xref: /openbmc/linux/fs/xfs/xfs_ioctl.c (revision 763f96944c954ce0e00a10a7bdfe29adbe4f92eb)
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_acl.h"
43 #include "xfs_btree.h"
44 #include <linux/fsmap.h>
45 #include "xfs_fsmap.h"
46 #include "scrub/xfs_scrub.h"
47 #include "xfs_sb.h"
48 
49 #include <linux/capability.h>
50 #include <linux/cred.h>
51 #include <linux/dcache.h>
52 #include <linux/mount.h>
53 #include <linux/namei.h>
54 #include <linux/pagemap.h>
55 #include <linux/slab.h>
56 #include <linux/exportfs.h>
57 
58 /*
59  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
60  * a file or fs handle.
61  *
62  * XFS_IOC_PATH_TO_FSHANDLE
63  *    returns fs handle for a mount point or path within that mount point
64  * XFS_IOC_FD_TO_HANDLE
65  *    returns full handle for a FD opened in user space
66  * XFS_IOC_PATH_TO_HANDLE
67  *    returns full handle for a path
68  */
69 int
70 xfs_find_handle(
71 	unsigned int		cmd,
72 	xfs_fsop_handlereq_t	*hreq)
73 {
74 	int			hsize;
75 	xfs_handle_t		handle;
76 	struct inode		*inode;
77 	struct fd		f = {NULL};
78 	struct path		path;
79 	int			error;
80 	struct xfs_inode	*ip;
81 
82 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
83 		f = fdget(hreq->fd);
84 		if (!f.file)
85 			return -EBADF;
86 		inode = file_inode(f.file);
87 	} else {
88 		error = user_lpath((const char __user *)hreq->path, &path);
89 		if (error)
90 			return error;
91 		inode = d_inode(path.dentry);
92 	}
93 	ip = XFS_I(inode);
94 
95 	/*
96 	 * We can only generate handles for inodes residing on a XFS filesystem,
97 	 * and only for regular files, directories or symbolic links.
98 	 */
99 	error = -EINVAL;
100 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
101 		goto out_put;
102 
103 	error = -EBADF;
104 	if (!S_ISREG(inode->i_mode) &&
105 	    !S_ISDIR(inode->i_mode) &&
106 	    !S_ISLNK(inode->i_mode))
107 		goto out_put;
108 
109 
110 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
111 
112 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
113 		/*
114 		 * This handle only contains an fsid, zero the rest.
115 		 */
116 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
117 		hsize = sizeof(xfs_fsid_t);
118 	} else {
119 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
120 					sizeof(handle.ha_fid.fid_len);
121 		handle.ha_fid.fid_pad = 0;
122 		handle.ha_fid.fid_gen = inode->i_generation;
123 		handle.ha_fid.fid_ino = ip->i_ino;
124 		hsize = sizeof(xfs_handle_t);
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 	uint		evmask,
315 	uint16_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 | XFS_MMAPLOCK_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, BREAK_UNMAP);
647 	if (error)
648 		goto out_unlock;
649 
650 	switch (bf->l_whence) {
651 	case 0: /*SEEK_SET*/
652 		break;
653 	case 1: /*SEEK_CUR*/
654 		bf->l_start += filp->f_pos;
655 		break;
656 	case 2: /*SEEK_END*/
657 		bf->l_start += XFS_ISIZE(ip);
658 		break;
659 	default:
660 		error = -EINVAL;
661 		goto out_unlock;
662 	}
663 
664 	/*
665 	 * length of <= 0 for resv/unresv/zero is invalid.  length for
666 	 * alloc/free is ignored completely and we have no idea what userspace
667 	 * might have set it to, so set it to zero to allow range
668 	 * checks to pass.
669 	 */
670 	switch (cmd) {
671 	case XFS_IOC_ZERO_RANGE:
672 	case XFS_IOC_RESVSP:
673 	case XFS_IOC_RESVSP64:
674 	case XFS_IOC_UNRESVSP:
675 	case XFS_IOC_UNRESVSP64:
676 		if (bf->l_len <= 0) {
677 			error = -EINVAL;
678 			goto out_unlock;
679 		}
680 		break;
681 	default:
682 		bf->l_len = 0;
683 		break;
684 	}
685 
686 	if (bf->l_start < 0 ||
687 	    bf->l_start > inode->i_sb->s_maxbytes ||
688 	    bf->l_start + bf->l_len < 0 ||
689 	    bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
690 		error = -EINVAL;
691 		goto out_unlock;
692 	}
693 
694 	switch (cmd) {
695 	case XFS_IOC_ZERO_RANGE:
696 		flags |= XFS_PREALLOC_SET;
697 		error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
698 		break;
699 	case XFS_IOC_RESVSP:
700 	case XFS_IOC_RESVSP64:
701 		flags |= XFS_PREALLOC_SET;
702 		error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
703 						XFS_BMAPI_PREALLOC);
704 		break;
705 	case XFS_IOC_UNRESVSP:
706 	case XFS_IOC_UNRESVSP64:
707 		error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
708 		break;
709 	case XFS_IOC_ALLOCSP:
710 	case XFS_IOC_ALLOCSP64:
711 	case XFS_IOC_FREESP:
712 	case XFS_IOC_FREESP64:
713 		flags |= XFS_PREALLOC_CLEAR;
714 		if (bf->l_start > XFS_ISIZE(ip)) {
715 			error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
716 					bf->l_start - XFS_ISIZE(ip), 0);
717 			if (error)
718 				goto out_unlock;
719 		}
720 
721 		iattr.ia_valid = ATTR_SIZE;
722 		iattr.ia_size = bf->l_start;
723 
724 		error = xfs_vn_setattr_size(file_dentry(filp), &iattr);
725 		break;
726 	default:
727 		ASSERT(0);
728 		error = -EINVAL;
729 	}
730 
731 	if (error)
732 		goto out_unlock;
733 
734 	error = xfs_update_prealloc_flags(ip, flags);
735 
736 out_unlock:
737 	xfs_iunlock(ip, iolock);
738 	mnt_drop_write_file(filp);
739 	return error;
740 }
741 
742 STATIC int
743 xfs_ioc_bulkstat(
744 	xfs_mount_t		*mp,
745 	unsigned int		cmd,
746 	void			__user *arg)
747 {
748 	xfs_fsop_bulkreq_t	bulkreq;
749 	int			count;	/* # of records returned */
750 	xfs_ino_t		inlast;	/* last inode number */
751 	int			done;
752 	int			error;
753 
754 	/* done = 1 if there are more stats to get and if bulkstat */
755 	/* should be called again (unused here, but used in dmapi) */
756 
757 	if (!capable(CAP_SYS_ADMIN))
758 		return -EPERM;
759 
760 	if (XFS_FORCED_SHUTDOWN(mp))
761 		return -EIO;
762 
763 	if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
764 		return -EFAULT;
765 
766 	if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
767 		return -EFAULT;
768 
769 	if ((count = bulkreq.icount) <= 0)
770 		return -EINVAL;
771 
772 	if (bulkreq.ubuffer == NULL)
773 		return -EINVAL;
774 
775 	if (cmd == XFS_IOC_FSINUMBERS)
776 		error = xfs_inumbers(mp, &inlast, &count,
777 					bulkreq.ubuffer, xfs_inumbers_fmt);
778 	else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
779 		error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
780 					sizeof(xfs_bstat_t), NULL, &done);
781 	else	/* XFS_IOC_FSBULKSTAT */
782 		error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
783 				     sizeof(xfs_bstat_t), bulkreq.ubuffer,
784 				     &done);
785 
786 	if (error)
787 		return error;
788 
789 	if (bulkreq.ocount != NULL) {
790 		if (copy_to_user(bulkreq.lastip, &inlast,
791 						sizeof(xfs_ino_t)))
792 			return -EFAULT;
793 
794 		if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
795 			return -EFAULT;
796 	}
797 
798 	return 0;
799 }
800 
801 STATIC int
802 xfs_ioc_fsgeometry_v1(
803 	xfs_mount_t		*mp,
804 	void			__user *arg)
805 {
806 	xfs_fsop_geom_t         fsgeo;
807 	int			error;
808 
809 	error = xfs_fs_geometry(&mp->m_sb, &fsgeo, 3);
810 	if (error)
811 		return error;
812 
813 	/*
814 	 * Caller should have passed an argument of type
815 	 * xfs_fsop_geom_v1_t.  This is a proper subset of the
816 	 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
817 	 */
818 	if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
819 		return -EFAULT;
820 	return 0;
821 }
822 
823 STATIC int
824 xfs_ioc_fsgeometry(
825 	xfs_mount_t		*mp,
826 	void			__user *arg)
827 {
828 	xfs_fsop_geom_t		fsgeo;
829 	int			error;
830 
831 	error = xfs_fs_geometry(&mp->m_sb, &fsgeo, 4);
832 	if (error)
833 		return error;
834 
835 	if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
836 		return -EFAULT;
837 	return 0;
838 }
839 
840 /*
841  * Linux extended inode flags interface.
