xref: /openbmc/linux/fs/xfs/xfs_ioctl.c (revision 020c5260)
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 
124 		hsize = XFS_HSIZE(handle);
125 	}
126 
127 	error = -EFAULT;
128 	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
129 	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
130 		goto out_put;
131 
132 	error = 0;
133 
134  out_put:
135 	if (cmd == XFS_IOC_FD_TO_HANDLE)
136 		fdput(f);
137 	else
138 		path_put(&path);
139 	return error;
140 }
141 
142 /*
143  * No need to do permission checks on the various pathname components
144  * as the handle operations are privileged.
145  */
146 STATIC int
147 xfs_handle_acceptable(
148 	void			*context,
149 	struct dentry		*dentry)
150 {
151 	return 1;
152 }
153 
154 /*
155  * Convert userspace handle data into a dentry.
156  */
157 struct dentry *
158 xfs_handle_to_dentry(
159 	struct file		*parfilp,
160 	void __user		*uhandle,
161 	u32			hlen)
162 {
163 	xfs_handle_t		handle;
164 	struct xfs_fid64	fid;
165 
166 	/*
167 	 * Only allow handle opens under a directory.
168 	 */
169 	if (!S_ISDIR(file_inode(parfilp)->i_mode))
170 		return ERR_PTR(-ENOTDIR);
171 
172 	if (hlen != sizeof(xfs_handle_t))
173 		return ERR_PTR(-EINVAL);
174 	if (copy_from_user(&handle, uhandle, hlen))
175 		return ERR_PTR(-EFAULT);
176 	if (handle.ha_fid.fid_len !=
177 	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
178 		return ERR_PTR(-EINVAL);
179 
180 	memset(&fid, 0, sizeof(struct fid));
181 	fid.ino = handle.ha_fid.fid_ino;
182 	fid.gen = handle.ha_fid.fid_gen;
183 
184 	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
185 			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
186 			xfs_handle_acceptable, NULL);
187 }
188 
189 STATIC struct dentry *
190 xfs_handlereq_to_dentry(
191 	struct file		*parfilp,
192 	xfs_fsop_handlereq_t	*hreq)
193 {
194 	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
195 }
196 
197 int
198 xfs_open_by_handle(
199 	struct file		*parfilp,
200 	xfs_fsop_handlereq_t	*hreq)
201 {
202 	const struct cred	*cred = current_cred();
203 	int			error;
204 	int			fd;
205 	int			permflag;
206 	struct file		*filp;
207 	struct inode		*inode;
208 	struct dentry		*dentry;
209 	fmode_t			fmode;
210 	struct path		path;
211 
212 	if (!capable(CAP_SYS_ADMIN))
213 		return -EPERM;
214 
215 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
216 	if (IS_ERR(dentry))
217 		return PTR_ERR(dentry);
218 	inode = d_inode(dentry);
219 
220 	/* Restrict xfs_open_by_handle to directories & regular files. */
221 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
222 		error = -EPERM;
223 		goto out_dput;
224 	}
225 
226 #if BITS_PER_LONG != 32
227 	hreq->oflags |= O_LARGEFILE;
228 #endif
229 
230 	permflag = hreq->oflags;
231 	fmode = OPEN_FMODE(permflag);
232 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
233 	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
234 		error = -EPERM;
235 		goto out_dput;
236 	}
237 
238 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
239 		error = -EPERM;
240 		goto out_dput;
241 	}
242 
243 	/* Can't write directories. */
244 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
245 		error = -EISDIR;
246 		goto out_dput;
247 	}
248 
249 	fd = get_unused_fd_flags(0);
250 	if (fd < 0) {
251 		error = fd;
252 		goto out_dput;
253 	}
254 
255 	path.mnt = parfilp->f_path.mnt;
256 	path.dentry = dentry;
257 	filp = dentry_open(&path, hreq->oflags, cred);
258 	dput(dentry);
259 	if (IS_ERR(filp)) {
260 		put_unused_fd(fd);
261 		return PTR_ERR(filp);
262 	}
263 
264 	if (S_ISREG(inode->i_mode)) {
265 		filp->f_flags |= O_NOATIME;
266 		filp->f_mode |= FMODE_NOCMTIME;
267 	}
268 
269 	fd_install(fd, filp);
270 	return fd;
271 
272  out_dput:
273 	dput(dentry);
274 	return error;
275 }
276 
277 int
278 xfs_readlink_by_handle(
279 	struct file		*parfilp,
280 	xfs_fsop_handlereq_t	*hreq)
281 {
282 	struct dentry		*dentry;
283 	__u32			olen;
284 	int			error;
285 
286 	if (!capable(CAP_SYS_ADMIN))
287 		return -EPERM;
288 
289 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
290 	if (IS_ERR(dentry))
291 		return PTR_ERR(dentry);
292 
293 	/* Restrict this handle operation to symlinks only. */
294 	if (!d_is_symlink(dentry)) {
295 		error = -EINVAL;
296 		goto out_dput;
297 	}
298 
299 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
300 		error = -EFAULT;
301 		goto out_dput;
302 	}
303 
304 	error = vfs_readlink(dentry, hreq->ohandle, olen);
305 
306  out_dput:
307 	dput(dentry);
308 	return error;
309 }
310 
311 int
312 xfs_set_dmattrs(
313 	xfs_inode_t     *ip,
314 	u_int		evmask,
315 	u_int16_t	state)
316 {
317 	xfs_mount_t	*mp = ip->i_mount;
318 	xfs_trans_t	*tp;
319 	int		error;
320 
321 	if (!capable(CAP_SYS_ADMIN))
322 		return -EPERM;
323 
324 	if (XFS_FORCED_SHUTDOWN(mp))
325 		return -EIO;
326 
327 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
328 	if (error)
329 		return error;
330 
331 	xfs_ilock(ip, XFS_ILOCK_EXCL);
332 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
333 
334 	ip->i_d.di_dmevmask = evmask;
335 	ip->i_d.di_dmstate  = state;
336 
337 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
338 	error = xfs_trans_commit(tp);
339 
340 	return error;
341 }
342 
343 STATIC int
344 xfs_fssetdm_by_handle(
345 	struct file		*parfilp,
346 	void			__user *arg)
347 {
348 	int			error;
349 	struct fsdmidata	fsd;
350 	xfs_fsop_setdm_handlereq_t dmhreq;
351 	struct dentry		*dentry;
352 
353 	if (!capable(CAP_MKNOD))
354 		return -EPERM;
355 	if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
356 		return -EFAULT;
357 
358 	error = mnt_want_write_file(parfilp);
359 	if (error)
360 		return error;
361 
362 	dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
363 	if (IS_ERR(dentry)) {
364 		mnt_drop_write_file(parfilp);
365 		return PTR_ERR(dentry);
366 	}
367 
368 	if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
369 		error = -EPERM;
370 		goto out;
371 	}
372 
373 	if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
374 		error = -EFAULT;
375 		goto out;
376 	}
377 
378 	error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
379 				 fsd.fsd_dmstate);
380 
381  out:
382 	mnt_drop_write_file(parfilp);
383 	dput(dentry);
384 	return error;
385 }
386 
387 STATIC int
388 xfs_attrlist_by_handle(
389 	struct file		*parfilp,
390 	void			__user *arg)
391 {
392 	int			error = -ENOMEM;
393 	attrlist_cursor_kern_t	*cursor;
394 	struct xfs_fsop_attrlist_handlereq __user	*p = arg;
395 	xfs_fsop_attrlist_handlereq_t al_hreq;
396 	struct dentry		*dentry;
397 	char			*kbuf;
398 
399 	if (!capable(CAP_SYS_ADMIN))
400 		return -EPERM;
401 	if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
402 		return -EFAULT;
403 	if (al_hreq.buflen < sizeof(struct attrlist) ||
404 	    al_hreq.buflen > XFS_XATTR_LIST_MAX)
405 		return -EINVAL;
406 
407 	/*
408 	 * Reject flags, only allow namespaces.
