilp,
518 	void			__user *arg)
519 {
520 	int			error;
521 	xfs_attr_multiop_t	*ops;
522 	xfs_fsop_attrmulti_handlereq_t am_hreq;
523 	struct dentry		*dentry;
524 	unsigned int		i, size;
525 	unsigned char		*attr_name;
526 
527 	if (!capable(CAP_SYS_ADMIN))
528 		return -EPERM;
529 	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
530 		return -EFAULT;
531 
532 	/* overflow check */
533 	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
534 		return -E2BIG;
535 
536 	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
537 	if (IS_ERR(dentry))
538 		return PTR_ERR(dentry);
539 
540 	error = -E2BIG;
541 	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
542 	if (!size || size > 16 * PAGE_SIZE)
543 		goto out_dput;
544 
545 	ops = memdup_user(am_hreq.ops, size);
546 	if (IS_ERR(ops)) {
547 		error = PTR_ERR(ops);
548 		goto out_dput;
549 	}
550 
551 	error = -ENOMEM;
552 	attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
553 	if (!attr_name)
554 		goto out_kfree_ops;
555 
556 	error = 0;
557 	for (i = 0; i < am_hreq.opcount; i++) {
558 		ops[i].am_error = strncpy_from_user((char *)attr_name,
559 				ops[i].am_attrname, MAXNAMELEN);
560 		if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
561 			error = -ERANGE;
562 		if (ops[i].am_error < 0)
563 			break;
564 
565 		switch (ops[i].am_opcode) {
566 		case ATTR_OP_GET:
567 			ops[i].am_error = xfs_attrmulti_attr_get(
568 					d_inode(dentry), attr_name,
569 					ops[i].am_attrvalue, &ops[i].am_length,
570 					ops[i].am_flags);
571 			break;
572 		case ATTR_OP_SET:
573 			ops[i].am_error = mnt_want_write_file(parfilp);
574 			if (ops[i].am_error)
575 				break;
576 			ops[i].am_error = xfs_attrmulti_attr_set(
577 					d_inode(dentry), attr_name,
578 					ops[i].am_attrvalue, ops[i].am_length,
579 					ops[i].am_flags);
580 			mnt_drop_write_file(parfilp);
581 			break;
582 		case ATTR_OP_REMOVE:
583 			ops[i].am_error = mnt_want_write_file(parfilp);
584 			if (ops[i].am_error)
585 				break;
586 			ops[i].am_error = xfs_attrmulti_attr_remove(
587 					d_inode(dentry), attr_name,
588 					ops[i].am_flags);
589 			mnt_drop_write_file(parfilp);
590 			break;
591 		default:
592 			ops[i].am_error = -EINVAL;
593 		}
594 	}
595 
596 	if (copy_to_user(am_hreq.ops, ops, size))
597 		error = -EFAULT;
598 
599 	kfree(attr_name);
600  out_kfree_ops:
601 	kfree(ops);
602  out_dput:
603 	dput(dentry);
604 	return error;
605 }
606 
607 int
608 xfs_ioc_space(
609 	struct file		*filp,
610 	unsigned int		cmd,
611 	xfs_flock64_t		*bf)
612 {
613 	struct inode		*inode = file_inode(filp);
614 	struct xfs_inode	*ip = XFS_I(inode);
615 	struct iattr		iattr;
616 	enum xfs_prealloc_flags	flags = 0;
617 	uint			iolock = XFS_IOLOCK_EXCL;
618 	int			error;
619 
620 	/*
621 	 * Only allow the sys admin to reserve space unless
622 	 * unwritten extents are enabled.
623 	 */
624 	if (!xfs_sb_version_hasextflgbit(&ip->i_mount->m_sb) &&
625 	    !capable(CAP_SYS_ADMIN))
626 		return -EPERM;
627 
628 	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
629 		return -EPERM;
630 
631 	if (!(filp->f_mode & FMODE_WRITE))
632 		return -EBADF;
633 
634 	if (!S_ISREG(inode->i_mode))
635 		return -EINVAL;
636 
637 	if (filp->f_flags & O_DSYNC)
638 		flags |= XFS_PREALLOC_SYNC;
639 	if (filp->f_mode & FMODE_NOCMTIME)
640 		flags |= XFS_PREALLOC_INVISIBLE;
641 
642 	error = mnt_want_write_file(filp);
643 	if (error)
644 		return error;
645 
646 	xfs_ilock(ip, iolock);
647 	error = xfs_break_layouts(inode, &iolock);
648 	if (error)
649 		goto out_unlock;
650 
651 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
652 	iolock |= XFS_MMAPLOCK_EXCL;
653 
654 	switch (bf->l_whence) {
655 	case 0: /*SEEK_SET*/
656 		break;
657 	case 1: /*SEEK_CUR*/
658 		bf->l_start += filp->f_pos;
659 		break;
660 	case 2: /*SEEK_END*/
661 		bf->l_start += XFS_ISIZE(ip);
662 		break;
663 	default:
664 		error = -EINVAL;
665 		goto out_unlock;
666 	}
667 
668 	/*
669 	 * length of <= 0 for resv/unresv/zero is invalid.  length for
670 	 * alloc/free is ignored completely and we have no idea what userspace
671 	 * might have set it to, so set it to zero to allow range
672 	 * checks to pass.