842  */
843 
844 STATIC unsigned int
845 xfs_merge_ioc_xflags(
846 	unsigned int	flags,
847 	unsigned int	start)
848 {
849 	unsigned int	xflags = start;
850 
851 	if (flags & FS_IMMUTABLE_FL)
852 		xflags |= FS_XFLAG_IMMUTABLE;
853 	else
854 		xflags &= ~FS_XFLAG_IMMUTABLE;
855 	if (flags & FS_APPEND_FL)
856 		xflags |= FS_XFLAG_APPEND;
857 	else
858 		xflags &= ~FS_XFLAG_APPEND;
859 	if (flags & FS_SYNC_FL)
860 		xflags |= FS_XFLAG_SYNC;
861 	else
862 		xflags &= ~FS_XFLAG_SYNC;
863 	if (flags & FS_NOATIME_FL)
864 		xflags |= FS_XFLAG_NOATIME;
865 	else
866 		xflags &= ~FS_XFLAG_NOATIME;
867 	if (flags & FS_NODUMP_FL)
868 		xflags |= FS_XFLAG_NODUMP;
869 	else
870 		xflags &= ~FS_XFLAG_NODUMP;
871 
872 	return xflags;
873 }
874 
875 STATIC unsigned int
876 xfs_di2lxflags(
877 	uint16_t	di_flags)
878 {
879 	unsigned int	flags = 0;
880 
881 	if (di_flags & XFS_DIFLAG_IMMUTABLE)
882 		flags |= FS_IMMUTABLE_FL;
883 	if (di_flags & XFS_DIFLAG_APPEND)
884 		flags |= FS_APPEND_FL;
885 	if (di_flags & XFS_DIFLAG_SYNC)
886 		flags |= FS_SYNC_FL;
887 	if (di_flags & XFS_DIFLAG_NOATIME)
888 		flags |= FS_NOATIME_FL;
889 	if (di_flags & XFS_DIFLAG_NODUMP)
890 		flags |= FS_NODUMP_FL;
891 	return flags;
892 }
893 
894 STATIC int
895 xfs_ioc_fsgetxattr(
896 	xfs_inode_t		*ip,
897 	int			attr,
898 	void			__user *arg)
899 {
900 	struct fsxattr		fa;
901 
902 	memset(&fa, 0, sizeof(struct fsxattr));
903 
904 	xfs_ilock(ip, XFS_ILOCK_SHARED);
905 	fa.fsx_xflags = xfs_ip2xflags(ip);
906 	fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
907 	fa.fsx_cowextsize = ip->i_d.di_cowextsize <<
908 			ip->i_mount->m_sb.sb_blocklog;
909 	fa.fsx_projid = xfs_get_projid(ip);
910 
911 	if (attr) {
912 		if (ip->i_afp) {
913 			if (ip->i_afp->if_flags & XFS_IFEXTENTS)
914 				fa.fsx_nextents = xfs_iext_count(ip->i_afp);
915 			else
916 				fa.fsx_nextents = ip->i_d.di_anextents;
917 		} else
918 			fa.fsx_nextents = 0;
919 	} else {
920 		if (ip->i_df.if_flags & XFS_IFEXTENTS)
921 			fa.fsx_nextents = xfs_iext_count(&ip->i_df);
922 		else
923 			fa.fsx_nextents = ip->i_d.di_nextents;
924 	}
925 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
926 
927 	if (copy_to_user(arg, &fa, sizeof(fa)))
928 		return -EFAULT;
929 	return 0;
930 }
931 
932 STATIC uint16_t
933 xfs_flags2diflags(
934 	struct xfs_inode	*ip,
935 	unsigned int		xflags)
936 {
937 	/* can't set PREALLOC this way, just preserve it */
938 	uint16_t		di_flags =
939 		(ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
940 
941 	if (xflags & FS_XFLAG_IMMUTABLE)
942 		di_flags |= XFS_DIFLAG_IMMUTABLE;
943 	if (xflags & FS_XFLAG_APPEND)
944 		di_flags |= XFS_DIFLAG_APPEND;
945 	if (xflags & FS_XFLAG_SYNC)
946 		di_flags |= XFS_DIFLAG_SYNC;
947 	if (xflags & FS_XFLAG_NOATIME)
948 		di_flags |= XFS_DIFLAG_NOATIME;
949 	if (xflags & FS_XFLAG_NODUMP)
950 		di_flags |= XFS_DIFLAG_NODUMP;
951 	if (xflags & FS_XFLAG_NODEFRAG)
952 		di_flags |= XFS_DIFLAG_NODEFRAG;
953 	if (xflags & FS_XFLAG_FILESTREAM)
954 		di_flags |= XFS_DIFLAG_FILESTREAM;
955 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
956 		if (xflags & FS_XFLAG_RTINHERIT)
957 			di_flags |= XFS_DIFLAG_RTINHERIT;
958 		if (xflags & FS_XFLAG_NOSYMLINKS)
959 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
960 		if (xflags & FS_XFLAG_EXTSZINHERIT)
961 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
962 		if (xflags & FS_XFLAG_PROJINHERIT)
963 			di_flags |= XFS_DIFLAG_PROJINHERIT;
964 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
965 		if (xflags & FS_XFLAG_REALTIME)
966 			di_flags |= XFS_DIFLAG_REALTIME;
967 		if (xflags & FS_XFLAG_EXTSIZE)
968 			di_flags |= XFS_DIFLAG_EXTSIZE;
969 	}
970 
971 	return di_flags;
972 }
973 
974 STATIC uint64_t
975 xfs_flags2diflags2(
976 	struct xfs_inode	*ip,
977 	unsigned int		xflags)
978 {
979 	uint64_t		di_flags2 =
980 		(ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK);
981 
982 	if (xflags & FS_XFLAG_DAX)
983 		di_flags2 |= XFS_DIFLAG2_DAX;
984 	if (xflags & FS_XFLAG_COWEXTSIZE)
985 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
986 
987 	return di_flags2;
988 }
989 
990 STATIC void
991 xfs_diflags_to_linux(
992 	struct xfs_inode	*ip)
993 {
994 	struct inode		*inode = VFS_I(ip);
995 	unsigned int		xflags = xfs_ip2xflags(ip);
996 
997 	if (xflags & FS_XFLAG_IMMUTABLE)
998 		inode->i_flags |= S_IMMUTABLE;
999 	else
1000 		inode->i_flags &= ~S_IMMUTABLE;
1001 	if (xflags & FS_XFLAG_APPEND)
1002 		inode->i_flags |= S_APPEND;
1003 	else
1004 		inode->i_flags &= ~S_APPEND;
1005 	if (xflags & FS_XFLAG_SYNC)
1006 		inode->i_flags |= S_SYNC;
1007 	else
1008 		inode->i_flags &= ~S_SYNC;
1009 	if (xflags & FS_XFLAG_NOATIME)
1010 		inode->i_flags |= S_NOATIME;
1011 	else
1012 		inode->i_flags &= ~S_NOATIME;
1013 #if 0	/* disabled until the flag switching races are sorted out */
1014 	if (xflags & FS_XFLAG_DAX)
1015 		inode->i_flags |= S_DAX;
1016 	else
1017 		inode->i_flags &= ~S_DAX;
1018 #endif
1019 }
1020 
1021 static int
1022 xfs_ioctl_setattr_xflags(
1023 	struct xfs_trans	*tp,
1024 	struct xfs_inode	*ip,
1025 	struct fsxattr		*fa)
1026 {
1027 	struct xfs_mount	*mp = ip->i_mount;
1028 	uint64_t		di_flags2;
1029 
1030 	/* Can't change realtime flag if any extents are allocated. */