409 	 */
410 	if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
411 		return -EINVAL;
412 
413 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
414 	if (IS_ERR(dentry))
415 		return PTR_ERR(dentry);
416 
417 	kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
418 	if (!kbuf)
419 		goto out_dput;
420 
421 	cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
422 	error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
423 					al_hreq.flags, cursor);
424 	if (error)
425 		goto out_kfree;
426 
427 	if (copy_to_user(&p->pos, cursor, sizeof(attrlist_cursor_kern_t))) {
428 		error = -EFAULT;
429 		goto out_kfree;
430 	}
431 
432 	if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
433 		error = -EFAULT;
434 
435 out_kfree:
436 	kmem_free(kbuf);
437 out_dput:
438 	dput(dentry);
439 	return error;
440 }
441 
442 int
443 xfs_attrmulti_attr_get(
444 	struct inode		*inode,
445 	unsigned char		*name,
446 	unsigned char		__user *ubuf,
447 	__uint32_t		*len,
448 	__uint32_t		flags)
449 {
450 	unsigned char		*kbuf;
451 	int			error = -EFAULT;
452 
453 	if (*len > XFS_XATTR_SIZE_MAX)
454 		return -EINVAL;
455 	kbuf = kmem_zalloc_large(*len, KM_SLEEP);
456 	if (!kbuf)
457 		return -ENOMEM;
458 
459 	error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
460 	if (error)
461 		goto out_kfree;
462 
463 	if (copy_to_user(ubuf, kbuf, *len))
464 		error = -EFAULT;
465 
466 out_kfree:
467 	kmem_free(kbuf);
468 	return error;
469 }
470 
471 int
472 xfs_attrmulti_attr_set(
473 	struct inode		*inode,
474 	unsigned char		*name,
475 	const unsigned char	__user *ubuf,
476 	__uint32_t		len,
477 	__uint32_t		flags)
478 {
479 	unsigned char		*kbuf;
480 	int			error;
481 
482 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
483 		return -EPERM;
484 	if (len > XFS_XATTR_SIZE_MAX)
485 		return -EINVAL;
486 
487 	kbuf = memdup_user(ubuf, len);
488 	if (IS_ERR(kbuf))
489 		return PTR_ERR(kbuf);
490 
491 	error = xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
492 	if (!error)
493 		xfs_forget_acl(inode, name, flags);
494 	kfree(kbuf);
495 	return error;
496 }
497 
498 int
499 xfs_attrmulti_attr_remove(
500 	struct inode		*inode,
501 	unsigned char		*name,
502 	__uint32_t		flags)
503 {
504 	int			error;
505 
506 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
507 		return -EPERM;
508 	error = xfs_attr_remove(XFS_I(inode), name, flags);
509 	if (!error)
510 		xfs_forget_acl(inode, name, flags);
511 	return error;
512 }
513 
514 STATIC int
515 xfs_attrmulti_by_handle(
516 	struct file		*parfilp,
517 	void			__user *arg)
518 {
519 	int			error;
520 	xfs_attr_multiop_t	*ops;
521 	xfs_fsop_attrmulti_handlereq_t am_hreq;
522 	struct dentry		*dentry;
523 	unsigned int		i, size;
524 	unsigned char		*attr_name;
525 
526 	if (!capable(CAP_SYS_ADMIN))
527 		return -EPERM;
528 	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
529 		return -EFAULT;
530 
531 	/* overflow check */
532 	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
533 		return -E2BIG;
534 
535 	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
536 	if (IS_ERR(dentry))
537 		return PTR_ERR(dentry);
538 
539 	error = -E2BIG;
540 	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
541 	if (!size || size > 16 * PAGE_SIZE)
542 		goto out_dput;
543 
544 	ops = memdup_user(am_hreq.ops, size);
545 	if (IS_ERR(ops)) {
546 		error = PTR_ERR(ops);
547 		goto out_dput;
548 	}
549 
550 	error = -ENOMEM;
551 	attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
552 	if (!attr_name)
553 		goto out_kfree_ops;
554 
555 	error = 0;
556 	for (i = 0; i < am_hreq.opcount; i++) {
557 		ops[i].am_error = strncpy_from_user((char *)attr_name,
558 				ops[i].am_attrname, MAXNAMELEN);
559 		if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
560 			error = -ERANGE;
561 		if (ops[i].am_error < 0)
562 			break;
563 
564 		switch (ops[i].am_opcode) {
565 		case ATTR_OP_GET:
566 			ops[i].am_error = xfs_attrmulti_attr_get(
567 					d_inode(dentry), attr_name,
568 					ops[i].am_attrvalue, &ops[i].am_length,
569 					ops[i].am_flags);
570 			break;
571 		case ATTR_OP_SET:
572 			ops[i].am_error = mnt_want_write_file(parfilp);
573 			if (ops[i].am_error)
574 				break;
575 			ops[i].am_error = xfs_attrmulti_attr_set(
576 					d_inode(dentry), attr_name,
577 					ops[i].am_attrvalue, ops[i].am_length,
578 					ops[i].am_flags);
579 			mnt_drop_write_file(parfilp);
580 			break;
581 		case ATTR_OP_REMOVE:
582 			ops[i].am_error = mnt_want_write_file(parfilp);
583 			if (ops[i].am_error)
584 				break;
585 			ops[i].am_error = xfs_attrmulti_attr_remove(
586 					d_inode(dentry), attr_name,
587 					ops[i].am_flags);
588 			mnt_drop_write_file(parfilp);
589 			break;
590 		default:
591 			ops[i].am_error = -EINVAL;
592 		}
593 	}
594 
595 	if (copy_to_user(am_hreq.ops, ops, size))
596 		error = -EFAULT;
597 
598 	kfree(attr_name);
599  out_kfree_ops:
600 	kfree(ops);
601  out_dput:
602 	dput(dentry);
603 	return error;
604 }
605 
606 int
607 xfs_ioc_space(
608 	struct file		*filp,
609 	unsigned int		cmd,
610 	xfs_flock64_t		*bf)
611 {
612 	struct inode		*inode = file_inode(filp);
613 	struct xfs_inode	*ip = XFS_I(inode);
614 	struct iattr		iattr;
615 	enum xfs_prealloc_flags	flags = 0;
616 	uint			iolock = XFS_IOLOCK_EXCL;
617 	int			error;
618 
619 	/*
620 	 * Only allow the sys admin to reserve space unless
621 	 * unwritten extents are enabled.
622 	 */
623 	if (!xfs_sb_version_hasextflgbit(&ip->i_mount->m_sb) &&
624 	    !capable(CAP_SYS_ADMIN))
625 		return -EPERM;
626 
627 	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
628 		return -EPERM;
629 
630 	if (!(filp->f_mode & FMODE_WRITE))
631 		return -EBADF;
632 
633 	if (!S_ISREG(inode->i_mode))
634 		return -EINVAL;
635 
636 	if (filp->f_flags & O_DSYNC)
637 		flags |= XFS_PREALLOC_SYNC;
638 	if (filp->f_mode & FMODE_NOCMTIME)
639 		flags |= XFS_PREALLOC_INVISIBLE;
640 
641 	error = mnt_want_write_file(filp);
642 	if (error)
643 		return error;
644 
645 	xfs_ilock(ip, iolock);
646 	error = xfs_break_layouts(inode, &iolock);
647 	if (error)
648 		goto out_unlock;
649 
650 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
651 	iolock |= XFS_MMAPLOCK_EXCL;
652 
653 	switch (bf->l_whence) {
654 	case 0: /*SEEK_SET*/
655 		break;
656 	case 1: /*SEEK_CUR*/
657 		bf->l_start += filp->f_pos;
658 		break;
659 	case 2: /*SEEK_END*/
660 		bf->l_start += XFS_ISIZE(ip);
661 		break;
662 	default:
663 		error = -EINVAL;
664 		goto out_unlock;
665 	}
666 
667 	/*
668 	 * length of <= 0 for resv/unresv/zero is invalid.  length for
669 	 * alloc/free is ignored completely and we have no idea what userspace
670 	 * might have set it to, so set it to zero to allow range
671 	 * checks to pass.
