xref: /linux/fs/xfs/xfs_ioctl.c (revision 6a61b70b43c9c4cbc7314bf6c8b5ba8b0d6e1e7b)
1 /*
2  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3  * All Rights Reserved.
4  *
5  * This program is free software; you can redistribute it and/or
6  * modify it under the terms of the GNU General Public License as
7  * published by the Free Software Foundation.
8  *
9  * This program is distributed in the hope that it would be useful,
10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
12  * GNU General Public License for more details.
13  *
14  * You should have received a copy of the GNU General Public License
15  * along with this program; if not, write the Free Software Foundation,
16  * Inc.,  51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA
17  */
18 #include "xfs.h"
19 #include "xfs_fs.h"
20 #include "xfs_shared.h"
21 #include "xfs_format.h"
22 #include "xfs_log_format.h"
23 #include "xfs_trans_resv.h"
24 #include "xfs_mount.h"
25 #include "xfs_inode.h"
26 #include "xfs_ioctl.h"
27 #include "xfs_alloc.h"
28 #include "xfs_rtalloc.h"
29 #include "xfs_itable.h"
30 #include "xfs_error.h"
31 #include "xfs_attr.h"
32 #include "xfs_bmap.h"
33 #include "xfs_bmap_util.h"
34 #include "xfs_fsops.h"
35 #include "xfs_discard.h"
36 #include "xfs_quota.h"
37 #include "xfs_export.h"
38 #include "xfs_trace.h"
39 #include "xfs_icache.h"
40 #include "xfs_symlink.h"
41 #include "xfs_trans.h"
42 #include "xfs_pnfs.h"
43 #include "xfs_acl.h"
44 #include "xfs_btree.h"
45 #include <linux/fsmap.h>
46 #include "xfs_fsmap.h"
47 #include "scrub/xfs_scrub.h"
48 #include "xfs_sb.h"
49 
50 #include <linux/capability.h>
51 #include <linux/cred.h>
52 #include <linux/dcache.h>
53 #include <linux/mount.h>
54 #include <linux/namei.h>
55 #include <linux/pagemap.h>
56 #include <linux/slab.h>
57 #include <linux/exportfs.h>
58 
59 /*
60  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
61  * a file or fs handle.
62  *
63  * XFS_IOC_PATH_TO_FSHANDLE
64  *    returns fs handle for a mount point or path within that mount point
65  * XFS_IOC_FD_TO_HANDLE
66  *    returns full handle for a FD opened in user space
67  * XFS_IOC_PATH_TO_HANDLE
68  *    returns full handle for a path
69  */
70 int
71 xfs_find_handle(
72 	unsigned int		cmd,
73 	xfs_fsop_handlereq_t	*hreq)
74 {
75 	int			hsize;
76 	xfs_handle_t		handle;
77 	struct inode		*inode;
78 	struct fd		f = {NULL};
79 	struct path		path;
80 	int			error;
81 	struct xfs_inode	*ip;
82 
83 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
84 		f = fdget(hreq->fd);
85 		if (!f.file)
86 			return -EBADF;
87 		inode = file_inode(f.file);
88 	} else {
89 		error = user_lpath((const char __user *)hreq->path, &path);
90 		if (error)
91 			return error;
92 		inode = d_inode(path.dentry);
93 	}
94 	ip = XFS_I(inode);
95 
96 	/*
97 	 * We can only generate handles for inodes residing on a XFS filesystem,
98 	 * and only for regular files, directories or symbolic links.
99 	 */
100 	error = -EINVAL;
101 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
102 		goto out_put;
103 
104 	error = -EBADF;
105 	if (!S_ISREG(inode->i_mode) &&
106 	    !S_ISDIR(inode->i_mode) &&
107 	    !S_ISLNK(inode->i_mode))
108 		goto out_put;
109 
110 
111 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
112 
113 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
114 		/*
115 		 * This handle only contains an fsid, zero the rest.
116 		 */
117 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
118 		hsize = sizeof(xfs_fsid_t);
119 	} else {
120 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
121 					sizeof(handle.ha_fid.fid_len);
122 		handle.ha_fid.fid_pad = 0;
123 		handle.ha_fid.fid_gen = inode->i_generation;
124 		handle.ha_fid.fid_ino = ip->i_ino;
125 		hsize = sizeof(xfs_handle_t);
126 	}
127 
128 	error = -EFAULT;
129 	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
130 	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
131 		goto out_put;
132 
133 	error = 0;
134 
135  out_put:
136 	if (cmd == XFS_IOC_FD_TO_HANDLE)
137 		fdput(f);
138 	else
139 		path_put(&path);
140 	return error;
141 }
142 
143 /*
144  * No need to do permission checks on the various pathname components
145  * as the handle operations are privileged.
146  */
147 STATIC int
148 xfs_handle_acceptable(
149 	void			*context,
150 	struct dentry		*dentry)
151 {
152 	return 1;
153 }
154 
155 /*
156  * Convert userspace handle data into a dentry.
157  */
158 struct dentry *
159 xfs_handle_to_dentry(
160 	struct file		*parfilp,
161 	void __user		*uhandle,
162 	u32			hlen)
163 {
164 	xfs_handle_t		handle;
165 	struct xfs_fid64	fid;
166 
167 	/*
168 	 * Only allow handle opens under a directory.