1031 	if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1032 	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1033 		return -EINVAL;
1034 
1035 	/* If realtime flag is set then must have realtime device */
1036 	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1037 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1038 		    (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1039 			return -EINVAL;
1040 	}
1041 
1042 	/* Clear reflink if we are actually able to set the rt flag. */
1043 	if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1044 		ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
1045 
1046 	/* Don't allow us to set DAX mode for a reflinked file for now. */
1047 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1048 		return -EINVAL;
1049 
1050 	/*
1051 	 * Can't modify an immutable/append-only file unless
1052 	 * we have appropriate permission.
1053 	 */
1054 	if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1055 	     (fa->fsx_xflags & (FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))) &&
1056 	    !capable(CAP_LINUX_IMMUTABLE))
1057 		return -EPERM;
1058 
1059 	/* diflags2 only valid for v3 inodes. */
1060 	di_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1061 	if (di_flags2 && ip->i_d.di_version < 3)
1062 		return -EINVAL;
1063 
1064 	ip->i_d.di_flags = xfs_flags2diflags(ip, fa->fsx_xflags);
1065 	ip->i_d.di_flags2 = di_flags2;
1066 
1067 	xfs_diflags_to_linux(ip);
1068 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1069 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1070 	XFS_STATS_INC(mp, xs_ig_attrchg);
1071 	return 0;
1072 }
1073 
1074 /*
1075  * If we are changing DAX flags, we have to ensure the file is clean and any
1076  * cached objects in the address space are invalidated and removed. This
1077  * requires us to lock out other IO and page faults similar to a truncate
1078  * operation. The locks need to be held until the transaction has been committed
1079  * so that the cache invalidation is atomic with respect to the DAX flag
1080  * manipulation.
1081  */
1082 static int
1083 xfs_ioctl_setattr_dax_invalidate(
1084 	struct xfs_inode	*ip,
1085 	struct fsxattr		*fa,
1086 	int			*join_flags)
1087 {
1088 	struct inode		*inode = VFS_I(ip);
1089 	struct super_block	*sb = inode->i_sb;
1090 	int			error;
1091 
1092 	*join_flags = 0;
1093 
1094 	/*
1095 	 * It is only valid to set the DAX flag on regular files and
1096 	 * directories on filesystems where the block size is equal to the page
1097 	 * size. On directories it serves as an inherit hint.
1098 	 */
1099 	if (fa->fsx_xflags & FS_XFLAG_DAX) {
1100 		if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1101 			return -EINVAL;
1102 		if (!bdev_dax_supported(xfs_find_bdev_for_inode(VFS_I(ip)),
1103 				sb->s_blocksize))
1104 			return -EINVAL;
1105 	}
1106 
1107 	/* If the DAX state is not changing, we have nothing to do here. */
1108 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
1109 		return 0;
1110 	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
1111 		return 0;
1112 
1113 	/* lock, flush and invalidate mapping in preparation for flag change */
1114 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1115 	error = filemap_write_and_wait(inode->i_mapping);
1116 	if (error)
1117 		goto out_unlock;
1118 	error = invalidate_inode_pages2(inode->i_mapping);
1119 	if (error)
1120 		goto out_unlock;
1121 
1122 	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
1123 	return 0;
1124 
1125 out_unlock:
1126 	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1127 	return error;
1128 
1129 }
1130 
1131 /*
1132  * Set up the transaction structure for the setattr operation, checking that we
1133  * have permission to do so. On success, return a clean transaction and the
1134  * inode locked exclusively ready for further operation specific checks. On
1135  * failure, return an error without modifying or locking the inode.
1136  *
1137  * The inode might already be IO locked on call. If this is the case, it is
1138  * indicated in @join_flags and we take full responsibility for ensuring they
1139  * are unlocked from now on. Hence if we have an error here, we still have to
1140  * unlock them. Otherwise, once they are joined to the transaction, they will
1141  * be unlocked on commit/cancel.
1142  */
1143 static struct xfs_trans *
1144 xfs_ioctl_setattr_get_trans(
1145 	struct xfs_inode	*ip,
1146 	int			join_flags)
1147 {
1148 	struct xfs_mount	*mp = ip->i_mount;
1149 	struct xfs_trans	*tp;
1150 	int			error = -EROFS;
1151 
1152 	if (mp->m_flags & XFS_MOUNT_RDONLY)
1153 		goto out_unlock;
1154 	error = -EIO;
1155 	if (XFS_FORCED_SHUTDOWN(mp))
1156 		goto out_unlock;
1157 
1158 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
1159 	if (error)
1160 		return ERR_PTR(error);
1161 
1162 	xfs_ilock(ip, XFS_ILOCK_EXCL);
1163 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
1164 	join_flags = 0;
1165 
1166 	/*
1167 	 * CAP_FOWNER overrides the following restrictions:
1168 	 *
1169 	 * The user ID of the calling process must be equal to the file owner
1170 	 * ID, except in cases where the CAP_FSETID capability is applicable.