672 	 */
673 	switch (cmd) {
674 	case XFS_IOC_ZERO_RANGE:
675 	case XFS_IOC_RESVSP:
676 	case XFS_IOC_RESVSP64:
677 	case XFS_IOC_UNRESVSP:
678 	case XFS_IOC_UNRESVSP64:
679 		if (bf->l_len <= 0) {
680 			error = -EINVAL;
681 			goto out_unlock;
682 		}
683 		break;
684 	default:
685 		bf->l_len = 0;
686 		break;
687 	}
688 
689 	if (bf->l_start < 0 ||
690 	    bf->l_start > inode->i_sb->s_maxbytes ||
691 	    bf->l_start + bf->l_len < 0 ||
692 	    bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
693 		error = -EINVAL;
694 		goto out_unlock;
695 	}
696 
697 	switch (cmd) {
698 	case XFS_IOC_ZERO_RANGE:
699 		flags |= XFS_PREALLOC_SET;
700 		error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
701 		break;
702 	case XFS_IOC_RESVSP:
703 	case XFS_IOC_RESVSP64:
704 		flags |= XFS_PREALLOC_SET;
705 		error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
706 						XFS_BMAPI_PREALLOC);
707 		break;
708 	case XFS_IOC_UNRESVSP:
709 	case XFS_IOC_UNRESVSP64:
710 		error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
711 		break;
712 	case XFS_IOC_ALLOCSP:
713 	case XFS_IOC_ALLOCSP64:
714 	case XFS_IOC_FREESP:
715 	case XFS_IOC_FREESP64:
716 		flags |= XFS_PREALLOC_CLEAR;
717 		if (bf->l_start > XFS_ISIZE(ip)) {
718 			error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
719 					bf->l_start - XFS_ISIZE(ip), 0);
720 			if (error)
721 				goto out_unlock;
722 		}
723 
724 		iattr.ia_valid = ATTR_SIZE;
725 		iattr.ia_size = bf->l_start;
726 
727 		error = xfs_vn_setattr_size(file_dentry(filp), &iattr);
728 		break;
729 	default:
730 		ASSERT(0);
731 		error = -EINVAL;
732 	}
733 
734 	if (error)
735 		goto out_unlock;
736 
737 	error = xfs_update_prealloc_flags(ip, flags);
738 
739 out_unlock:
740 	xfs_iunlock(ip, iolock);
741 	mnt_drop_write_file(filp);
742 	return error;
743 }
744 
745 STATIC int
746 xfs_ioc_bulkstat(
747 	xfs_mount_t		*mp,
748 	unsigned int		cmd,
749 	void			__user *arg)
750 {
751 	xfs_fsop_bulkreq_t	bulkreq;
752 	int			count;	/* # of records returned */
753 	xfs_ino_t		inlast;	/* last inode number */
754 	int			done;
755 	int			error;
756 
757 	/* done = 1 if there are more stats to get and if bulkstat */
758 	/* should be called again (unused here, but used in dmapi) */
759 
760 	if (!capable(CAP_SYS_ADMIN))
761 		return -EPERM;
762 
763 	if (XFS_FORCED_SHUTDOWN(mp))
764 		return -EIO;
765 
766 	if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
767 		return -EFAULT;
768 
769 	if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
770 		return -EFAULT;
771 
772 	if ((count = bulkreq.icount) <= 0)
773 		return -EINVAL;
774 
775 	if (bulkreq.ubuffer == NULL)
776 		return -EINVAL;
777 
778 	if (cmd == XFS_IOC_FSINUMBERS)
779 		error = xfs_inumbers(mp, &inlast, &count,
780 					bulkreq.ubuffer, xfs_inumbers_fmt);
781 	else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
782 		error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
783 					sizeof(xfs_bstat_t), NULL, &done);
784 	else	/* XFS_IOC_FSBULKSTAT */
785 		error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
786 				     sizeof(xfs_bstat_t), bulkreq.ubuffer,
787 				     &done);
788 
789 	if (error)
790 		return error;
791 
792 	if (bulkreq.ocount != NULL) {
793 		if (copy_to_user(bulkreq.lastip, &inlast,
794 						sizeof(xfs_ino_t)))
795 			return -EFAULT;
796 
797 		if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
798 			return -EFAULT;
799 	}
800 
801 	return 0;
802 }
803 
804 STATIC int
805 xfs_ioc_fsgeometry_v1(
806 	xfs_mount_t		*mp,
807 	void			__user *arg)
808 {
809 	xfs_fsop_geom_t         fsgeo;
810 	int			error;
811 
812 	error = xfs_fs_geometry(mp, &fsgeo, 3);
813 	if (error)
814 		return error;
815 
816 	/*
817 	 * Caller should have passed an argument of type
818 	 * xfs_fsop_geom_v1_t.  This is a proper subset of the
819 	 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
820 	 */
821 	if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
822 		return -EFAULT;
823 	return 0;
824 }
825 
826 STATIC int
827 xfs_ioc_fsgeometry(
828 	xfs_mount_t		*mp,
829 	void			__user *arg)
830 {
831 	xfs_fsop_geom_t		fsgeo;
832 	int			error;
833 
834 	error = xfs_fs_geometry(mp, &fsgeo, 4);
835 	if (error)
836 		return error;
837 
838 	if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
839 		return -EFAULT;
840 	return 0;
841 }
842 
843 /*
844  * Linux extended inode flags interface.
845  */
846 
847 STATIC unsigned int
848 xfs_merge_ioc_xflags(
849 	unsigned int	flags,
850 	unsigned int	start)
851 {
852 	unsigned int	xflags = start;
853 
854 	if (flags & FS_IMMUTABLE_FL)
855 		xflags |= FS_XFLAG_IMMUTABLE;
856 	else
857 		xflags &= ~FS_XFLAG_IMMUTABLE;
858 	if (flags & FS_APPEND_FL)
859 		xflags |= FS_XFLAG_APPEND;
860 	else
861 		xflags &= ~FS_XFLAG_APPEND;
862 	if (flags & FS_SYNC_FL)
863 		xflags |= FS_XFLAG_SYNC;
864 	else
865 		xflags &= ~FS_XFLAG_SYNC;
866 	if (flags & FS_NOATIME_FL)
867 		xflags |= FS_XFLAG_NOATIME;
868 	else
869 		xflags &= ~FS_XFLAG_NOATIME;
870 	if (flags & FS_NODUMP_FL)
871 		xflags |= FS_XFLAG_NODUMP;
872 	else
873 		xflags &= ~FS_XFLAG_NODUMP;
874 
875 	return xflags;
876 }
877 
878 STATIC unsigned int
879 xfs_di2lxflags(
880 	__uint16_t	di_flags)
881 {
882 	unsigned int	flags = 0;
883 
884 	if (di_flags & XFS_DIFLAG_IMMUTABLE)
885 		flags |= FS_IMMUTABLE_FL;
886 	if (di_flags & XFS_DIFLAG_APPEND)
887 		flags |= FS_APPEND_FL;
888 	if (di_flags & XFS_DIFLAG_SYNC)
889 		flags |= FS_SYNC_FL;
890 	if (di_flags & XFS_DIFLAG_NOATIME)
891 		flags |= FS_NOATIME_FL;
892 	if (di_flags & XFS_DIFLAG_NODUMP)
893 		flags |= FS_NODUMP_FL;
894 	return flags;
895 }
896 
897 STATIC int
898 xfs_ioc_fsgetxattr(
899 	xfs_inode_t		*ip,
900 	int			attr,
901 	void			__user *arg)
902 {
903 	struct fsxattr		fa;
904 
905 	memset(&fa, 0, sizeof(struct fsxattr));
906 
907 	xfs_ilock(ip, XFS_ILOCK_SHARED);
908 	fa.fsx_xflags = xfs_ip2xflags(ip);
909 	fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
910 	fa.fsx_cowextsize = ip->i_d.di_cowextsize <<
911 			ip->i_mount->m_sb.sb_blocklog;
912 	fa.fsx_projid = xfs_get_projid(ip);
913 
914 	if (attr) {
915 		if (ip->i_afp) {
916 			if (ip->i_afp->if_flags & XFS_IFEXTENTS)
917 				fa.fsx_nextents = xfs_iext_count(ip->i_afp);
918 			else
919 				fa.fsx_nextents = ip->i_d.di_anextents;
920 		} else
921 			fa.fsx_nextents = 0;
922 	} else {
923 		if (ip->i_df.if_flags & XFS_IFEXTENTS)
924 			fa.fsx_nextents = xfs_iext_count(&ip->i_df);
925 		else
926 			fa.fsx_nextents = ip->i_d.di_nextents;
927 	}
928 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
929 
930 	if (copy_to_user(arg, &fa, sizeof(fa)))
931 		return -EFAULT;
932 	return 0;