673 	 */
674 	switch (cmd) {
675 	case XFS_IOC_ZERO_RANGE:
676 	case XFS_IOC_RESVSP:
677 	case XFS_IOC_RESVSP64:
678 	case XFS_IOC_UNRESVSP:
679 	case XFS_IOC_UNRESVSP64:
680 		if (bf->l_len <= 0) {
681 			error = -EINVAL;
682 			goto out_unlock;
683 		}
684 		break;
685 	default:
686 		bf->l_len = 0;
687 		break;
688 	}
689 
690 	if (bf->l_start < 0 ||
691 	    bf->l_start > inode->i_sb->s_maxbytes ||
692 	    bf->l_start + bf->l_len < 0 ||
693 	    bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
694 		error = -EINVAL;
695 		goto out_unlock;
696 	}
697 
698 	switch (cmd) {
699 	case XFS_IOC_ZERO_RANGE:
700 		flags |= XFS_PREALLOC_SET;
701 		error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
702 		break;
703 	case XFS_IOC_RESVSP:
704 	case XFS_IOC_RESVSP64:
705 		flags |= XFS_PREALLOC_SET;
706 		error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
707 						XFS_BMAPI_PREALLOC);
708 		break;
709 	case XFS_IOC_UNRESVSP:
710 	case XFS_IOC_UNRESVSP64:
711 		error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
712 		break;
713 	case XFS_IOC_ALLOCSP:
714 	case XFS_IOC_ALLOCSP64:
715 	case XFS_IOC_FREESP:
716 	case XFS_IOC_FREESP64:
717 		flags |= XFS_PREALLOC_CLEAR;
718 		if (bf->l_start > XFS_ISIZE(ip)) {
719 			error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
720 					bf->l_start - XFS_ISIZE(ip), 0);
721 			if (error)
722 				goto out_unlock;
723 		}
724 
725 		iattr.ia_valid = ATTR_SIZE;
726 		iattr.ia_size = bf->l_start;
727 
728 		error = xfs_vn_setattr_size(file_dentry(filp), &iattr);
729 		break;
730 	default:
731 		ASSERT(0);
732 		error = -EINVAL;
733 	}
734 
735 	if (error)
736 		goto out_unlock;
737 
738 	error = xfs_update_prealloc_flags(ip, flags);
739 
740 out_unlock:
741 	xfs_iunlock(ip, iolock);
742 	mnt_drop_write_file(filp);
743 	return error;
744 }
745 
746 STATIC int
747 xfs_ioc_bulkstat(
748 	xfs_mount_t		*mp,
749 	unsigned int		cmd,
750 	void			__user *arg)
751 {
752 	xfs_fsop_bulkreq_t	bulkreq;
753 	int			count;	/* # of records returned */
754 	xfs_ino_t		inlast;	/* last inode number */
755 	int			done;
756 	int			error;
757 
758 	/* done = 1 if there are more stats to get and if bulkstat */
759 	/* should be called again (unused here, but used in dmapi) */
760 
761 	if (!capable(CAP_SYS_ADMIN))
762 		return -EPERM;
763 
764 	if (XFS_FORCED_SHUTDOWN(mp))
765 		return -EIO;
766 
767 	if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
768 		return -EFAULT;
769 
770 	if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
771 		return -EFAULT;
772 
773 	if ((count = bulkreq.icount) <= 0)
774 		return -EINVAL;
775 
776 	if (bulkreq.ubuffer == NULL)
777 		return -EINVAL;
778 
779 	if (cmd == XFS_IOC_FSINUMBERS)
780 		error = xfs_inumbers(mp, &inlast, &count,
781 					bulkreq.ubuffer, xfs_inumbers_fmt);
782 	else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
783 		error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
784 					sizeof(xfs_bstat_t), NULL, &done);
785 	else	/* XFS_IOC_FSBULKSTAT */
786 		error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
787 				     sizeof(xfs_bstat_t), bulkreq.ubuffer,
788 				     &done);
789 
790 	if (error)
791 		return error;
792 
793 	if (bulkreq.ocount != NULL) {
794 		if (copy_to_user(bulkreq.lastip, &inlast,
795 						sizeof(xfs_ino_t)))
796 			return -EFAULT;
797 
798 		if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
799 			return -EFAULT;
800 	}
801 
802 	return 0;
803 }
804 
805 STATIC int
806 xfs_ioc_fsgeometry_v1(
807 	xfs_mount_t		*mp,
808 	void			__user *arg)
809 {
810 	xfs_fsop_geom_t         fsgeo;
811 	int			error;
812 
813 	error = xfs_fs_geometry(&mp->m_sb, &fsgeo, 3);
814 	if (error)
815 		return error;
816 
817 	/*
818 	 * Caller should have passed an argument of type
819 	 * xfs_fsop_geom_v1_t.  This is a proper subset of the
820 	 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
821 	 */
822 	if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
823 		return -EFAULT;
824 	return 0;
825 }
826 
827 STATIC int
828 xfs_ioc_fsgeometry(
829 	xfs_mount_t		*mp,
830 	void			__user *arg)
831 {
832 	xfs_fsop_geom_t		fsgeo;
833 	int			error;
834 
835 	error = xfs_fs_geometry(&mp->m_sb, &fsgeo, 4);
836 	if (error)
837 		return error;
838 
839 	if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
840 		return -EFAULT;
841 	return 0;
842 }
843 
844 /*
845  * Linux extended inode flags interface.
846  */
847 
848 STATIC unsigned int
849 xfs_merge_ioc_xflags(
850 	unsigned int	flags,
851 	unsigned int	start)
852 {
853 	unsigned int	xflags = start;
854 
855 	if (flags & FS_IMMUTABLE_FL)
856 		xflags |= FS_XFLAG_IMMUTABLE;
857 	else
858 		xflags &= ~FS_XFLAG_IMMUTABLE;
859 	if (flags & FS_APPEND_FL)
860 		xflags |= FS_XFLAG_APPEND;
861 	else
862 		xflags &= ~FS_XFLAG_APPEND;
863 	if (flags & FS_SYNC_FL)
864 		xflags |= FS_XFLAG_SYNC;
865 	else
866 		xflags &= ~FS_XFLAG_SYNC;
867 	if (flags & FS_NOATIME_FL)
868 		xflags |= FS_XFLAG_NOATIME;
869 	else
870 		xflags &= ~FS_XFLAG_NOATIME;
871 	if (flags & FS_NODUMP_FL)
872 		xflags |= FS_XFLAG_NODUMP;
873 	else
874 		xflags &= ~FS_XFLAG_NODUMP;
875 
876 	return xflags;
877 }
878 
879 STATIC unsigned int
880 xfs_di2lxflags(
881 	uint16_t	di_flags)
882 {
883 	unsigned int	flags = 0;
884 
885 	if (di_flags & XFS_DIFLAG_IMMUTABLE)
886 		flags |= FS_IMMUTABLE_FL;
887 	if (di_flags & XFS_DIFLAG_APPEND)
888 		flags |= FS_APPEND_FL;
889 	if (di_flags & XFS_DIFLAG_SYNC)
890 		flags |= FS_SYNC_FL;
891 	if (di_flags & XFS_DIFLAG_NOATIME)
892 		flags |= FS_NOATIME_FL;
893 	if (di_flags & XFS_DIFLAG_NODUMP)
894 		flags |= FS_NODUMP_FL;
895 	return flags;
896 }
897 
898 STATIC int
899 xfs_ioc_fsgetxattr(
900 	xfs_inode_t		*ip,
901 	int			attr,
902 	void			__user *arg)
903 {
904 	struct fsxattr		fa;
905 
906 	memset(&fa, 0, sizeof(struct fsxattr));
907 
908 	xfs_ilock(ip, XFS_ILOCK_SHARED);
909 	fa.fsx_xflags = xfs_ip2xflags(ip);
910 	fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
911 	fa.fsx_cowextsize = ip->i_d.di_cowextsize <<
912 			ip->i_mount->m_sb.sb_blocklog;
913 	fa.fsx_projid = xfs_get_projid(ip);
914 
915 	if (attr) {