1171 	 */
1172 	if (!inode_owner_or_capable(VFS_I(ip))) {
1173 		error = -EPERM;
1174 		goto out_cancel;
1175 	}
1176 
1177 	if (mp->m_flags & XFS_MOUNT_WSYNC)
1178 		xfs_trans_set_sync(tp);
1179 
1180 	return tp;
1181 
1182 out_cancel:
1183 	xfs_trans_cancel(tp);
1184 out_unlock:
1185 	if (join_flags)
1186 		xfs_iunlock(ip, join_flags);
1187 	return ERR_PTR(error);
1188 }
1189 
1190 /*
1191  * extent size hint validation is somewhat cumbersome. Rules are:
1192  *
1193  * 1. extent size hint is only valid for directories and regular files
1194  * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1195  * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1196  * 4. can only be changed on regular files if no extents are allocated
1197  * 5. can be changed on directories at any time
1198  * 6. extsize hint of 0 turns off hints, clears inode flags.
1199  * 7. Extent size must be a multiple of the appropriate block size.
1200  * 8. for non-realtime files, the extent size hint must be limited
1201  *    to half the AG size to avoid alignment extending the extent beyond the
1202  *    limits of the AG.
1203  *
1204  * Please keep this function in sync with xfs_scrub_inode_extsize.
1205  */
1206 static int
1207 xfs_ioctl_setattr_check_extsize(
1208 	struct xfs_inode	*ip,
1209 	struct fsxattr		*fa)
1210 {
1211 	struct xfs_mount	*mp = ip->i_mount;
1212 
1213 	if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(VFS_I(ip)->i_mode))
1214 		return -EINVAL;
1215 
1216 	if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
1217 	    !S_ISDIR(VFS_I(ip)->i_mode))
1218 		return -EINVAL;
1219 
1220 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_d.di_nextents &&
1221 	    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1222 		return -EINVAL;
1223 
1224 	if (fa->fsx_extsize != 0) {
1225 		xfs_extlen_t    size;
1226 		xfs_fsblock_t   extsize_fsb;
1227 
1228 		extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1229 		if (extsize_fsb > MAXEXTLEN)
1230 			return -EINVAL;
1231 
1232 		if (XFS_IS_REALTIME_INODE(ip) ||
1233 		    (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1234 			size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1235 		} else {
1236 			size = mp->m_sb.sb_blocksize;
1237 			if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1238 				return -EINVAL;
1239 		}
1240 
1241 		if (fa->fsx_extsize % size)
1242 			return -EINVAL;
1243 	} else
1244 		fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
1245 
1246 	return 0;
1247 }
1248 
1249 /*
1250  * CoW extent size hint validation rules are:
1251  *
1252  * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
1253  *    The inode does not have to have any shared blocks, but it must be a v3.
1254  * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
1255  *    for a directory, the hint is propagated to new files.
1256  * 3. Can be changed on files & directories at any time.
1257  * 4. CoW extsize hint of 0 turns off hints, clears inode flags.
1258  * 5. Extent size must be a multiple of the appropriate block size.
1259  * 6. The extent size hint must be limited to half the AG size to avoid
1260  *    alignment extending the extent beyond the limits of the AG.
1261  *
1262  * Please keep this function in sync with xfs_scrub_inode_cowextsize.
1263  */
1264 static int
1265 xfs_ioctl_setattr_check_cowextsize(
1266 	struct xfs_inode	*ip,
1267 	struct fsxattr		*fa)
1268 {
1269 	struct xfs_mount	*mp = ip->i_mount;
1270 
1271 	if (!(fa->fsx_xflags & FS_XFLAG_COWEXTSIZE))
1272 		return 0;
1273 
1274 	if (!xfs_sb_version_hasreflink(&ip->i_mount->m_sb) ||
1275 	    ip->i_d.di_version != 3)
1276 		return -EINVAL;
1277 
1278 	if (!S_ISREG(VFS_I(ip)->i_mode) && !S_ISDIR(VFS_I(ip)->i_mode))
1279 		return -EINVAL;
1280 
1281 	if (fa->fsx_cowextsize != 0) {
1282 		xfs_extlen_t    size;
1283 		xfs_fsblock_t   cowextsize_fsb;
1284 
1285 		cowextsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1286 		if (cowextsize_fsb > MAXEXTLEN)
1287 			return -EINVAL;
1288 
1289 		size = mp->m_sb.sb_blocksize;
1290 		if (cowextsize_fsb > mp->m_sb.sb_agblocks / 2)
1291 			return -EINVAL;
1292 
1293 		if (fa->fsx_cowextsize % size)
1294 			return -EINVAL;
1295 	} else
1296 		fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
1297 
1298 	return 0;
1299 }
1300 
1301 static int
1302 xfs_ioctl_setattr_check_projid(
1303 	struct xfs_inode	*ip,
1304 	struct fsxattr		*fa)
1305 {
1306 	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1307 	if (fa->fsx_projid > (uint16_t)-1 &&
1308 	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1309 		return -EINVAL;
1310 
1311 	/*
1312 	 * Project Quota ID state is only allowed to change from within the init
1313 	 * namespace. Enforce that restriction only if we are trying to change
1314 	 * the quota ID state. Everything else is allowed in user namespaces.
1315 	 */
1316 	if (current_user_ns() == &init_user_ns)
1317 		return 0;
1318 
1319 	if (xfs_get_projid(ip) != fa->fsx_projid)
1320 		return -EINVAL;
1321 	if ((fa->fsx_xflags & FS_XFLAG_PROJINHERIT) !=
1322 	    (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1323 		return -EINVAL;
1324 
1325 	return 0;
1326 }
1327 
1328 STATIC int
1329 xfs_ioctl_setattr(
1330 	xfs_inode_t		*ip,
1331 	struct fsxattr		*fa)
1332 {
1333 	struct xfs_mount	*mp = ip->i_mount;
1334 	struct xfs_trans	*tp;
1335 	struct xfs_dquot	*udqp = NULL;
1336 	struct xfs_dquot	*pdqp = NULL;
1337 	struct xfs_dquot	*olddquot = NULL;
1338 	int			code;
1339 	int			join_flags = 0;
1340 
1341 	trace_xfs_ioctl_setattr(ip);
1342 
1343 	code = xfs_ioctl_setattr_check_projid(ip, fa);
1344 	if (code)
1345 		return code;
1346 
1347 	/*
1348 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1349 	 * before we start any other transactions. Trying to do this later
1350 	 * is messy. We don't care to take a readlock to look at the ids
1351 	 * in inode here, because we can't hold it across the trans_reserve.
1352 	 * If the IDs do change before we take the ilock, we're covered
1353 	 * because the i_*dquot fields will get updated anyway.
1354 	 */
1355 	if (XFS_IS_QUOTA_ON(mp)) {
1356 		code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1357 					 ip->i_d.di_gid, fa->fsx_projid,
1358 					 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1359 		if (code)
1360 			return code;
1361 	}
1362 
1363 	/*
1364 	 * Changing DAX config may require inode locking for mapping
1365 	 * invalidation. These need to be held all the way to transaction commit
1366 	 * or cancel time, so need to be passed through to
1367 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1368 	 * appropriately.