933 }
934 
935 STATIC void
936 xfs_set_diflags(
937 	struct xfs_inode	*ip,
938 	unsigned int		xflags)
939 {
940 	unsigned int		di_flags;
941 	uint64_t		di_flags2;
942 
943 	/* can't set PREALLOC this way, just preserve it */
944 	di_flags = (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
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 	ip->i_d.di_flags = di_flags;
975 
976 	/* diflags2 only valid for v3 inodes. */
977 	if (ip->i_d.di_version < 3)
978 		return;
979 
980 	di_flags2 = (ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK);
981 	if (xflags & FS_XFLAG_DAX)
982 		di_flags2 |= XFS_DIFLAG2_DAX;
983 	if (xflags & FS_XFLAG_COWEXTSIZE)
984 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
985 
986 	ip->i_d.di_flags2 = di_flags2;
987 }
988 
989 STATIC void
990 xfs_diflags_to_linux(
991 	struct xfs_inode	*ip)
992 {
993 	struct inode		*inode = VFS_I(ip);
994 	unsigned int		xflags = xfs_ip2xflags(ip);
995 
996 	if (xflags & FS_XFLAG_IMMUTABLE)
997 		inode->i_flags |= S_IMMUTABLE;
998 	else
999 		inode->i_flags &= ~S_IMMUTABLE;
1000 	if (xflags & FS_XFLAG_APPEND)
1001 		inode->i_flags |= S_APPEND;
1002 	else
1003 		inode->i_flags &= ~S_APPEND;
1004 	if (xflags & FS_XFLAG_SYNC)
1005 		inode->i_flags |= S_SYNC;
1006 	else
1007 		inode->i_flags &= ~S_SYNC;
1008 	if (xflags & FS_XFLAG_NOATIME)
1009 		inode->i_flags |= S_NOATIME;
1010 	else
1011 		inode->i_flags &= ~S_NOATIME;
1012 	if (xflags & FS_XFLAG_DAX)
1013 		inode->i_flags |= S_DAX;
1014 	else
1015 		inode->i_flags &= ~S_DAX;
1016 
1017 }
1018 
1019 static int
1020 xfs_ioctl_setattr_xflags(
1021 	struct xfs_trans	*tp,
1022 	struct xfs_inode	*ip,
1023 	struct fsxattr		*fa)
1024 {
1025 	struct xfs_mount	*mp = ip->i_mount;
1026 
1027 	/* Can't change realtime flag if any extents are allocated. */
1028 	if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1029 	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1030 		return -EINVAL;
1031 
1032 	/* If realtime flag is set then must have realtime device */
1033 	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1034 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1035 		    (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1036 			return -EINVAL;
1037 	}
1038 
1039 	/* Clear reflink if we are actually able to set the rt flag. */
1040 	if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1041 		ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
1042 
1043 	/* Don't allow us to set DAX mode for a reflinked file for now. */
1044 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1045 		return -EINVAL;
1046 
1047 	/*
1048 	 * Can't modify an immutable/append-only file unless
1049 	 * we have appropriate permission.
1050 	 */
1051 	if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1052 	     (fa->fsx_xflags & (FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))) &&
1053 	    !capable(CAP_LINUX_IMMUTABLE))
1054 		return -EPERM;
1055 
1056 	xfs_set_diflags(ip, fa->fsx_xflags);
1057 	xfs_diflags_to_linux(ip);
1058 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1059 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1060 	XFS_STATS_INC(mp, xs_ig_attrchg);
1061 	return 0;
1062 }
1063 
1064 /*
1065  * If we are changing DAX flags, we have to ensure the file is clean and any
1066  * cached objects in the address space are invalidated and removed. This
1067  * requires us to lock out other IO and page faults similar to a truncate
1068  * operation. The locks need to be held until the transaction has been committed
1069  * so that the cache invalidation is atomic with respect to the DAX flag
1070  * manipulation.
1071  */
1072 static int
1073 xfs_ioctl_setattr_dax_invalidate(
1074 	struct xfs_inode	*ip,
1075 	struct fsxattr		*fa,
1076 	int			*join_flags)
1077 {
1078 	struct inode		*inode = VFS_I(ip);
1079 	int			error;
1080 
1081 	*join_flags = 0;
1082 
1083 	/*
1084 	 * It is only valid to set the DAX flag on regular files and
1085 	 * directories on filesystems where the block size is equal to the page
1086 	 * size. On directories it serves as an inherit hint.
1087 	 */
1088 	if (fa->fsx_xflags & FS_XFLAG_DAX) {
1089 		if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1090 			return -EINVAL;
1091 		if (ip->i_mount->m_sb.sb_blocksize != PAGE_SIZE)
1092 			return -EINVAL;
1093 	}
1094 
1095 	/* If the DAX state is not changing, we have nothing to do here. */
1096 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
1097 		return 0;
1098 	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
1099 		return 0;
1100 
1101 	/* lock, flush and invalidate mapping in preparation for flag change */
1102 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1103 	error = filemap_write_and_wait(inode->i_mapping);
1104 	if (error)
1105 		goto out_unlock;
1106 	error = invalidate_inode_pages2(inode->i_mapping);
1107 	if (error)
1108 		goto out_unlock;
1109 
1110 	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
1111 	return 0;
1112 
1113 out_unlock:
1114 	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1115 	return error;
1116 
1117 }
1118 
1119 /*
1120  * Set up the transaction structure for the setattr operation, checking that we
1121  * have permission to do so. On success, return a clean transaction and the
1122  * inode locked exclusively ready for further operation specific checks. On
1123  * failure, return an error without modifying or locking the inode.
1124  *
1125  * The inode might already be IO locked on call. If this is the case, it is
1126  * indicated in @join_flags and we take full responsibility for ensuring they
1127  * are unlocked from now on. Hence if we have an error here, we still have to
1128  * unlock them. Otherwise, once they are joined to the transaction, they will
1129  * be unlocked on commit/cancel.
1130  */
1131 static struct xfs_trans *
1132 xfs_ioctl_setattr_get_trans(
1133 	struct xfs_inode	*ip,
1134 	int			join_flags)
1135 {
1136 	struct xfs_mount	*mp = ip->i_mount;
1137 	struct xfs_trans	*tp;
1138 	int			error = -EROFS;
1139 
1140 	if (mp->m_flags & XFS_MOUNT_RDONLY)
1141 		goto out_unlock;
1142 	error = -EIO;
1143 	if (XFS_FORCED_SHUTDOWN(mp))
1144 		goto out_unlock;
1145 
1146 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
1147 	if (error)
1148 		return ERR_PTR(error);
1149 
1150 	xfs_ilock(ip, XFS_ILOCK_EXCL);
1151 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
1152 	join_flags = 0;
1153 
1154 	/*
1155 	 * CAP_FOWNER overrides the following restrictions:
1156 	 *
1157 	 * The user ID of the calling process must be equal to the file owner
1158 	 * ID, except in cases where the CAP_FSETID capability is applicable.