916 		if (ip->i_afp) {
917 			if (ip->i_afp->if_flags & XFS_IFEXTENTS)
918 				fa.fsx_nextents = xfs_iext_count(ip->i_afp);
919 			else
920 				fa.fsx_nextents = ip->i_d.di_anextents;
921 		} else
922 			fa.fsx_nextents = 0;
923 	} else {
924 		if (ip->i_df.if_flags & XFS_IFEXTENTS)
925 			fa.fsx_nextents = xfs_iext_count(&ip->i_df);
926 		else
927 			fa.fsx_nextents = ip->i_d.di_nextents;
928 	}
929 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
930 
931 	if (copy_to_user(arg, &fa, sizeof(fa)))
932 		return -EFAULT;
933 	return 0;
934 }
935 
936 STATIC uint16_t
937 xfs_flags2diflags(
938 	struct xfs_inode	*ip,
939 	unsigned int		xflags)
940 {
941 	/* can't set PREALLOC this way, just preserve it */
942 	uint16_t		di_flags =
943 		(ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
944 
945 	if (xflags & FS_XFLAG_IMMUTABLE)
946 		di_flags |= XFS_DIFLAG_IMMUTABLE;
947 	if (xflags & FS_XFLAG_APPEND)
948 		di_flags |= XFS_DIFLAG_APPEND;
949 	if (xflags & FS_XFLAG_SYNC)
950 		di_flags |= XFS_DIFLAG_SYNC;
951 	if (xflags & FS_XFLAG_NOATIME)
952 		di_flags |= XFS_DIFLAG_NOATIME;
953 	if (xflags & FS_XFLAG_NODUMP)
954 		di_flags |= XFS_DIFLAG_NODUMP;
955 	if (xflags & FS_XFLAG_NODEFRAG)
956 		di_flags |= XFS_DIFLAG_NODEFRAG;
957 	if (xflags & FS_XFLAG_FILESTREAM)
958 		di_flags |= XFS_DIFLAG_FILESTREAM;
959 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
960 		if (xflags & FS_XFLAG_RTINHERIT)
961 			di_flags |= XFS_DIFLAG_RTINHERIT;
962 		if (xflags & FS_XFLAG_NOSYMLINKS)
963 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
964 		if (xflags & FS_XFLAG_EXTSZINHERIT)
965 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
966 		if (xflags & FS_XFLAG_PROJINHERIT)
967 			di_flags |= XFS_DIFLAG_PROJINHERIT;
968 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
969 		if (xflags & FS_XFLAG_REALTIME)
970 			di_flags |= XFS_DIFLAG_REALTIME;
971 		if (xflags & FS_XFLAG_EXTSIZE)
972 			di_flags |= XFS_DIFLAG_EXTSIZE;
973 	}
974 
975 	return di_flags;
976 }
977 
978 STATIC uint64_t
979 xfs_flags2diflags2(
980 	struct xfs_inode	*ip,
981 	unsigned int		xflags)
982 {
983 	uint64_t		di_flags2 =
984 		(ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK);
985 
986 	if (xflags & FS_XFLAG_DAX)
987 		di_flags2 |= XFS_DIFLAG2_DAX;
988 	if (xflags & FS_XFLAG_COWEXTSIZE)
989 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
990 
991 	return di_flags2;
992 }
993 
994 STATIC void
995 xfs_diflags_to_linux(
996 	struct xfs_inode	*ip)
997 {
998 	struct inode		*inode = VFS_I(ip);
999 	unsigned int		xflags = xfs_ip2xflags(ip);
1000 
1001 	if (xflags & FS_XFLAG_IMMUTABLE)
1002 		inode->i_flags |= S_IMMUTABLE;
1003 	else
1004 		inode->i_flags &= ~S_IMMUTABLE;
1005 	if (xflags & FS_XFLAG_APPEND)
1006 		inode->i_flags |= S_APPEND;
1007 	else
1008 		inode->i_flags &= ~S_APPEND;
1009 	if (xflags & FS_XFLAG_SYNC)
1010 		inode->i_flags |= S_SYNC;
1011 	else
1012 		inode->i_flags &= ~S_SYNC;
1013 	if (xflags & FS_XFLAG_NOATIME)
1014 		inode->i_flags |= S_NOATIME;
1015 	else
1016 		inode->i_flags &= ~S_NOATIME;
1017 #if 0	/* disabled until the flag switching races are sorted out */
1018 	if (xflags & FS_XFLAG_DAX)
1019 		inode->i_flags |= S_DAX;
1020 	else
1021 		inode->i_flags &= ~S_DAX;
1022 #endif
1023 }
1024 
1025 static int
1026 xfs_ioctl_setattr_xflags(
1027 	struct xfs_trans	*tp,
1028 	struct xfs_inode	*ip,
1029 	struct fsxattr		*fa)
1030 {
1031 	struct xfs_mount	*mp = ip->i_mount;
1032 	uint64_t		di_flags2;
1033 
1034 	/* Can't change realtime flag if any extents are allocated. */
1035 	if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1036 	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1037 		return -EINVAL;
1038 
1039 	/* If realtime flag is set then must have realtime device */
1040 	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1041 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1042 		    (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1043 			return -EINVAL;
1044 	}
1045 
1046 	/* Clear reflink if we are actually able to set the rt flag. */
1047 	if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1048 		ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
1049 
1050 	/* Don't allow us to set DAX mode for a reflinked file for now. */
1051 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1052 		return -EINVAL;
1053 
1054 	/*
1055 	 * Can't modify an immutable/append-only file unless
1056 	 * we have appropriate permission.
1057 	 */
1058 	if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1059 	     (fa->fsx_xflags & (FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))) &&
1060 	    !capable(CAP_LINUX_IMMUTABLE))
1061 		return -EPERM;
1062 
1063 	/* diflags2 only valid for v3 inodes. */
1064 	di_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1065 	if (di_flags2 && ip->i_d.di_version < 3)
1066 		return -EINVAL;
1067 
1068 	ip->i_d.di_flags = xfs_flags2diflags(ip, fa->fsx_xflags);
1069 	ip->i_d.di_flags2 = di_flags2;
1070 
1071 	xfs_diflags_to_linux(ip);
1072 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1073 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1074 	XFS_STATS_INC(mp, xs_ig_attrchg);
1075 	return 0;
1076 }
1077 
1078 /*
1079  * If we are changing DAX flags, we have to ensure the file is clean and any
1080  * cached objects in the address space are invalidated and removed. This
1081  * requires us to lock out other IO and page faults similar to a truncate
1082  * operation. The locks need to be held until the transaction has been committed
1083  * so that the cache invalidation is atomic with respect to the DAX flag
1084  * manipulation.
1085  */
1086 static int
1087 xfs_ioctl_setattr_dax_invalidate(
1088 	struct xfs_inode	*ip,
1089 	struct fsxattr		*fa,
1090 	int			*join_flags)
1091 {
1092 	struct inode		*inode = VFS_I(ip);
1093 	struct super_block	*sb = inode->i_sb;
1094 	int			error;
1095 
1096 	*join_flags = 0;
1097 
1098 	/*
1099 	 * It is only valid to set the DAX flag on regular files and
1100 	 * directories on filesystems where the block size is equal to the page
1101 	 * size. On directories it serves as an inherit hint.