1369 	 */
1370 	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
1371 	if (code)
1372 		goto error_free_dquots;
1373 
1374 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1375 	if (IS_ERR(tp)) {
1376 		code = PTR_ERR(tp);
1377 		goto error_free_dquots;
1378 	}
1379 
1380 
1381 	if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1382 	    xfs_get_projid(ip) != fa->fsx_projid) {
1383 		code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1384 				capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
1385 		if (code)	/* out of quota */
1386 			goto error_trans_cancel;
1387 	}
1388 
1389 	code = xfs_ioctl_setattr_check_extsize(ip, fa);
1390 	if (code)
1391 		goto error_trans_cancel;
1392 
1393 	code = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1394 	if (code)
1395 		goto error_trans_cancel;
1396 
1397 	code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1398 	if (code)
1399 		goto error_trans_cancel;
1400 
1401 	/*
1402 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1403 	 * overrides the following restrictions:
1404 	 *
1405 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1406 	 * successful return from chown()
1407 	 */
1408 
1409 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1410 	    !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1411 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1412 
1413 	/* Change the ownerships and register project quota modifications */
1414 	if (xfs_get_projid(ip) != fa->fsx_projid) {
1415 		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1416 			olddquot = xfs_qm_vop_chown(tp, ip,
1417 						&ip->i_pdquot, pdqp);
1418 		}
1419 		ASSERT(ip->i_d.di_version > 1);
1420 		xfs_set_projid(ip, fa->fsx_projid);
1421 	}
1422 
1423 	/*
1424 	 * Only set the extent size hint if we've already determined that the
1425 	 * extent size hint should be set on the inode. If no extent size flags
1426 	 * are set on the inode then unconditionally clear the extent size hint.
1427 	 */
1428 	if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1429 		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1430 	else
1431 		ip->i_d.di_extsize = 0;
1432 	if (ip->i_d.di_version == 3 &&
1433 	    (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
1434 		ip->i_d.di_cowextsize = fa->fsx_cowextsize >>
1435 				mp->m_sb.sb_blocklog;
1436 	else
1437 		ip->i_d.di_cowextsize = 0;
1438 
1439 	code = xfs_trans_commit(tp);
1440 
1441 	/*
1442 	 * Release any dquot(s) the inode had kept before chown.
1443 	 */
1444 	xfs_qm_dqrele(olddquot);
1445 	xfs_qm_dqrele(udqp);
1446 	xfs_qm_dqrele(pdqp);
1447 
1448 	return code;
1449 
1450 error_trans_cancel:
1451 	xfs_trans_cancel(tp);
1452 error_free_dquots:
1453 	xfs_qm_dqrele(udqp);
1454 	xfs_qm_dqrele(pdqp);
1455 	return code;
1456 }
1457 
1458 STATIC int
1459 xfs_ioc_fssetxattr(
1460 	xfs_inode_t		*ip,
1461 	struct file		*filp,
1462 	void			__user *arg)
1463 {
1464 	struct fsxattr		fa;
1465 	int error;
1466 
1467 	if (copy_from_user(&fa, arg, sizeof(fa)))
1468 		return -EFAULT;
1469 
1470 	error = mnt_want_write_file(filp);
1471 	if (error)
1472 		return error;
1473 	error = xfs_ioctl_setattr(ip, &fa);
1474 	mnt_drop_write_file(filp);
1475 	return error;
1476 }
1477 
1478 STATIC int
1479 xfs_ioc_getxflags(
1480 	xfs_inode_t		*ip,
1481 	void			__user *arg)
1482 {
1483 	unsigned int		flags;
1484 
1485 	flags = xfs_di2lxflags(ip->i_d.di_flags);
1486 	if (copy_to_user(arg, &flags, sizeof(flags)))
1487 		return -EFAULT;
1488 	return 0;
1489 }
1490 
1491 STATIC int
1492 xfs_ioc_setxflags(
1493 	struct xfs_inode	*ip,
1494 	struct file		*filp,
1495 	void			__user *arg)
1496 {
1497 	struct xfs_trans	*tp;
1498 	struct fsxattr		fa;
1499 	unsigned int		flags;
1500 	int			join_flags = 0;
1501 	int			error;
1502 
1503 	if (copy_from_user(&flags, arg, sizeof(flags)))
1504 		return -EFAULT;
1505 
1506 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1507 		      FS_NOATIME_FL | FS_NODUMP_FL | \
1508 		      FS_SYNC_FL))
1509 		return -EOPNOTSUPP;
1510 
1511 	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1512 
1513 	error = mnt_want_write_file(filp);
1514 	if (error)
1515 		return error;
1516 
1517 	/*
1518 	 * Changing DAX config may require inode locking for mapping
1519 	 * invalidation. These need to be held all the way to transaction commit
1520 	 * or cancel time, so need to be passed through to
1521 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1522 	 * appropriately.
1523 	 */
1524 	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
1525 	if (error)
1526 		goto out_drop_write;
1527 
1528 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1529 	if (IS_ERR(tp)) {
1530 		error = PTR_ERR(tp);
1531 		goto out_drop_write;
1532 	}
1533 
1534 	error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1535 	if (error) {
1536 		xfs_trans_cancel(tp);
1537 		goto out_drop_write;
1538 	}
1539 
1540 	error = xfs_trans_commit(tp);
1541 out_drop_write:
1542 	mnt_drop_write_file(filp);
1543 	return error;
1544 }
1545 
1546 static bool
1547 xfs_getbmap_format(
1548 	struct kgetbmap		*p,
1549 	struct getbmapx __user	*u,
1550 	size_t			recsize)
1551 {
1552 	if (put_user(p->bmv_offset, &u->bmv_offset) ||
1553 	    put_user(p->bmv_block, &u->bmv_block) ||
1554 	    put_user(p->bmv_length, &u->bmv_length) ||
1555 	    put_user(0, &u->bmv_count) ||
1556 	    put_user(0, &u->bmv_entries))
1557 		return false;
1558 	if (recsize < sizeof(struct getbmapx))
1559 		return true;
1560 	if (put_user(0, &u->bmv_iflags) ||
1561 	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
1562 	    put_user(0, &u->bmv_unused1) ||
1563 	    put_user(0, &u->bmv_unused2))
1564 		return false;
1565 	return true;