1159 	 */
1160 	if (!inode_owner_or_capable(VFS_I(ip))) {
1161 		error = -EPERM;
1162 		goto out_cancel;
1163 	}
1164 
1165 	if (mp->m_flags & XFS_MOUNT_WSYNC)
1166 		xfs_trans_set_sync(tp);
1167 
1168 	return tp;
1169 
1170 out_cancel:
1171 	xfs_trans_cancel(tp);
1172 out_unlock:
1173 	if (join_flags)
1174 		xfs_iunlock(ip, join_flags);
1175 	return ERR_PTR(error);
1176 }
1177 
1178 /*
1179  * extent size hint validation is somewhat cumbersome. Rules are:
1180  *
1181  * 1. extent size hint is only valid for directories and regular files
1182  * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1183  * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1184  * 4. can only be changed on regular files if no extents are allocated
1185  * 5. can be changed on directories at any time
1186  * 6. extsize hint of 0 turns off hints, clears inode flags.
1187  * 7. Extent size must be a multiple of the appropriate block size.
1188  * 8. for non-realtime files, the extent size hint must be limited
1189  *    to half the AG size to avoid alignment extending the extent beyond the
1190  *    limits of the AG.
1191  */
1192 static int
1193 xfs_ioctl_setattr_check_extsize(
1194 	struct xfs_inode	*ip,
1195 	struct fsxattr		*fa)
1196 {
1197 	struct xfs_mount	*mp = ip->i_mount;
1198 
1199 	if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(VFS_I(ip)->i_mode))
1200 		return -EINVAL;
1201 
1202 	if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
1203 	    !S_ISDIR(VFS_I(ip)->i_mode))
1204 		return -EINVAL;
1205 
1206 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_d.di_nextents &&
1207 	    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1208 		return -EINVAL;
1209 
1210 	if (fa->fsx_extsize != 0) {
1211 		xfs_extlen_t    size;
1212 		xfs_fsblock_t   extsize_fsb;
1213 
1214 		extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1215 		if (extsize_fsb > MAXEXTLEN)
1216 			return -EINVAL;
1217 
1218 		if (XFS_IS_REALTIME_INODE(ip) ||
1219 		    (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1220 			size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1221 		} else {
1222 			size = mp->m_sb.sb_blocksize;
1223 			if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1224 				return -EINVAL;
1225 		}
1226 
1227 		if (fa->fsx_extsize % size)
1228 			return -EINVAL;
1229 	} else
1230 		fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
1231 
1232 	return 0;
1233 }
1234 
1235 /*
1236  * CoW extent size hint validation rules are:
1237  *
1238  * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
1239  *    The inode does not have to have any shared blocks, but it must be a v3.
1240  * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
1241  *    for a directory, the hint is propagated to new files.
1242  * 3. Can be changed on files & directories at any time.
1243  * 4. CoW extsize hint of 0 turns off hints, clears inode flags.
1244  * 5. Extent size must be a multiple of the appropriate block size.
1245  * 6. The extent size hint must be limited to half the AG size to avoid
1246  *    alignment extending the extent beyond the limits of the AG.
1247  */
1248 static int
1249 xfs_ioctl_setattr_check_cowextsize(
1250 	struct xfs_inode	*ip,
1251 	struct fsxattr		*fa)
1252 {
1253 	struct xfs_mount	*mp = ip->i_mount;
1254 
1255 	if (!(fa->fsx_xflags & FS_XFLAG_COWEXTSIZE))
1256 		return 0;
1257 
1258 	if (!xfs_sb_version_hasreflink(&ip->i_mount->m_sb) ||
1259 	    ip->i_d.di_version != 3)
1260 		return -EINVAL;
1261 
1262 	if (!S_ISREG(VFS_I(ip)->i_mode) && !S_ISDIR(VFS_I(ip)->i_mode))
1263 		return -EINVAL;
1264 
1265 	if (fa->fsx_cowextsize != 0) {
1266 		xfs_extlen_t    size;
1267 		xfs_fsblock_t   cowextsize_fsb;
1268 
1269 		cowextsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1270 		if (cowextsize_fsb > MAXEXTLEN)
1271 			return -EINVAL;
1272 
1273 		size = mp->m_sb.sb_blocksize;
1274 		if (cowextsize_fsb > mp->m_sb.sb_agblocks / 2)
1275 			return -EINVAL;
1276 
1277 		if (fa->fsx_cowextsize % size)
1278 			return -EINVAL;
1279 	} else
1280 		fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
1281 
1282 	return 0;
1283 }
1284 
1285 static int
1286 xfs_ioctl_setattr_check_projid(
1287 	struct xfs_inode	*ip,
1288 	struct fsxattr		*fa)
1289 {
1290 	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1291 	if (fa->fsx_projid > (__uint16_t)-1 &&
1292 	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1293 		return -EINVAL;
1294 
1295 	/*
1296 	 * Project Quota ID state is only allowed to change from within the init
1297 	 * namespace. Enforce that restriction only if we are trying to change
1298 	 * the quota ID state. Everything else is allowed in user namespaces.
1299 	 */
1300 	if (current_user_ns() == &init_user_ns)
1301 		return 0;
1302 
1303 	if (xfs_get_projid(ip) != fa->fsx_projid)
1304 		return -EINVAL;
1305 	if ((fa->fsx_xflags & FS_XFLAG_PROJINHERIT) !=
1306 	    (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1307 		return -EINVAL;
1308 
1309 	return 0;
1310 }
1311 
1312 STATIC int
1313 xfs_ioctl_setattr(
1314 	xfs_inode_t		*ip,
1315 	struct fsxattr		*fa)
1316 {
1317 	struct xfs_mount	*mp = ip->i_mount;
1318 	struct xfs_trans	*tp;
1319 	struct xfs_dquot	*udqp = NULL;
1320 	struct xfs_dquot	*pdqp = NULL;
1321 	struct xfs_dquot	*olddquot = NULL;
1322 	int			code;
1323 	int			join_flags = 0;
1324 
1325 	trace_xfs_ioctl_setattr(ip);
1326 
1327 	code = xfs_ioctl_setattr_check_projid(ip, fa);
1328 	if (code)
1329 		return code;
1330 
1331 	/*
1332 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1333 	 * before we start any other transactions. Trying to do this later
1334 	 * is messy. We don't care to take a readlock to look at the ids
1335 	 * in inode here, because we can't hold it across the trans_reserve.
1336 	 * If the IDs do change before we take the ilock, we're covered
1337 	 * because the i_*dquot fields will get updated anyway.
1338 	 */
1339 	if (XFS_IS_QUOTA_ON(mp)) {
1340 		code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1341 					 ip->i_d.di_gid, fa->fsx_projid,
1342 					 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1343 		if (code)
1344 			return code;
1345 	}
1346 
1347 	/*
1348 	 * Changing DAX config may require inode locking for mapping
1349 	 * invalidation. These need to be held all the way to transaction commit
1350 	 * or cancel time, so need to be passed through to
1351 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1352 	 * appropriately.
1353 	 */
1354 	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
1355 	if (code)
1356 		goto error_free_dquots;
1357 
1358 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1359 	if (IS_ERR(tp)) {
1360 		code = PTR_ERR(tp);
1361 		goto error_free_dquots;
1362 	}
1363 
1364 
1365 	if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1366 	    xfs_get_projid(ip) != fa->fsx_projid) {
1367 		code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1368 				capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
1369 		if (code)	/* out of quota */
1370 			goto error_trans_cancel;
1371 	}
1372 
1373 	code = xfs_ioctl_setattr_check_extsize(ip, fa);
1374 	if (code)
1375 		goto error_trans_cancel;
1376 
1377 	code = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1378 	if (code)
1379 		goto error_trans_cancel;
1380 
1381 	code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1382 	if (code)
1383 		goto error_trans_cancel;
1384 
1385 	/*
1386 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1387 	 * overrides the following restrictions:
1388 	 *
1389 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1390 	 * successful return from chown()
1391 	 */
1392 
1393 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1394 	    !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1395 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1396 
1397 	/* Change the ownerships and register project quota modifications */
1398 	if (xfs_get_projid(ip) != fa->fsx_projid) {
1399 		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1400 			olddquot = xfs_qm_vop_chown(tp, ip,
1401 						&ip->i_pdquot, pdqp);
1402 		}
1403 		ASSERT(ip->i_d.di_version > 1);
1404 		xfs_set_projid(ip, fa->fsx_projid);
1405 	}
1406 
1407 	/*
1408 	 * Only set the extent size hint if we've already determined that the
1409 	 * extent size hint should be set on the inode. If no extent size flags
1410 	 * are set on the inode then unconditionally clear the extent size hint.