1102 	 */
1103 	if (fa->fsx_xflags & FS_XFLAG_DAX) {
1104 		if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1105 			return -EINVAL;
1106 		if (bdev_dax_supported(sb, sb->s_blocksize) < 0)
1107 			return -EINVAL;
1108 	}
1109 
1110 	/* If the DAX state is not changing, we have nothing to do here. */
1111 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
1112 		return 0;
1113 	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
1114 		return 0;
1115 
1116 	/* lock, flush and invalidate mapping in preparation for flag change */
1117 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1118 	error = filemap_write_and_wait(inode->i_mapping);
1119 	if (error)
1120 		goto out_unlock;
1121 	error = invalidate_inode_pages2(inode->i_mapping);
1122 	if (error)
1123 		goto out_unlock;
1124 
1125 	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
1126 	return 0;
1127 
1128 out_unlock:
1129 	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1130 	return error;
1131 
1132 }
1133 
1134 /*
1135  * Set up the transaction structure for the setattr operation, checking that we
1136  * have permission to do so. On success, return a clean transaction and the
1137  * inode locked exclusively ready for further operation specific checks. On
1138  * failure, return an error without modifying or locking the inode.
1139  *
1140  * The inode might already be IO locked on call. If this is the case, it is
1141  * indicated in @join_flags and we take full responsibility for ensuring they
1142  * are unlocked from now on. Hence if we have an error here, we still have to
1143  * unlock them. Otherwise, once they are joined to the transaction, they will
1144  * be unlocked on commit/cancel.
1145  */
1146 static struct xfs_trans *
1147 xfs_ioctl_setattr_get_trans(
1148 	struct xfs_inode	*ip,
1149 	int			join_flags)
1150 {
1151 	struct xfs_mount	*mp = ip->i_mount;
1152 	struct xfs_trans	*tp;
1153 	int			error = -EROFS;
1154 
1155 	if (mp->m_flags & XFS_MOUNT_RDONLY)
1156 		goto out_unlock;
1157 	error = -EIO;
1158 	if (XFS_FORCED_SHUTDOWN(mp))
1159 		goto out_unlock;
1160 
1161 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
1162 	if (error)
1163 		return ERR_PTR(error);
1164 
1165 	xfs_ilock(ip, XFS_ILOCK_EXCL);
1166 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
1167 	join_flags = 0;
1168 
1169 	/*
1170 	 * CAP_FOWNER overrides the following restrictions:
1171 	 *
1172 	 * The user ID of the calling process must be equal to the file owner
1173 	 * ID, except in cases where the CAP_FSETID capability is applicable.
1174 	 */
1175 	if (!inode_owner_or_capable(VFS_I(ip))) {
1176 		error = -EPERM;
1177 		goto out_cancel;
1178 	}
1179 
1180 	if (mp->m_flags & XFS_MOUNT_WSYNC)
1181 		xfs_trans_set_sync(tp);
1182 
1183 	return tp;
1184 
1185 out_cancel:
1186 	xfs_trans_cancel(tp);
1187 out_unlock:
1188 	if (join_flags)
1189 		xfs_iunlock(ip, join_flags);
1190 	return ERR_PTR(error);
1191 }
1192 
1193 /*
1194  * extent size hint validation is somewhat cumbersome. Rules are:
1195  *
1196  * 1. extent size hint is only valid for directories and regular files
1197  * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1198  * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1199  * 4. can only be changed on regular files if no extents are allocated
1200  * 5. can be changed on directories at any time
1201  * 6. extsize hint of 0 turns off hints, clears inode flags.
1202  * 7. Extent size must be a multiple of the appropriate block size.
1203  * 8. for non-realtime files, the extent size hint must be limited
1204  *    to half the AG size to avoid alignment extending the extent beyond the
1205  *    limits of the AG.
1206  *
1207  * Please keep this function in sync with xfs_scrub_inode_extsize.
1208  */
1209 static int
1210 xfs_ioctl_setattr_check_extsize(
1211 	struct xfs_inode	*ip,
1212 	struct fsxattr		*fa)
1213 {
1214 	struct xfs_mount	*mp = ip->i_mount;
1215 
1216 	if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(VFS_I(ip)->i_mode))
1217 		return -EINVAL;
1218 
1219 	if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
1220 	    !S_ISDIR(VFS_I(ip)->i_mode))
1221 		return -EINVAL;
1222 
1223 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_d.di_nextents &&
1224 	    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1225 		return -EINVAL;
1226 
1227 	if (fa->fsx_extsize != 0) {
1228 		xfs_extlen_t    size;
1229 		xfs_fsblock_t   extsize_fsb;
1230 
1231 		extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1232 		if (extsize_fsb > MAXEXTLEN)
1233 			return -EINVAL;
1234 
1235 		if (XFS_IS_REALTIME_INODE(ip) ||
1236 		    (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1237 			size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1238 		} else {
1239 			size = mp->m_sb.sb_blocksize;
1240 			if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1241 				return -EINVAL;
1242 		}
1243 
1244 		if (fa->fsx_extsize % size)
1245 			return -EINVAL;
1246 	} else
1247 		fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
1248 
1249 	return 0;
1250 }
1251 
1252 /*
1253  * CoW extent size hint validation rules are:
1254  *
1255  * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
1256  *    The inode does not have to have any shared blocks, but it must be a v3.
1257  * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
1258  *    for a directory, the hint is propagated to new files.
1259  * 3. Can be changed on files & directories at any time.
1260  * 4. CoW extsize hint of 0 turns off hints, clears inode flags.
1261  * 5. Extent size must be a multiple of the appropriate block size.
1262  * 6. The extent size hint must be limited to half the AG size to avoid
1263  *    alignment extending the extent beyond the limits of the AG.
1264  *
1265  * Please keep this function in sync with xfs_scrub_inode_cowextsize.
1266  */
1267 static int
1268 xfs_ioctl_setattr_check_cowextsize(
1269 	struct xfs_inode	*ip,
1270 	struct fsxattr		*fa)
1271 {
1272 	struct xfs_mount	*mp = ip->i_mount;
1273 
1274 	if (!(fa->fsx_xflags & FS_XFLAG_COWEXTSIZE))
1275 		return 0;
1276 
1277 	if (!xfs_sb_version_hasreflink(&ip->i_mount->m_sb) ||
1278 	    ip->i_d.di_version != 3)
1279 		return -EINVAL;
1280 
1281 	if (!S_ISREG(VFS_I(ip)->i_mode) && !S_ISDIR(VFS_I(ip)->i_mode))
1282 		return -EINVAL;
1283 
1284 	if (fa->fsx_cowextsize != 0) {
1285 		xfs_extlen_t    size;
1286 		xfs_fsblock_t   cowextsize_fsb;
1287 
1288 		cowextsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1289 		if (cowextsize_fsb > MAXEXTLEN)
1290 			return -EINVAL;
1291 
1292 		size = mp->m_sb.sb_blocksize;
1293 		if (cowextsize_fsb > mp->m_sb.sb_agblocks / 2)
1294 			return -EINVAL;
1295 
1296 		if (fa->fsx_cowextsize % size)
1297 			return -EINVAL;
1298 	} else
1299 		fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
1300 
1301 	return 0;
1302 }
1303 
1304 static int
1305 xfs_ioctl_setattr_check_projid(
1306 	struct xfs_inode	*ip,
1307 	struct fsxattr		*fa)
1308 {
1309 	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1310 	if (fa->fsx_projid > (uint16_t)-1 &&
1311 	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1312 		return -EINVAL;
1313 
1314 	/*
1315 	 * Project Quota ID state is only allowed to change from within the init
1316 	 * namespace. Enforce that restriction only if we are trying to change
1317 	 * the quota ID state. Everything else is allowed in user namespaces.