1566 }
1567 
1568 STATIC int
1569 xfs_ioc_getbmap(
1570 	struct file		*file,
1571 	unsigned int		cmd,
1572 	void			__user *arg)
1573 {
1574 	struct getbmapx		bmx = { 0 };
1575 	struct kgetbmap		*buf;
1576 	size_t			recsize;
1577 	int			error, i;
1578 
1579 	switch (cmd) {
1580 	case XFS_IOC_GETBMAPA:
1581 		bmx.bmv_iflags = BMV_IF_ATTRFORK;
1582 		/*FALLTHRU*/
1583 	case XFS_IOC_GETBMAP:
1584 		if (file->f_mode & FMODE_NOCMTIME)
1585 			bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1586 		/* struct getbmap is a strict subset of struct getbmapx. */
1587 		recsize = sizeof(struct getbmap);
1588 		break;
1589 	case XFS_IOC_GETBMAPX:
1590 		recsize = sizeof(struct getbmapx);
1591 		break;
1592 	default:
1593 		return -EINVAL;
1594 	}
1595 
1596 	if (copy_from_user(&bmx, arg, recsize))
1597 		return -EFAULT;
1598 
1599 	if (bmx.bmv_count < 2)
1600 		return -EINVAL;
1601 	if (bmx.bmv_count > ULONG_MAX / recsize)
1602 		return -ENOMEM;
1603 
1604 	buf = kmem_zalloc_large(bmx.bmv_count * sizeof(*buf), 0);
1605 	if (!buf)
1606 		return -ENOMEM;
1607 
1608 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1609 	if (error)
1610 		goto out_free_buf;
1611 
1612 	error = -EFAULT;
1613 	if (copy_to_user(arg, &bmx, recsize))
1614 		goto out_free_buf;
1615 	arg += recsize;
1616 
1617 	for (i = 0; i < bmx.bmv_entries; i++) {
1618 		if (!xfs_getbmap_format(buf + i, arg, recsize))
1619 			goto out_free_buf;
1620 		arg += recsize;
1621 	}
1622 
1623 	error = 0;
1624 out_free_buf:
1625 	kmem_free(buf);
1626 	return 0;
1627 }
1628 
1629 struct getfsmap_info {
1630 	struct xfs_mount	*mp;
1631 	struct fsmap_head __user *data;
1632 	unsigned int		idx;
1633 	__u32			last_flags;
1634 };
1635 
1636 STATIC int
1637 xfs_getfsmap_format(struct xfs_fsmap *xfm, void *priv)
1638 {
1639 	struct getfsmap_info	*info = priv;
1640 	struct fsmap		fm;
1641 
1642 	trace_xfs_getfsmap_mapping(info->mp, xfm);
1643 
1644 	info->last_flags = xfm->fmr_flags;
1645 	xfs_fsmap_from_internal(&fm, xfm);
1646 	if (copy_to_user(&info->data->fmh_recs[info->idx++], &fm,
1647 			sizeof(struct fsmap)))
1648 		return -EFAULT;
1649 
1650 	return 0;
1651 }
1652 
1653 STATIC int
1654 xfs_ioc_getfsmap(
1655 	struct xfs_inode	*ip,
1656 	struct fsmap_head	__user *arg)
1657 {
1658 	struct getfsmap_info	info = { NULL };
1659 	struct xfs_fsmap_head	xhead = {0};
1660 	struct fsmap_head	head;
1661 	bool			aborted = false;
1662 	int			error;
1663 
1664 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1665 		return -EFAULT;
1666 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1667 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1668 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1669 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1670 		       sizeof(head.fmh_keys[1].fmr_reserved)))
1671 		return -EINVAL;
1672 
1673 	xhead.fmh_iflags = head.fmh_iflags;
1674 	xhead.fmh_count = head.fmh_count;
1675 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1676 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1677 
1678 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1679 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1680 
1681 	info.mp = ip->i_mount;
1682 	info.data = arg;
1683 	error = xfs_getfsmap(ip->i_mount, &xhead, xfs_getfsmap_format, &info);
1684 	if (error == XFS_BTREE_QUERY_RANGE_ABORT) {
1685 		error = 0;
1686 		aborted = true;
1687 	} else if (error)
1688 		return error;
1689 
1690 	/* If we didn't abort, set the "last" flag in the last fmx */
1691 	if (!aborted && info.idx) {
1692 		info.last_flags |= FMR_OF_LAST;
1693 		if (copy_to_user(&info.data->fmh_recs[info.idx - 1].fmr_flags,
1694 				&info.last_flags, sizeof(info.last_flags)))
1695 			return -EFAULT;
1696 	}
1697 
1698 	/* copy back header */
1699 	head.fmh_entries = xhead.fmh_entries;
1700 	head.fmh_oflags = xhead.fmh_oflags;
1701 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
1702 		return -EFAULT;
1703 
1704 	return 0;
1705 }
1706 
1707 STATIC int
1708 xfs_ioc_scrub_metadata(
1709 	struct xfs_inode		*ip,
1710 	void				__user *arg)
1711 {
1712 	struct xfs_scrub_metadata	scrub;
1713 	int				error;
1714 
1715 	if (!capable(CAP_SYS_ADMIN))
1716 		return -EPERM;
1717 
1718 	if (copy_from_user(&scrub, arg, sizeof(scrub)))
1719 		return -EFAULT;
1720 
1721 	error = xfs_scrub_metadata(ip, &scrub);
1722 	if (error)
1723 		return error;
1724 
1725 	if (copy_to_user(arg, &scrub, sizeof(scrub)))
1726 		return -EFAULT;
1727 
1728 	return 0;
1729 }
1730 
1731 int
1732 xfs_ioc_swapext(
1733 	xfs_swapext_t	*sxp)
1734 {
1735 	xfs_inode_t     *ip, *tip;
1736 	struct fd	f, tmp;
1737 	int		error = 0;
1738 
1739 	/* Pull information for the target fd */
1740 	f = fdget((int)sxp->sx_fdtarget);
1741 	if (!f.file) {
1742 		error = -EINVAL;
1743 		goto out;
1744 	}
1745 
1746 	if (!(f.file->f_mode & FMODE_WRITE) ||
1747 	    !(f.file->f_mode & FMODE_READ) ||
1748 	    (f.file->f_flags & O_APPEND)) {
1749 		error = -EBADF;
1750 		goto out_put_file;
1751 	}
1752 
1753 	tmp = fdget((int)sxp->sx_fdtmp);
1754 	if (!tmp.file) {
1755 		error = -EINVAL;
1756 		goto out_put_file;
1757 	}
1758 
1759 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1760 	    !(tmp.file->f_mode & FMODE_READ) ||
1761 	    (tmp.file->f_flags & O_APPEND)) {
1762 		error = -EBADF;
1763 		goto out_put_tmp_file;
1764 	}
1765 
1766 	if (IS_SWAPFILE(file_inode(f.file)) ||
1767 	    IS_SWAPFILE(file_inode(tmp.file))) {
1768 		error = -EINVAL;
1769 		goto out_put_tmp_file;
1770 	}
1771 
1772 	/*
1773 	 * We need to ensure that the fds passed in point to XFS inodes
1774 	 * before we cast and access them as XFS structures as we have no
1775 	 * control over what the user passes us here.