1411 	 */
1412 	if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1413 		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1414 	else
1415 		ip->i_d.di_extsize = 0;
1416 	if (ip->i_d.di_version == 3 &&
1417 	    (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
1418 		ip->i_d.di_cowextsize = fa->fsx_cowextsize >>
1419 				mp->m_sb.sb_blocklog;
1420 	else
1421 		ip->i_d.di_cowextsize = 0;
1422 
1423 	code = xfs_trans_commit(tp);
1424 
1425 	/*
1426 	 * Release any dquot(s) the inode had kept before chown.
1427 	 */
1428 	xfs_qm_dqrele(olddquot);
1429 	xfs_qm_dqrele(udqp);
1430 	xfs_qm_dqrele(pdqp);
1431 
1432 	return code;
1433 
1434 error_trans_cancel:
1435 	xfs_trans_cancel(tp);
1436 error_free_dquots:
1437 	xfs_qm_dqrele(udqp);
1438 	xfs_qm_dqrele(pdqp);
1439 	return code;
1440 }
1441 
1442 STATIC int
1443 xfs_ioc_fssetxattr(
1444 	xfs_inode_t		*ip,
1445 	struct file		*filp,
1446 	void			__user *arg)
1447 {
1448 	struct fsxattr		fa;
1449 	int error;
1450 
1451 	if (copy_from_user(&fa, arg, sizeof(fa)))
1452 		return -EFAULT;
1453 
1454 	error = mnt_want_write_file(filp);
1455 	if (error)
1456 		return error;
1457 	error = xfs_ioctl_setattr(ip, &fa);
1458 	mnt_drop_write_file(filp);
1459 	return error;
1460 }
1461 
1462 STATIC int
1463 xfs_ioc_getxflags(
1464 	xfs_inode_t		*ip,
1465 	void			__user *arg)
1466 {
1467 	unsigned int		flags;
1468 
1469 	flags = xfs_di2lxflags(ip->i_d.di_flags);
1470 	if (copy_to_user(arg, &flags, sizeof(flags)))
1471 		return -EFAULT;
1472 	return 0;
1473 }
1474 
1475 STATIC int
1476 xfs_ioc_setxflags(
1477 	struct xfs_inode	*ip,
1478 	struct file		*filp,
1479 	void			__user *arg)
1480 {
1481 	struct xfs_trans	*tp;
1482 	struct fsxattr		fa;
1483 	unsigned int		flags;
1484 	int			join_flags = 0;
1485 	int			error;
1486 
1487 	if (copy_from_user(&flags, arg, sizeof(flags)))
1488 		return -EFAULT;
1489 
1490 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1491 		      FS_NOATIME_FL | FS_NODUMP_FL | \
1492 		      FS_SYNC_FL))
1493 		return -EOPNOTSUPP;
1494 
1495 	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1496 
1497 	error = mnt_want_write_file(filp);
1498 	if (error)
1499 		return error;
1500 
1501 	/*
1502 	 * Changing DAX config may require inode locking for mapping
1503 	 * invalidation. These need to be held all the way to transaction commit
1504 	 * or cancel time, so need to be passed through to
1505 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1506 	 * appropriately.
1507 	 */
1508 	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
1509 	if (error)
1510 		goto out_drop_write;
1511 
1512 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1513 	if (IS_ERR(tp)) {
1514 		error = PTR_ERR(tp);
1515 		goto out_drop_write;
1516 	}
1517 
1518 	error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1519 	if (error) {
1520 		xfs_trans_cancel(tp);
1521 		goto out_drop_write;
1522 	}
1523 
1524 	error = xfs_trans_commit(tp);
1525 out_drop_write:
1526 	mnt_drop_write_file(filp);
1527 	return error;
1528 }
1529 
1530 STATIC int
1531 xfs_getbmap_format(void **ap, struct getbmapx *bmv)
1532 {
1533 	struct getbmap __user	*base = (struct getbmap __user *)*ap;
1534 
1535 	/* copy only getbmap portion (not getbmapx) */
1536 	if (copy_to_user(base, bmv, sizeof(struct getbmap)))
1537 		return -EFAULT;
1538 
1539 	*ap += sizeof(struct getbmap);
1540 	return 0;
1541 }
1542 
1543 STATIC int
1544 xfs_ioc_getbmap(
1545 	struct file		*file,
1546 	unsigned int		cmd,
1547 	void			__user *arg)
1548 {
1549 	struct getbmapx		bmx = { 0 };
1550 	int			error;
1551 
1552 	/* struct getbmap is a strict subset of struct getbmapx. */
1553 	if (copy_from_user(&bmx, arg, offsetof(struct getbmapx, bmv_iflags)))
1554 		return -EFAULT;
1555 
1556 	if (bmx.bmv_count < 2)
1557 		return -EINVAL;
1558 
1559 	bmx.bmv_iflags = (cmd == XFS_IOC_GETBMAPA ? BMV_IF_ATTRFORK : 0);
1560 	if (file->f_mode & FMODE_NOCMTIME)
1561 		bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1562 
1563 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, xfs_getbmap_format,
1564 			    (__force struct getbmap *)arg+1);
1565 	if (error)
1566 		return error;
1567 
1568 	/* copy back header - only size of getbmap */
1569 	if (copy_to_user(arg, &bmx, sizeof(struct getbmap)))
1570 		return -EFAULT;
1571 	return 0;
1572 }
1573 
1574 STATIC int
1575 xfs_getbmapx_format(void **ap, struct getbmapx *bmv)
1576 {
1577 	struct getbmapx __user	*base = (struct getbmapx __user *)*ap;
1578 
1579 	if (copy_to_user(base, bmv, sizeof(struct getbmapx)))
1580 		return -EFAULT;
1581 
1582 	*ap += sizeof(struct getbmapx);
1583 	return 0;
1584 }
1585 
1586 STATIC int
1587 xfs_ioc_getbmapx(
1588 	struct xfs_inode	*ip,
1589 	void			__user *arg)
1590 {
1591 	struct getbmapx		bmx;
1592 	int			error;
1593 
1594 	if (copy_from_user(&bmx, arg, sizeof(bmx)))
1595 		return -EFAULT;
1596 
1597 	if (bmx.bmv_count < 2)
1598 		return -EINVAL;
1599 
1600 	if (bmx.bmv_iflags & (~BMV_IF_VALID))
1601 		return -EINVAL;
1602 
1603 	error = xfs_getbmap(ip, &bmx, xfs_getbmapx_format,
1604 			    (__force struct getbmapx *)arg+1);
1605 	if (error)
1606 		return error;
1607 
1608 	/* copy back header */
1609 	if (copy_to_user(arg, &bmx, sizeof(struct getbmapx)))
1610 		return -EFAULT;
1611 
1612 	return 0;
1613 }
1614 
1615 struct getfsmap_info {
1616 	struct xfs_mount	*mp;
1617 	struct fsmap_head __user *data;
1618 	unsigned int		idx;
1619 	__u32			last_flags;