1318 	 */
1319 	if (current_user_ns() == &init_user_ns)
1320 		return 0;
1321 
1322 	if (xfs_get_projid(ip) != fa->fsx_projid)
1323 		return -EINVAL;
1324 	if ((fa->fsx_xflags & FS_XFLAG_PROJINHERIT) !=
1325 	    (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1326 		return -EINVAL;
1327 
1328 	return 0;
1329 }
1330 
1331 STATIC int
1332 xfs_ioctl_setattr(
1333 	xfs_inode_t		*ip,
1334 	struct fsxattr		*fa)
1335 {
1336 	struct xfs_mount	*mp = ip->i_mount;
1337 	struct xfs_trans	*tp;
1338 	struct xfs_dquot	*udqp = NULL;
1339 	struct xfs_dquot	*pdqp = NULL;
1340 	struct xfs_dquot	*olddquot = NULL;
1341 	int			code;
1342 	int			join_flags = 0;
1343 
1344 	trace_xfs_ioctl_setattr(ip);
1345 
1346 	code = xfs_ioctl_setattr_check_projid(ip, fa);
1347 	if (code)
1348 		return code;
1349 
1350 	/*
1351 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1352 	 * before we start any other transactions. Trying to do this later
1353 	 * is messy. We don't care to take a readlock to look at the ids
1354 	 * in inode here, because we can't hold it across the trans_reserve.
1355 	 * If the IDs do change before we take the ilock, we're covered
1356 	 * because the i_*dquot fields will get updated anyway.
1357 	 */
1358 	if (XFS_IS_QUOTA_ON(mp)) {
1359 		code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1360 					 ip->i_d.di_gid, fa->fsx_projid,
1361 					 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1362 		if (code)
1363 			return code;
1364 	}
1365 
1366 	/*
1367 	 * Changing DAX config may require inode locking for mapping
1368 	 * invalidation. These need to be held all the way to transaction commit
1369 	 * or cancel time, so need to be passed through to
1370 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1371 	 * appropriately.
1372 	 */
1373 	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
1374 	if (code)
1375 		goto error_free_dquots;
1376 
1377 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1378 	if (IS_ERR(tp)) {
1379 		code = PTR_ERR(tp);
1380 		goto error_free_dquots;
1381 	}
1382 
1383 
1384 	if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1385 	    xfs_get_projid(ip) != fa->fsx_projid) {
1386 		code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1387 				capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
1388 		if (code)	/* out of quota */
1389 			goto error_trans_cancel;
1390 	}
1391 
1392 	code = xfs_ioctl_setattr_check_extsize(ip, fa);
1393 	if (code)
1394 		goto error_trans_cancel;
1395 
1396 	code = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1397 	if (code)
1398 		goto error_trans_cancel;
1399 
1400 	code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1401 	if (code)
1402 		goto error_trans_cancel;
1403 
1404 	/*
1405 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1406 	 * overrides the following restrictions:
1407 	 *
1408 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1409 	 * successful return from chown()
1410 	 */
1411 
1412 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1413 	    !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1414 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1415 
1416 	/* Change the ownerships and register project quota modifications */
1417 	if (xfs_get_projid(ip) != fa->fsx_projid) {
1418 		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1419 			olddquot = xfs_qm_vop_chown(tp, ip,
1420 						&ip->i_pdquot, pdqp);
1421 		}
1422 		ASSERT(ip->i_d.di_version > 1);
1423 		xfs_set_projid(ip, fa->fsx_projid);
1424 	}
1425 
1426 	/*
1427 	 * Only set the extent size hint if we've already determined that the
1428 	 * extent size hint should be set on the inode. If no extent size flags
1429 	 * are set on the inode then unconditionally clear the extent size hint.
1430 	 */
1431 	if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1432 		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1433 	else
1434 		ip->i_d.di_extsize = 0;
1435 	if (ip->i_d.di_version == 3 &&
1436 	    (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
1437 		ip->i_d.di_cowextsize = fa->fsx_cowextsize >>
1438 				mp->m_sb.sb_blocklog;
1439 	else
1440 		ip->i_d.di_cowextsize = 0;
1441 
1442 	code = xfs_trans_commit(tp);
1443 
1444 	/*
1445 	 * Release any dquot(s) the inode had kept before chown.
1446 	 */
1447 	xfs_qm_dqrele(olddquot);
1448 	xfs_qm_dqrele(udqp);
1449 	xfs_qm_dqrele(pdqp);
1450 
1451 	return code;
1452 
1453 error_trans_cancel:
1454 	xfs_trans_cancel(tp);
1455 error_free_dquots:
1456 	xfs_qm_dqrele(udqp);
1457 	xfs_qm_dqrele(pdqp);
1458 	return code;
1459 }
1460 
1461 STATIC int
1462 xfs_ioc_fssetxattr(
1463 	xfs_inode_t		*ip,
1464 	struct file		*filp,
1465 	void			__user *arg)
1466 {
1467 	struct fsxattr		fa;
1468 	int error;
1469 
1470 	if (copy_from_user(&fa, arg, sizeof(fa)))
1471 		return -EFAULT;
1472 
1473 	error = mnt_want_write_file(filp);
1474 	if (error)
1475 		return error;
1476 	error = xfs_ioctl_setattr(ip, &fa);
1477 	mnt_drop_write_file(filp);
1478 	return error;
1479 }
1480 
1481 STATIC int
1482 xfs_ioc_getxflags(
1483 	xfs_inode_t		*ip,
1484 	void			__user *arg)
1485 {
1486 	unsigned int		flags;
1487 
1488 	flags = xfs_di2lxflags(ip->i_d.di_flags);
1489 	if (copy_to_user(arg, &flags, sizeof(flags)))
1490 		return -EFAULT;
1491 	return 0;
1492 }
1493 
1494 STATIC int
1495 xfs_ioc_setxflags(
1496 	struct xfs_inode	*ip,
1497 	struct file		*filp,
1498 	void			__user *arg)
1499 {
1500 	struct xfs_trans	*tp;
1501 	struct fsxattr		fa;
1502 	unsigned int		flags;
1503 	int			join_flags = 0;
1504 	int			error;
1505 
1506 	if (copy_from_user(&flags, arg, sizeof(flags)))
1507 		return -EFAULT;
1508 
1509 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1510 		      FS_NOATIME_FL | FS_NODUMP_FL | \
1511 		      FS_SYNC_FL))
1512 		return -EOPNOTSUPP;
1513 
1514 	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1515 
1516 	error = mnt_want_write_file(filp);
1517 	if (error)
1518 		return error;
1519 
1520 	/*
1521 	 * Changing DAX config may require inode locking for mapping
1522 	 * invalidation. These need to be held all the way to transaction commit
1523 	 * or cancel time, so need to be passed through to
1524 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1525 	 * appropriately.