1776 	 */
1777 	if (f.file->f_op != &xfs_file_operations ||
1778 	    tmp.file->f_op != &xfs_file_operations) {
1779 		error = -EINVAL;
1780 		goto out_put_tmp_file;
1781 	}
1782 
1783 	ip = XFS_I(file_inode(f.file));
1784 	tip = XFS_I(file_inode(tmp.file));
1785 
1786 	if (ip->i_mount != tip->i_mount) {
1787 		error = -EINVAL;
1788 		goto out_put_tmp_file;
1789 	}
1790 
1791 	if (ip->i_ino == tip->i_ino) {
1792 		error = -EINVAL;
1793 		goto out_put_tmp_file;
1794 	}
1795 
1796 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1797 		error = -EIO;
1798 		goto out_put_tmp_file;
1799 	}
1800 
1801 	error = xfs_swap_extents(ip, tip, sxp);
1802 
1803  out_put_tmp_file:
1804 	fdput(tmp);
1805  out_put_file:
1806 	fdput(f);
1807  out:
1808 	return error;
1809 }
1810 
1811 static int
1812 xfs_ioc_getlabel(
1813 	struct xfs_mount	*mp,
1814 	char			__user *user_label)
1815 {
1816 	struct xfs_sb		*sbp = &mp->m_sb;
1817 	char			label[XFSLABEL_MAX + 1];
1818 
1819 	/* Paranoia */
1820 	BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1821 
1822 	spin_lock(&mp->m_sb_lock);
1823 	strncpy(label, sbp->sb_fname, sizeof(sbp->sb_fname));
1824 	spin_unlock(&mp->m_sb_lock);
1825 
1826 	/* xfs on-disk label is 12 chars, be sure we send a null to user */
1827 	label[XFSLABEL_MAX] = '\0';
1828 	if (copy_to_user(user_label, label, sizeof(sbp->sb_fname)))
1829 		return -EFAULT;
1830 	return 0;
1831 }
1832 
1833 static int
1834 xfs_ioc_setlabel(
1835 	struct file		*filp,
1836 	struct xfs_mount	*mp,
1837 	char			__user *newlabel)
1838 {
1839 	struct xfs_sb		*sbp = &mp->m_sb;
1840 	char			label[XFSLABEL_MAX + 1];
1841 	size_t			len;
1842 	int			error;
1843 
1844 	if (!capable(CAP_SYS_ADMIN))
1845 		return -EPERM;
1846 	/*
1847 	 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1848 	 * smaller, at 12 bytes.  We copy one more to be sure we find the
1849 	 * (required) NULL character to test the incoming label length.
1850 	 * NB: The on disk label doesn't need to be null terminated.
1851 	 */
1852 	if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1853 		return -EFAULT;
1854 	len = strnlen(label, XFSLABEL_MAX + 1);
1855 	if (len > sizeof(sbp->sb_fname))
1856 		return -EINVAL;
1857 
1858 	error = mnt_want_write_file(filp);
1859 	if (error)
1860 		return error;
1861 
1862 	spin_lock(&mp->m_sb_lock);
1863 	memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1864 	strncpy(sbp->sb_fname, label, sizeof(sbp->sb_fname));
1865 	spin_unlock(&mp->m_sb_lock);
1866 
1867 	/*
1868 	 * Now we do several things to satisfy userspace.
1869 	 * In addition to normal logging of the primary superblock, we also
1870 	 * immediately write these changes to sector zero for the primary, then
1871 	 * update all backup supers (as xfs_db does for a label change), then
1872 	 * invalidate the block device page cache.  This is so that any prior
1873 	 * buffered reads from userspace (i.e. from blkid) are invalidated,
1874 	 * and userspace will see the newly-written label.
1875 	 */
1876 	error = xfs_sync_sb_buf(mp);
1877 	if (error)
1878 		goto out;
1879 	/*
1880 	 * growfs also updates backup supers so lock against that.
1881 	 */
1882 	mutex_lock(&mp->m_growlock);
1883 	error = xfs_update_secondary_sbs(mp);
1884 	mutex_unlock(&mp->m_growlock);
1885 
1886 	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1887 
1888 out:
1889 	mnt_drop_write_file(filp);
1890 	return error;
1891 }
1892 
1893 /*
1894  * Note: some of the ioctl's return positive numbers as a
1895  * byte count indicating success, such as readlink_by_handle.
1896  * So we don't "sign flip" like most other routines.  This means
1897  * true errors need to be returned as a negative value.
1898  */
1899 long
1900 xfs_file_ioctl(
1901 	struct file		*filp,
1902 	unsigned int		cmd,
1903 	unsigned long		p)
1904 {
1905 	struct inode		*inode = file_inode(filp);
1906 	struct xfs_inode	*ip = XFS_I(inode);
1907 	struct xfs_mount	*mp = ip->i_mount;
1908 	void			__user *arg = (void __user *)p;
1909 	int			error;
1910 
1911 	trace_xfs_file_ioctl(ip);
1912 
1913 	switch (cmd) {
1914 	case FITRIM:
1915 		return xfs_ioc_trim(mp, arg);
1916 	case FS_IOC_GETFSLABEL:
1917 		return xfs_ioc_getlabel(mp, arg);
1918 	case FS_IOC_SETFSLABEL:
1919 		return xfs_ioc_setlabel(filp, mp, arg);
1920 	case XFS_IOC_ALLOCSP:
1921 	case XFS_IOC_FREESP:
1922 	case XFS_IOC_RESVSP:
1923 	case XFS_IOC_UNRESVSP:
1924 	case XFS_IOC_ALLOCSP64:
1925 	case XFS_IOC_FREESP64:
1926 	case XFS_IOC_RESVSP64:
1927 	case XFS_IOC_UNRESVSP64:
1928 	case XFS_IOC_ZERO_RANGE: {
1929 		xfs_flock64_t		bf;
1930 
1931 		if (copy_from_user(&bf, arg, sizeof(bf)))
1932 			return -EFAULT;
1933 		return xfs_ioc_space(filp, cmd, &bf);
1934 	}
1935 	case XFS_IOC_DIOINFO: {
1936 		struct dioattr	da;