1620 };
1621 
1622 STATIC int
1623 xfs_getfsmap_format(struct xfs_fsmap *xfm, void *priv)
1624 {
1625 	struct getfsmap_info	*info = priv;
1626 	struct fsmap		fm;
1627 
1628 	trace_xfs_getfsmap_mapping(info->mp, xfm);
1629 
1630 	info->last_flags = xfm->fmr_flags;
1631 	xfs_fsmap_from_internal(&fm, xfm);
1632 	if (copy_to_user(&info->data->fmh_recs[info->idx++], &fm,
1633 			sizeof(struct fsmap)))
1634 		return -EFAULT;
1635 
1636 	return 0;
1637 }
1638 
1639 STATIC int
1640 xfs_ioc_getfsmap(
1641 	struct xfs_inode	*ip,
1642 	struct fsmap_head	__user *arg)
1643 {
1644 	struct getfsmap_info	info = { NULL };
1645 	struct xfs_fsmap_head	xhead = {0};
1646 	struct fsmap_head	head;
1647 	bool			aborted = false;
1648 	int			error;
1649 
1650 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1651 		return -EFAULT;
1652 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1653 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1654 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1655 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1656 		       sizeof(head.fmh_keys[1].fmr_reserved)))
1657 		return -EINVAL;
1658 
1659 	xhead.fmh_iflags = head.fmh_iflags;
1660 	xhead.fmh_count = head.fmh_count;
1661 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1662 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1663 
1664 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1665 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1666 
1667 	info.mp = ip->i_mount;
1668 	info.data = arg;
1669 	error = xfs_getfsmap(ip->i_mount, &xhead, xfs_getfsmap_format, &info);
1670 	if (error == XFS_BTREE_QUERY_RANGE_ABORT) {
1671 		error = 0;
1672 		aborted = true;
1673 	} else if (error)
1674 		return error;
1675 
1676 	/* If we didn't abort, set the "last" flag in the last fmx */
1677 	if (!aborted && info.idx) {
1678 		info.last_flags |= FMR_OF_LAST;
1679 		if (copy_to_user(&info.data->fmh_recs[info.idx - 1].fmr_flags,
1680 				&info.last_flags, sizeof(info.last_flags)))
1681 			return -EFAULT;
1682 	}
1683 
1684 	/* copy back header */
1685 	head.fmh_entries = xhead.fmh_entries;
1686 	head.fmh_oflags = xhead.fmh_oflags;
1687 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
1688 		return -EFAULT;
1689 
1690 	return 0;
1691 }
1692 
1693 int
1694 xfs_ioc_swapext(
1695 	xfs_swapext_t	*sxp)
1696 {
1697 	xfs_inode_t     *ip, *tip;
1698 	struct fd	f, tmp;
1699 	int		error = 0;
1700 
1701 	/* Pull information for the target fd */
1702 	f = fdget((int)sxp->sx_fdtarget);
1703 	if (!f.file) {
1704 		error = -EINVAL;
1705 		goto out;
1706 	}
1707 
1708 	if (!(f.file->f_mode & FMODE_WRITE) ||
1709 	    !(f.file->f_mode & FMODE_READ) ||
1710 	    (f.file->f_flags & O_APPEND)) {
1711 		error = -EBADF;
1712 		goto out_put_file;
1713 	}
1714 
1715 	tmp = fdget((int)sxp->sx_fdtmp);
1716 	if (!tmp.file) {
1717 		error = -EINVAL;
1718 		goto out_put_file;
1719 	}
1720 
1721 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1722 	    !(tmp.file->f_mode & FMODE_READ) ||
1723 	    (tmp.file->f_flags & O_APPEND)) {
1724 		error = -EBADF;
1725 		goto out_put_tmp_file;
1726 	}
1727 
1728 	if (IS_SWAPFILE(file_inode(f.file)) ||
1729 	    IS_SWAPFILE(file_inode(tmp.file))) {
1730 		error = -EINVAL;
1731 		goto out_put_tmp_file;
1732 	}
1733 
1734 	/*
1735 	 * We need to ensure that the fds passed in point to XFS inodes
1736 	 * before we cast and access them as XFS structures as we have no
1737 	 * control over what the user passes us here.
1738 	 */
1739 	if (f.file->f_op != &xfs_file_operations ||
1740 	    tmp.file->f_op != &xfs_file_operations) {
1741 		error = -EINVAL;
1742 		goto out_put_tmp_file;
1743 	}
1744 
1745 	ip = XFS_I(file_inode(f.file));
1746 	tip = XFS_I(file_inode(tmp.file));
1747 
1748 	if (ip->i_mount != tip->i_mount) {
1749 		error = -EINVAL;
1750 		goto out_put_tmp_file;
1751 	}
1752 
1753 	if (ip->i_ino == tip->i_ino) {
1754 		error = -EINVAL;
1755 		goto out_put_tmp_file;
1756 	}
1757 
1758 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1759 		error = -EIO;
1760 		goto out_put_tmp_file;
1761 	}
1762 
1763 	error = xfs_swap_extents(ip, tip, sxp);
1764 
1765  out_put_tmp_file:
1766 	fdput(tmp);
1767  out_put_file:
1768 	fdput(f);
1769  out:
1770 	return error;
1771 }
1772 
1773 /*
1774  * Note: some of the ioctl's return positive numbers as a
1775  * byte count indicating success, such as readlink_by_handle.
1776  * So we don't "sign flip" like most other routines.  This means
1777  * true errors need to be returned as a negative value.
1778  */
1779 long
1780 xfs_file_ioctl(
1781 	struct file		*filp,
1782 	unsigned int		cmd,
1783 	unsigned long		p)
1784 {
1785 	struct inode		*inode = file_inode(filp);
1786 	struct xfs_inode	*ip = XFS_I(inode);
1787 	struct xfs_mount	*mp = ip->i_mount;
1788 	void			__user *arg = (void __user *)p;
1789 	int			error;
1790 
1791 	trace_xfs_file_ioctl(ip);
1792 
1793 	switch (cmd) {
1794 	case FITRIM:
1795 		return xfs_ioc_trim(mp, arg);
1796 	case XFS_IOC_ALLOCSP:
1797 	case XFS_IOC_FREESP:
1798 	case XFS_IOC_RESVSP:
1799 	case XFS_IOC_UNRESVSP:
1800 	case XFS_IOC_ALLOCSP64:
1801 	case XFS_IOC_FREESP64:
1802 	case XFS_IOC_RESVSP64:
1803 	case XFS_IOC_UNRESVSP64:
1804 	case XFS_IOC_ZERO_RANGE: {
1805 		xfs_flock64_t		bf;
1806 
1807 		if (copy_from_user(&bf, arg, sizeof(bf)))
1808 			return -EFAULT;
1809 		return xfs_ioc_space(filp, cmd, &bf);
1810 	}
1811 	case XFS_IOC_DIOINFO: {
1812 		struct dioattr	da;