1526 	 */
1527 	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
1528 	if (error)
1529 		goto out_drop_write;
1530 
1531 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1532 	if (IS_ERR(tp)) {
1533 		error = PTR_ERR(tp);
1534 		goto out_drop_write;
1535 	}
1536 
1537 	error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1538 	if (error) {
1539 		xfs_trans_cancel(tp);
1540 		goto out_drop_write;
1541 	}
1542 
1543 	error = xfs_trans_commit(tp);
1544 out_drop_write:
1545 	mnt_drop_write_file(filp);
1546 	return error;
1547 }
1548 
1549 static bool
1550 xfs_getbmap_format(
1551 	struct kgetbmap		*p,
1552 	struct getbmapx __user	*u,
1553 	size_t			recsize)
1554 {
1555 	if (put_user(p->bmv_offset, &u->bmv_offset) ||
1556 	    put_user(p->bmv_block, &u->bmv_block) ||
1557 	    put_user(p->bmv_length, &u->bmv_length) ||
1558 	    put_user(0, &u->bmv_count) ||
1559 	    put_user(0, &u->bmv_entries))
1560 		return false;
1561 	if (recsize < sizeof(struct getbmapx))
1562 		return true;
1563 	if (put_user(0, &u->bmv_iflags) ||
1564 	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
1565 	    put_user(0, &u->bmv_unused1) ||
1566 	    put_user(0, &u->bmv_unused2))
1567 		return false;
1568 	return true;
1569 }
1570 
1571 STATIC int
1572 xfs_ioc_getbmap(
1573 	struct file		*file,
1574 	unsigned int		cmd,
1575 	void			__user *arg)
1576 {
1577 	struct getbmapx		bmx = { 0 };
1578 	struct kgetbmap		*buf;
1579 	size_t			recsize;
1580 	int			error, i;
1581 
1582 	switch (cmd) {
1583 	case XFS_IOC_GETBMAPA:
1584 		bmx.bmv_iflags = BMV_IF_ATTRFORK;
1585 		/*FALLTHRU*/
1586 	case XFS_IOC_GETBMAP:
1587 		if (file->f_mode & FMODE_NOCMTIME)
1588 			bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1589 		/* struct getbmap is a strict subset of struct getbmapx. */
1590 		recsize = sizeof(struct getbmap);
1591 		break;
1592 	case XFS_IOC_GETBMAPX:
1593 		recsize = sizeof(struct getbmapx);
1594 		break;
1595 	default:
1596 		return -EINVAL;
1597 	}
1598 
1599 	if (copy_from_user(&bmx, arg, recsize))
1600 		return -EFAULT;
1601 
1602 	if (bmx.bmv_count < 2)
1603 		return -EINVAL;
1604 	if (bmx.bmv_count > ULONG_MAX / recsize)
1605 		return -ENOMEM;
1606 
1607 	buf = kmem_zalloc_large(bmx.bmv_count * sizeof(*buf), 0);
1608 	if (!buf)
1609 		return -ENOMEM;
1610 
1611 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1612 	if (error)
1613 		goto out_free_buf;
1614 
1615 	error = -EFAULT;
1616 	if (copy_to_user(arg, &bmx, recsize))
1617 		goto out_free_buf;
1618 	arg += recsize;
1619 
1620 	for (i = 0; i < bmx.bmv_entries; i++) {
1621 		if (!xfs_getbmap_format(buf + i, arg, recsize))
1622 			goto out_free_buf;
1623 		arg += recsize;
1624 	}
1625 
1626 	error = 0;
1627 out_free_buf:
1628 	kmem_free(buf);
1629 	return 0;
1630 }
1631 
1632 struct getfsmap_info {
1633 	struct xfs_mount	*mp;
1634 	struct fsmap_head __user *data;
1635 	unsigned int		idx;
1636 	__u32			last_flags;
1637 };
1638 
1639 STATIC int
1640 xfs_getfsmap_format(struct xfs_fsmap *xfm, void *priv)
1641 {
1642 	struct getfsmap_info	*info = priv;
1643 	struct fsmap		fm;
1644 
1645 	trace_xfs_getfsmap_mapping(info->mp, xfm);
1646 
1647 	info->last_flags = xfm->fmr_flags;
1648 	xfs_fsmap_from_internal(&fm, xfm);
1649 	if (copy_to_user(&info->data->fmh_recs[info->idx++], &fm,
1650 			sizeof(struct fsmap)))
1651 		return -EFAULT;
1652 
1653 	return 0;
1654 }
1655 
1656 STATIC int
1657 xfs_ioc_getfsmap(
1658 	struct xfs_inode	*ip,
1659 	struct fsmap_head	__user *arg)
1660 {
1661 	struct getfsmap_info	info = { NULL };
1662 	struct xfs_fsmap_head	xhead = {0};
1663 	struct fsmap_head	head;
1664 	bool			aborted = false;
1665 	int			error;
1666 
1667 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1668 		return -EFAULT;
1669 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1670 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1671 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1672 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1673 		       sizeof(head.fmh_keys[1].fmr_reserved)))
1674 		return -EINVAL;
1675 
1676 	xhead.fmh_iflags = head.fmh_iflags;
1677 	xhead.fmh_count = head.fmh_count;
1678 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1679 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1680 
1681 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1682 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1683 
1684 	info.mp = ip->i_mount;
1685 	info.data = arg;
1686 	error = xfs_getfsmap(ip->i_mount, &xhead, xfs_getfsmap_format, &info);
1687 	if (error == XFS_BTREE_QUERY_RANGE_ABORT) {
1688 		error = 0;
1689 		aborted = true;
1690 	} else if (error)
1691 		return error;
1692 
1693 	/* If we didn't abort, set the "last" flag in the last fmx */
1694 	if (!aborted && info.idx) {
1695 		info.last_flags |= FMR_OF_LAST;
1696 		if (copy_to_user(&info.data->fmh_recs[info.idx - 1].fmr_flags,
1697 				&info.last_flags, sizeof(info.last_flags)))
1698 			return -EFAULT;
1699 	}
1700 
1701 	/* copy back header */
1702 	head.fmh_entries = xhead.fmh_entries;
1703 	head.fmh_oflags = xhead.fmh_oflags;
1704 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
1705 		return -EFAULT;
1706 
1707 	return 0;
1708 }
1709 
1710 STATIC int
1711 xfs_ioc_scrub_metadata(
1712 	struct xfs_inode		*ip,
1713 	void				__user *arg)
1714 {
1715 	struct xfs_scrub_metadata	scrub;
1716 	int				error;
1717 
1718 	if (!capable(CAP_SYS_ADMIN))
1719 		return -EPERM;
1720 
1721 	if (copy_from_user(&scrub, arg, sizeof(scrub)))
1722 		return -EFAULT;
1723 
1724 	error = xfs_scrub_metadata(ip, &scrub);
1725 	if (error)
1726 		return error;
1727 
1728 	if (copy_to_user(arg, &scrub, sizeof(scrub)))
1729 		return -EFAULT;
1730 
1731 	return 0;
1732 }
1733 
1734 int
1735 xfs_ioc_swapext(
1736 	xfs_swapext_t	*sxp)
1737 {
1738 	xfs_inode_t     *ip, *tip;
1739 	struct fd	f, tmp;
1740 	int		error = 0;
1741 
1742 	/* Pull information for the target fd */
1743 	f = fdget((int)sxp->sx_fdtarget);
1744 	if (!f.file) {
1745 		error = -EINVAL;
1746 		goto out;
1747 	}
1748 
1749 	if (!(f.file->f_mode & FMODE_WRITE) ||
1750 	    !(f.file->f_mode & FMODE_READ) ||
1751 	    (f.file->f_flags & O_APPEND)) {
1752 		error = -EBADF;
1753 		goto out_put_file;
1754 	}
1755 
1756 	tmp = fdget((int)sxp->sx_fdtmp);
1757 	if (!tmp.file) {
1758 		error = -EINVAL;
1759 		goto out_put_file;
1760 	}
1761 
1762 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1763 	    !(tmp.file->f_mode & FMODE_READ) ||
1764 	    (tmp.file->f_flags & O_APPEND)) {
1765 		error = -EBADF;
1766 		goto out_put_tmp_file;
1767 	}
1768 
1769 	if (IS_SWAPFILE(file_inode(f.file)) ||
1770 	    IS_SWAPFILE(file_inode(tmp.file))) {
1771 		error = -EINVAL;
1772 		goto out_put_tmp_file;
1773 	}
1774 
1775 	/*
1776 	 * We need to ensure that the fds passed in point to XFS inodes
1777 	 * before we cast and access them as XFS structures as we have no
1778 	 * control over what the user passes us here.