1937 		xfs_buftarg_t	*target =
1938 			XFS_IS_REALTIME_INODE(ip) ?
1939 			mp->m_rtdev_targp : mp->m_ddev_targp;
1940 
1941 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1942 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1943 
1944 		if (copy_to_user(arg, &da, sizeof(da)))
1945 			return -EFAULT;
1946 		return 0;
1947 	}
1948 
1949 	case XFS_IOC_FSBULKSTAT_SINGLE:
1950 	case XFS_IOC_FSBULKSTAT:
1951 	case XFS_IOC_FSINUMBERS:
1952 		return xfs_ioc_bulkstat(mp, cmd, arg);
1953 
1954 	case XFS_IOC_FSGEOMETRY_V1:
1955 		return xfs_ioc_fsgeometry_v1(mp, arg);
1956 
1957 	case XFS_IOC_FSGEOMETRY:
1958 		return xfs_ioc_fsgeometry(mp, arg);
1959 
1960 	case XFS_IOC_GETVERSION:
1961 		return put_user(inode->i_generation, (int __user *)arg);
1962 
1963 	case XFS_IOC_FSGETXATTR:
1964 		return xfs_ioc_fsgetxattr(ip, 0, arg);
1965 	case XFS_IOC_FSGETXATTRA:
1966 		return xfs_ioc_fsgetxattr(ip, 1, arg);
1967 	case XFS_IOC_FSSETXATTR:
1968 		return xfs_ioc_fssetxattr(ip, filp, arg);
1969 	case XFS_IOC_GETXFLAGS:
1970 		return xfs_ioc_getxflags(ip, arg);
1971 	case XFS_IOC_SETXFLAGS:
1972 		return xfs_ioc_setxflags(ip, filp, arg);
1973 
1974 	case XFS_IOC_FSSETDM: {
1975 		struct fsdmidata	dmi;
1976 
1977 		if (copy_from_user(&dmi, arg, sizeof(dmi)))
1978 			return -EFAULT;
1979 
1980 		error = mnt_want_write_file(filp);
1981 		if (error)
1982 			return error;
1983 
1984 		error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1985 				dmi.fsd_dmstate);
1986 		mnt_drop_write_file(filp);
1987 		return error;
1988 	}
1989 
1990 	case XFS_IOC_GETBMAP:
1991 	case XFS_IOC_GETBMAPA:
1992 	case XFS_IOC_GETBMAPX:
1993 		return xfs_ioc_getbmap(filp, cmd, arg);
1994 
1995 	case FS_IOC_GETFSMAP:
1996 		return xfs_ioc_getfsmap(ip, arg);
1997 
1998 	case XFS_IOC_SCRUB_METADATA:
1999 		return xfs_ioc_scrub_metadata(ip, arg);
2000 
2001 	case XFS_IOC_FD_TO_HANDLE:
2002 	case XFS_IOC_PATH_TO_HANDLE:
2003 	case XFS_IOC_PATH_TO_FSHANDLE: {
2004 		xfs_fsop_handlereq_t	hreq;
2005 
2006 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
2007 			return -EFAULT;
2008 		return xfs_find_handle(cmd, &hreq);
2009 	}
2010 	case XFS_IOC_OPEN_BY_HANDLE: {
2011 		xfs_fsop_handlereq_t	hreq;
2012 
2013 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2014 			return -EFAULT;
2015 		return xfs_open_by_handle(filp, &hreq);
2016 	}
2017 	case XFS_IOC_FSSETDM_BY_HANDLE:
2018 		return xfs_fssetdm_by_handle(filp, arg);
2019 
2020 	case XFS_IOC_READLINK_BY_HANDLE: {
2021 		xfs_fsop_handlereq_t	hreq;
2022 
2023 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2024 			return -EFAULT;
2025 		return xfs_readlink_by_handle(filp, &hreq);
2026 	}
2027 	case XFS_IOC_ATTRLIST_BY_HANDLE:
2028 		return xfs_attrlist_by_handle(filp, arg);
2029 
2030 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
2031 		return xfs_attrmulti_by_handle(filp, arg);
2032 
2033 	case XFS_IOC_SWAPEXT: {
2034 		struct xfs_swapext	sxp;
2035 
2036 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
2037 			return -EFAULT;
2038 		error = mnt_want_write_file(filp);
2039 		if (error)
2040 			return error;
2041 		error = xfs_ioc_swapext(&sxp);
2042 		mnt_drop_write_file(filp);
2043 		return error;
2044 	}
2045 
2046 	case XFS_IOC_FSCOUNTS: {
2047 		xfs_fsop_counts_t out;
2048 
2049 		error = xfs_fs_counts(mp, &out);
2050 		if (error)
2051 			return error;
2052 
2053 		if (copy_to_user(arg, &out, sizeof(out)))
2054 			return -EFAULT;
2055 		return 0;
2056 	}
2057 
2058 	case XFS_IOC_SET_RESBLKS: {
2059 		xfs_fsop_resblks_t inout;
2060 		uint64_t	   in;
2061 
2062 		if (!capable(CAP_SYS_ADMIN))
2063 			return -EPERM;
2064 
2065 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2066 			return -EROFS;
2067 
2068 		if (copy_from_user(&inout, arg, sizeof(inout)))
2069 			return -EFAULT;
2070 
2071 		error = mnt_want_write_file(filp);
2072 		if (error)
2073 			return error;
2074 
2075 		/* input parameter is passed in resblks field of structure */
2076 		in = inout.resblks;
2077 		error = xfs_reserve_blocks(mp, &in, &inout);
2078 		mnt_drop_write_file(filp);
2079 		if (error)
2080 			return error;
2081 
2082 		if (copy_to_user(arg, &inout, sizeof(inout)))
2083 			return -EFAULT;
2084 		return 0;
2085 	}
2086 
2087 	case XFS_IOC_GET_RESBLKS: {
2088 		xfs_fsop_resblks_t out;
2089 
2090 		if (!capable(CAP_SYS_ADMIN))
2091 			return -EPERM;
2092 
2093 		error = xfs_reserve_blocks(mp, NULL, &out);
2094 		if (error)
2095 			return error;
2096 
2097 		if (copy_to_user(arg, &out, sizeof(out)))
2098 			return -EFAULT;
2099 
2100 		return 0;
2101 	}
2102 
2103 	case XFS_IOC_FSGROWFSDATA: {
2104 		xfs_growfs_data_t in;
2105 
2106 		if (copy_from_user(&in, arg, sizeof(in)))
2107 			return -EFAULT;
2108 
2109 		error = mnt_want_write_file(filp);
2110 		if (error)
2111 			return error;
2112 		error = xfs_growfs_data(mp, &in);
2113 		mnt_drop_write_file(filp);
2114 		return error;
2115 	}
2116 
2117 	case XFS_IOC_FSGROWFSLOG: {
2118 		xfs_growfs_log_t in;
2119 
2120 		if (copy_from_user(&in, arg, sizeof(in)))
2121 			return -EFAULT;
2122 
2123 		error = mnt_want_write_file(filp);
2124 		if (error)
2125 			return error;
2126 		error = xfs_growfs_log(mp, &in);
2127 		mnt_drop_write_file(filp);
2128 		return error;
2129 	}
2130 
2131 	case XFS_IOC_FSGROWFSRT: {
2132 		xfs_growfs_rt_t in;
2133 
2134 		if (copy_from_user(&in, arg, sizeof(in)))
2135 			return -EFAULT;
2136 
2137 		error = mnt_want_write_file(filp);
2138 		if (error)
2139 			return error;
2140 		error = xfs_growfs_rt(mp, &in);
2141 		mnt_drop_write_file(filp);
2142 		return error;
2143 	}
2144 
2145 	case XFS_IOC_GOINGDOWN: {
2146 		uint32_t in;
2147 
2148 		if (!capable(CAP_SYS_ADMIN))
2149 			return -EPERM;
2150 
2151 		if (get_user(in, (uint32_t __user *)arg))
2152 			return -EFAULT;
2153 
2154 		return xfs_fs_goingdown(mp, in);
2155 	}
2156 
2157 	case XFS_IOC_ERROR_INJECTION: {
2158 		xfs_error_injection_t in;
2159 
2160 		if (!capable(CAP_SYS_ADMIN))
2161 			return -EPERM;
2162 
2163 		if (copy_from_user(&in, arg, sizeof(in)))
2164 			return -EFAULT;
2165 
2166 		return xfs_errortag_add(mp, in.errtag);
2167 	}
2168 
2169 	case XFS_IOC_ERROR_CLEARALL:
2170 		if (!capable(CAP_SYS_ADMIN))
2171 			return -EPERM;
2172 
2173 		return xfs_errortag_clearall(mp);
2174 
2175 	case XFS_IOC_FREE_EOFBLOCKS: {
2176 		struct xfs_fs_eofblocks eofb;
2177 		struct xfs_eofblocks keofb;
2178 
2179 		if (!capable(CAP_SYS_ADMIN))
2180 			return -EPERM;
2181 
2182 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2183 			return -EROFS;
2184 
2185 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2186 			return -EFAULT;
2187 
2188 		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2189 		if (error)
2190 			return error;
2191 
2192 		return xfs_icache_free_eofblocks(mp, &keofb);
2193 	}
2194 
2195 	default:
2196 		return -ENOTTY;
2197 	}
2198 }
2199