1813 		xfs_buftarg_t	*target =
1814 			XFS_IS_REALTIME_INODE(ip) ?
1815 			mp->m_rtdev_targp : mp->m_ddev_targp;
1816 
1817 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1818 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1819 
1820 		if (copy_to_user(arg, &da, sizeof(da)))
1821 			return -EFAULT;
1822 		return 0;
1823 	}
1824 
1825 	case XFS_IOC_FSBULKSTAT_SINGLE:
1826 	case XFS_IOC_FSBULKSTAT:
1827 	case XFS_IOC_FSINUMBERS:
1828 		return xfs_ioc_bulkstat(mp, cmd, arg);
1829 
1830 	case XFS_IOC_FSGEOMETRY_V1:
1831 		return xfs_ioc_fsgeometry_v1(mp, arg);
1832 
1833 	case XFS_IOC_FSGEOMETRY:
1834 		return xfs_ioc_fsgeometry(mp, arg);
1835 
1836 	case XFS_IOC_GETVERSION:
1837 		return put_user(inode->i_generation, (int __user *)arg);
1838 
1839 	case XFS_IOC_FSGETXATTR:
1840 		return xfs_ioc_fsgetxattr(ip, 0, arg);
1841 	case XFS_IOC_FSGETXATTRA:
1842 		return xfs_ioc_fsgetxattr(ip, 1, arg);
1843 	case XFS_IOC_FSSETXATTR:
1844 		return xfs_ioc_fssetxattr(ip, filp, arg);
1845 	case XFS_IOC_GETXFLAGS:
1846 		return xfs_ioc_getxflags(ip, arg);
1847 	case XFS_IOC_SETXFLAGS:
1848 		return xfs_ioc_setxflags(ip, filp, arg);
1849 
1850 	case XFS_IOC_FSSETDM: {
1851 		struct fsdmidata	dmi;
1852 
1853 		if (copy_from_user(&dmi, arg, sizeof(dmi)))
1854 			return -EFAULT;
1855 
1856 		error = mnt_want_write_file(filp);
1857 		if (error)
1858 			return error;
1859 
1860 		error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1861 				dmi.fsd_dmstate);
1862 		mnt_drop_write_file(filp);
1863 		return error;
1864 	}
1865 
1866 	case XFS_IOC_GETBMAP:
1867 	case XFS_IOC_GETBMAPA:
1868 		return xfs_ioc_getbmap(filp, cmd, arg);
1869 
1870 	case XFS_IOC_GETBMAPX:
1871 		return xfs_ioc_getbmapx(ip, arg);
1872 
1873 	case FS_IOC_GETFSMAP:
1874 		return xfs_ioc_getfsmap(ip, arg);
1875 
1876 	case XFS_IOC_FD_TO_HANDLE:
1877 	case XFS_IOC_PATH_TO_HANDLE:
1878 	case XFS_IOC_PATH_TO_FSHANDLE: {
1879 		xfs_fsop_handlereq_t	hreq;
1880 
1881 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1882 			return -EFAULT;
1883 		return xfs_find_handle(cmd, &hreq);
1884 	}
1885 	case XFS_IOC_OPEN_BY_HANDLE: {
1886 		xfs_fsop_handlereq_t	hreq;
1887 
1888 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1889 			return -EFAULT;
1890 		return xfs_open_by_handle(filp, &hreq);
1891 	}
1892 	case XFS_IOC_FSSETDM_BY_HANDLE:
1893 		return xfs_fssetdm_by_handle(filp, arg);
1894 
1895 	case XFS_IOC_READLINK_BY_HANDLE: {
1896 		xfs_fsop_handlereq_t	hreq;
1897 
1898 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1899 			return -EFAULT;
1900 		return xfs_readlink_by_handle(filp, &hreq);
1901 	}
1902 	case XFS_IOC_ATTRLIST_BY_HANDLE:
1903 		return xfs_attrlist_by_handle(filp, arg);
1904 
1905 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
1906 		return xfs_attrmulti_by_handle(filp, arg);
1907 
1908 	case XFS_IOC_SWAPEXT: {
1909 		struct xfs_swapext	sxp;
1910 
1911 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1912 			return -EFAULT;
1913 		error = mnt_want_write_file(filp);
1914 		if (error)
1915 			return error;
1916 		error = xfs_ioc_swapext(&sxp);
1917 		mnt_drop_write_file(filp);
1918 		return error;
1919 	}
1920 
1921 	case XFS_IOC_FSCOUNTS: {
1922 		xfs_fsop_counts_t out;
1923 
1924 		error = xfs_fs_counts(mp, &out);
1925 		if (error)
1926 			return error;
1927 
1928 		if (copy_to_user(arg, &out, sizeof(out)))
1929 			return -EFAULT;
1930 		return 0;
1931 	}
1932 
1933 	case XFS_IOC_SET_RESBLKS: {
1934 		xfs_fsop_resblks_t inout;
1935 		__uint64_t	   in;
1936 
1937 		if (!capable(CAP_SYS_ADMIN))
1938 			return -EPERM;
1939 
1940 		if (mp->m_flags & XFS_MOUNT_RDONLY)
1941 			return -EROFS;
1942 
1943 		if (copy_from_user(&inout, arg, sizeof(inout)))
1944 			return -EFAULT;
1945 
1946 		error = mnt_want_write_file(filp);
1947 		if (error)
1948 			return error;
1949 
1950 		/* input parameter is passed in resblks field of structure */
1951 		in = inout.resblks;
1952 		error = xfs_reserve_blocks(mp, &in, &inout);
1953 		mnt_drop_write_file(filp);
1954 		if (error)
1955 			return error;
1956 
1957 		if (copy_to_user(arg, &inout, sizeof(inout)))
1958 			return -EFAULT;
1959 		return 0;
1960 	}
1961 
1962 	case XFS_IOC_GET_RESBLKS: {
1963 		xfs_fsop_resblks_t out;
1964 
1965 		if (!capable(CAP_SYS_ADMIN))
1966 			return -EPERM;
1967 
1968 		error = xfs_reserve_blocks(mp, NULL, &out);
1969 		if (error)
1970 			return error;
1971 
1972 		if (copy_to_user(arg, &out, sizeof(out)))
1973 			return -EFAULT;
1974 
1975 		return 0;
1976 	}
1977 
1978 	case XFS_IOC_FSGROWFSDATA: {
1979 		xfs_growfs_data_t in;
1980 
1981 		if (copy_from_user(&in, arg, sizeof(in)))
1982 			return -EFAULT;
1983 
1984 		error = mnt_want_write_file(filp);
1985 		if (error)
1986 			return error;
1987 		error = xfs_growfs_data(mp, &in);
1988 		mnt_drop_write_file(filp);
1989 		return error;
1990 	}
1991 
1992 	case XFS_IOC_FSGROWFSLOG: {
1993 		xfs_growfs_log_t in;
1994 
1995 		if (copy_from_user(&in, arg, sizeof(in)))
1996 			return -EFAULT;
1997 
1998 		error = mnt_want_write_file(filp);
1999 		if (error)
2000 			return error;
2001 		error = xfs_growfs_log(mp, &in);
2002 		mnt_drop_write_file(filp);
2003 		return error;
2004 	}
2005 
2006 	case XFS_IOC_FSGROWFSRT: {
2007 		xfs_growfs_rt_t in;
2008 
2009 		if (copy_from_user(&in, arg, sizeof(in)))
2010 			return -EFAULT;
2011 
2012 		error = mnt_want_write_file(filp);
2013 		if (error)
2014 			return error;
2015 		error = xfs_growfs_rt(mp, &in);
2016 		mnt_drop_write_file(filp);
2017 		return error;
2018 	}
2019 
2020 	case XFS_IOC_GOINGDOWN: {
2021 		__uint32_t in;
2022 
2023 		if (!capable(CAP_SYS_ADMIN))
2024 			return -EPERM;
2025 
2026 		if (get_user(in, (__uint32_t __user *)arg))
2027 			return -EFAULT;
2028 
2029 		return xfs_fs_goingdown(mp, in);
2030 	}
2031 
2032 	case XFS_IOC_ERROR_INJECTION: {
2033 		xfs_error_injection_t in;
2034 
2035 		if (!capable(CAP_SYS_ADMIN))
2036 			return -EPERM;
2037 
2038 		if (copy_from_user(&in, arg, sizeof(in)))
2039 			return -EFAULT;
2040 
2041 		return xfs_errortag_add(in.errtag, mp);
2042 	}
2043 
2044 	case XFS_IOC_ERROR_CLEARALL:
2045 		if (!capable(CAP_SYS_ADMIN))
2046 			return -EPERM;
2047 
2048 		return xfs_errortag_clearall(mp, 1);
2049 
2050 	case XFS_IOC_FREE_EOFBLOCKS: {
2051 		struct xfs_fs_eofblocks eofb;
2052 		struct xfs_eofblocks keofb;
2053 
2054 		if (!capable(CAP_SYS_ADMIN))
2055 			return -EPERM;
2056 
2057 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2058 			return -EROFS;
2059 
2060 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2061 			return -EFAULT;
2062 
2063 		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2064 		if (error)
2065 			return error;
2066 
2067 		return xfs_icache_free_eofblocks(mp, &keofb);
2068 	}
2069 
2070 	default:
2071 		return -ENOTTY;
2072 	}
2073 }
2074