1779 	 */
1780 	if (f.file->f_op != &xfs_file_operations ||
1781 	    tmp.file->f_op != &xfs_file_operations) {
1782 		error = -EINVAL;
1783 		goto out_put_tmp_file;
1784 	}
1785 
1786 	ip = XFS_I(file_inode(f.file));
1787 	tip = XFS_I(file_inode(tmp.file));
1788 
1789 	if (ip->i_mount != tip->i_mount) {
1790 		error = -EINVAL;
1791 		goto out_put_tmp_file;
1792 	}
1793 
1794 	if (ip->i_ino == tip->i_ino) {
1795 		error = -EINVAL;
1796 		goto out_put_tmp_file;
1797 	}
1798 
1799 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1800 		error = -EIO;
1801 		goto out_put_tmp_file;
1802 	}
1803 
1804 	error = xfs_swap_extents(ip, tip, sxp);
1805 
1806  out_put_tmp_file:
1807 	fdput(tmp);
1808  out_put_file:
1809 	fdput(f);
1810  out:
1811 	return error;
1812 }
1813 
1814 /*
1815  * Note: some of the ioctl's return positive numbers as a
1816  * byte count indicating success, such as readlink_by_handle.
1817  * So we don't "sign flip" like most other routines.  This means
1818  * true errors need to be returned as a negative value.
1819  */
1820 long
1821 xfs_file_ioctl(
1822 	struct file		*filp,
1823 	unsigned int		cmd,
1824 	unsigned long		p)
1825 {
1826 	struct inode		*inode = file_inode(filp);
1827 	struct xfs_inode	*ip = XFS_I(inode);
1828 	struct xfs_mount	*mp = ip->i_mount;
1829 	void			__user *arg = (void __user *)p;
1830 	int			error;
1831 
1832 	trace_xfs_file_ioctl(ip);
1833 
1834 	switch (cmd) {
1835 	case FITRIM:
1836 		return xfs_ioc_trim(mp, arg);
1837 	case XFS_IOC_ALLOCSP:
1838 	case XFS_IOC_FREESP:
1839 	case XFS_IOC_RESVSP:
1840 	case XFS_IOC_UNRESVSP:
1841 	case XFS_IOC_ALLOCSP64:
1842 	case XFS_IOC_FREESP64:
1843 	case XFS_IOC_RESVSP64:
1844 	case XFS_IOC_UNRESVSP64:
1845 	case XFS_IOC_ZERO_RANGE: {
1846 		xfs_flock64_t		bf;
1847 
1848 		if (copy_from_user(&bf, arg, sizeof(bf)))
1849 			return -EFAULT;
1850 		return xfs_ioc_space(filp, cmd, &bf);
1851 	}
1852 	case XFS_IOC_DIOINFO: {
1853 		struct dioattr	da;
1854 		xfs_buftarg_t	*target =
1855 			XFS_IS_REALTIME_INODE(ip) ?
1856 			mp->m_rtdev_targp : mp->m_ddev_targp;
1857 
1858 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1859 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1860 
1861 		if (copy_to_user(arg, &da, sizeof(da)))
1862 			return -EFAULT;
1863 		return 0;
1864 	}
1865 
1866 	case XFS_IOC_FSBULKSTAT_SINGLE:
1867 	case XFS_IOC_FSBULKSTAT:
1868 	case XFS_IOC_FSINUMBERS:
1869 		return xfs_ioc_bulkstat(mp, cmd, arg);
1870 
1871 	case XFS_IOC_FSGEOMETRY_V1:
1872 		return xfs_ioc_fsgeometry_v1(mp, arg);
1873 
1874 	case XFS_IOC_FSGEOMETRY:
1875 		return xfs_ioc_fsgeometry(mp, arg);
1876 
1877 	case XFS_IOC_GETVERSION:
1878 		return put_user(inode->i_generation, (int __user *)arg);
1879 
1880 	case XFS_IOC_FSGETXATTR:
1881 		return xfs_ioc_fsgetxattr(ip, 0, arg);
1882 	case XFS_IOC_FSGETXATTRA:
1883 		return xfs_ioc_fsgetxattr(ip, 1, arg);
1884 	case XFS_IOC_FSSETXATTR:
1885 		return xfs_ioc_fssetxattr(ip, filp, arg);
1886 	case XFS_IOC_GETXFLAGS:
1887 		return xfs_ioc_getxflags(ip, arg);
1888 	case XFS_IOC_SETXFLAGS:
1889 		return xfs_ioc_setxflags(ip, filp, arg);
1890 
1891 	case XFS_IOC_FSSETDM: {
1892 		struct fsdmidata	dmi;
1893 
1894 		if (copy_from_user(&dmi, arg, sizeof(dmi)))
1895 			return -EFAULT;
1896 
1897 		error = mnt_want_write_file(filp);
1898 		if (error)
1899 			return error;
1900 
1901 		error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1902 				dmi.fsd_dmstate);
1903 		mnt_drop_write_file(filp);
1904 		return error;
1905 	}
1906 
1907 	case XFS_IOC_GETBMAP:
1908 	case XFS_IOC_GETBMAPA:
1909 	case XFS_IOC_GETBMAPX:
1910 		return xfs_ioc_getbmap(filp, cmd, arg);
1911 
1912 	case FS_IOC_GETFSMAP:
1913 		return xfs_ioc_getfsmap(ip, arg);
1914 
1915 	case XFS_IOC_SCRUB_METADATA:
1916 		return xfs_ioc_scrub_metadata(ip, arg);
1917 
1918 	case XFS_IOC_FD_TO_HANDLE:
1919 	case XFS_IOC_PATH_TO_HANDLE:
1920 	case XFS_IOC_PATH_TO_FSHANDLE: {
1921 		xfs_fsop_handlereq_t	hreq;
1922 
1923 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1924 			return -EFAULT;
1925 		return xfs_find_handle(cmd, &hreq);
1926 	}
1927 	case XFS_IOC_OPEN_BY_HANDLE: {
1928 		xfs_fsop_handlereq_t	hreq;
1929 
1930 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1931 			return -EFAULT;
1932 		return xfs_open_by_handle(filp, &hreq);
1933 	}
1934 	case XFS_IOC_FSSETDM_BY_HANDLE:
1935 		return xfs_fssetdm_by_handle(filp, arg);
1936 
1937 	case XFS_IOC_READLINK_BY_HANDLE: {
1938 		xfs_fsop_handlereq_t	hreq;
1939 
1940 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1941 			return -EFAULT;
1942 		return xfs_readlink_by_handle(filp, &hreq);
1943 	}
1944 	case XFS_IOC_ATTRLIST_BY_HANDLE:
1945 		return xfs_attrlist_by_handle(filp, arg);
1946 
1947 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
1948 		return xfs_attrmulti_by_handle(filp, arg);
1949 
1950 	case XFS_IOC_SWAPEXT: {
1951 		struct xfs_swapext	sxp;
1952 
1953 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1954 			return -EFAULT;
1955 		error = mnt_want_write_file(filp);
1956 		if (error)
1957 			return error;
1958 		error = xfs_ioc_swapext(&sxp);
1959 		mnt_drop_write_file(filp);
1960 		return error;
1961 	}
1962 
1963 	case XFS_IOC_FSCOUNTS: {
1964 		xfs_fsop_counts_t out;
1965 
1966 		error = xfs_fs_counts(mp, &out);
1967 		if (error)
1968 			return error;
1969 
1970 		if (copy_to_user(arg, &out, sizeof(out)))
1971 			return -EFAULT;
1972 		return 0;
1973 	}
1974 
1975 	case XFS_IOC_SET_RESBLKS: {
1976 		xfs_fsop_resblks_t inout;
1977 		uint64_t	   in;
1978 
1979 		if (!capable(CAP_SYS_ADMIN))
1980 			return -EPERM;
1981 
1982 		if (mp->m_flags & XFS_MOUNT_RDONLY)
1983 			return -EROFS;
1984 
1985 		if (copy_from_user(&inout, arg, sizeof(inout)))
1986 			return -EFAULT;
1987 
1988 		error = mnt_want_write_file(filp);
1989 		if (error)
1990 			return error;
1991 
1992 		/* input parameter is passed in resblks field of structure */
1993 		in = inout.resblks;
1994 		error = xfs_reserve_blocks(mp, &in, &inout);
1995 		mnt_drop_write_file(filp);
1996 		if (error)
1997 			return error;
1998 
1999 		if (copy_to_user(arg, &inout, sizeof(inout)))
2000 			return -EFAULT;
2001 		return 0;
2002 	}
2003 
2004 	case XFS_IOC_GET_RESBLKS: {
2005 		xfs_fsop_resblks_t out;
2006 
2007 		if (!capable(CAP_SYS_ADMIN))
2008 			return -EPERM;
2009 
2010 		error = xfs_reserve_blocks(mp, NULL, &out);
2011 		if (error)
2012 			return error;
2013 
2014 		if (copy_to_user(arg, &out, sizeof(out)))
2015 			return -EFAULT;
2016 
2017 		return 0;
2018 	}
2019 
2020 	case XFS_IOC_FSGROWFSDATA: {
2021 		xfs_growfs_data_t in;
2022 
2023 		if (copy_from_user(&in, arg, sizeof(in)))
2024 			return -EFAULT;
2025 
2026 		error = mnt_want_write_file(filp);
2027 		if (error)
2028 			return error;
2029 		error = xfs_growfs_data(mp, &in);
2030 		mnt_drop_write_file(filp);
2031 		return error;
2032 	}
2033 
2034 	case XFS_IOC_FSGROWFSLOG: {
2035 		xfs_growfs_log_t in;
2036 
2037 		if (copy_from_user(&in, arg, sizeof(in)))
2038 			return -EFAULT;
2039 
2040 		error = mnt_want_write_file(filp);
2041 		if (error)
2042 			return error;
2043 		error = xfs_growfs_log(mp, &in);
2044 		mnt_drop_write_file(filp);
2045 		return error;
2046 	}
2047 
2048 	case XFS_IOC_FSGROWFSRT: {
2049 		xfs_growfs_rt_t in;
2050 
2051 		if (copy_from_user(&in, arg, sizeof(in)))
2052 			return -EFAULT;
2053 
2054 		error = mnt_want_write_file(filp);
2055 		if (error)
2056 			return error;
2057 		error = xfs_growfs_rt(mp, &in);
2058 		mnt_drop_write_file(filp);
2059 		return error;
2060 	}
2061 
2062 	case XFS_IOC_GOINGDOWN: {
2063 		uint32_t in;
2064 
2065 		if (!capable(CAP_SYS_ADMIN))
2066 			return -EPERM;
2067 
2068 		if (get_user(in, (uint32_t __user *)arg))
2069 			return -EFAULT;
2070 
2071 		return xfs_fs_goingdown(mp, in);
2072 	}
2073 
2074 	case XFS_IOC_ERROR_INJECTION: {
2075 		xfs_error_injection_t in;
2076 
2077 		if (!capable(CAP_SYS_ADMIN))
2078 			return -EPERM;
2079 
2080 		if (copy_from_user(&in, arg, sizeof(in)))
2081 			return -EFAULT;
2082 
2083 		return xfs_errortag_add(mp, in.errtag);
2084 	}
2085 
2086 	case XFS_IOC_ERROR_CLEARALL:
2087 		if (!capable(CAP_SYS_ADMIN))
2088 			return -EPERM;
2089 
2090 		return xfs_errortag_clearall(mp);
2091 
2092 	case XFS_IOC_FREE_EOFBLOCKS: {
2093 		struct xfs_fs_eofblocks eofb;
2094 		struct xfs_eofblocks keofb;
2095 
2096 		if (!capable(CAP_SYS_ADMIN))
2097 			return -EPERM;
2098 
2099 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2100 			return -EROFS;
2101 
2102 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2103 			return -EFAULT;
2104 
2105 		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2106 		if (error)
2107 			return error;
2108 
2109 		return xfs_icache_free_eofblocks(mp, &keofb);
2110 	}
2111 
2112 	default:
2113 		return -ENOTTY;
2114 	}
2115 }
2116