169 	 */
170 	if (!S_ISDIR(file_inode(parfilp)->i_mode))
171 		return ERR_PTR(-ENOTDIR);
172 
173 	if (hlen != sizeof(xfs_handle_t))
174 		return ERR_PTR(-EINVAL);
175 	if (copy_from_user(&handle, uhandle, hlen))
176 		return ERR_PTR(-EFAULT);
177 	if (handle.ha_fid.fid_len !=
178 	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
179 		return ERR_PTR(-EINVAL);
180 
181 	memset(&fid, 0, sizeof(struct fid));
182 	fid.ino = handle.ha_fid.fid_ino;
183 	fid.gen = handle.ha_fid.fid_gen;
184 
185 	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
186 			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
187 			xfs_handle_acceptable, NULL);
188 }
189 
190 STATIC struct dentry *
191 xfs_handlereq_to_dentry(
192 	struct file		*parfilp,
193 	xfs_fsop_handlereq_t	*hreq)
194 {
195 	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
196 }
197 
198 int
199 xfs_open_by_handle(
200 	struct file		*parfilp,
201 	xfs_fsop_handlereq_t	*hreq)
202 {
203 	const struct cred	*cred = current_cred();
204 	int			error;
205 	int			fd;
206 	int			permflag;
207 	struct file		*filp;
208 	struct inode		*inode;
209 	struct dentry		*dentry;
210 	fmode_t			fmode;
211 	struct path		path;
212 
213 	if (!capable(CAP_SYS_ADMIN))
214 		return -EPERM;
215 
216 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
217 	if (IS_ERR(dentry))
218 		return PTR_ERR(dentry);
219 	inode = d_inode(dentry);
220 
221 	/* Restrict xfs_open_by_handle to directories & regular files. */
222 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
223 		error = -EPERM;
224 		goto out_dput;
225 	}
226 
227 #if BITS_PER_LONG != 32
228 	hreq->oflags |= O_LARGEFILE;
229 #endif
230 
231 	permflag = hreq->oflags;
232 	fmode = OPEN_FMODE(permflag);
233 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
234 	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
235 		error = -EPERM;
236 		goto out_dput;
237 	}
238 
239 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
240 		error = -EPERM;
241 		goto out_dput;
242 	}
243 
244 	/* Can't write directories. */
245 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
246 		error = -EISDIR;
247 		goto out_dput;
248 	}
249 
250 	fd = get_unused_fd_flags(0);
251 	if (fd < 0) {
252 		error = fd;
253 		goto out_dput;
254 	}
255 
256 	path.mnt = parfilp->f_path.mnt;
257 	path.dentry = dentry;
258 	filp = dentry_open(&path, hreq->oflags, cred);
259 	dput(dentry);
260 	if (IS_ERR(filp)) {
261 		put_unused_fd(fd);
262 		return PTR_ERR(filp);
263 	}
264 
265 	if (S_ISREG(inode->i_mode)) {
266 		filp->f_flags |= O_NOATIME;
267 		filp->f_mode |= FMODE_NOCMTIME;
268 	}
269 
270 	fd_install(fd, filp);
271 	return fd;
272 
273  out_dput:
274 	dput(dentry);
275 	return error;
276 }
277 
278 int
279 xfs_readlink_by_handle(
280 	struct file		*parfilp,
281 	xfs_fsop_handlereq_t	*hreq)
282 {
283 	struct dentry		*dentry;
284 	__u32			olen;
285 	int			error;
286 
287 	if (!capable(CAP_SYS_ADMIN))
288 		return -EPERM;
289 
290 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
291 	if (IS_ERR(dentry))
292 		return PTR_ERR(dentry);
293 
294 	/* Restrict this handle operation to symlinks only. */
295 	if (!d_is_symlink(dentry)) {
296 		error = -EINVAL;
297 		goto out_dput;
298 	}
299 
300 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
301 		error = -EFAULT;
302 		goto out_dput;
303 	}
304 
305 	error = vfs_readlink(dentry, hreq->ohandle, olen);
306 
307  out_dput:
308 	dput(dentry);
309 	return error;
310 }
311 
312 int
313 xfs_set_dmattrs(
314 	xfs_inode_t     *ip,
315 	uint		evmask,
316 	uint16_t	state)
317 {
318 	xfs_mount_t	*mp = ip->i_mount;
319 	xfs_trans_t	*tp;
320 	int		error;
321 
322 	if (!capable(CAP_SYS_ADMIN))
323 		return -EPERM;
324 
325 	if (XFS_FORCED_SHUTDOWN(mp))
326 		return -EIO;
327 
328 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
329 	if (error)
330 		return error;
331 
332 	xfs_ilock(ip, XFS_ILOCK_EXCL);
333 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
334 
335 	ip->i_d.di_dmevmask = evmask;
336 	ip->i_d.di_dmstate  = state;
337 
338 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
339 	error = xfs_trans_commit(tp);
340 
341 	return error;
342 }
343 
344 STATIC int
345 xfs_fssetdm_by_handle(
346 	struct file		*parfilp,
347 	void			__user *arg)
348 {
349 	int			error;
350 	struct fsdmidata	fsd;
351 	xfs_fsop_setdm_handlereq_t dmhreq;
352 	struct dentry		*dentry;
353 
354 	if (!capable(CAP_MKNOD))
355 		return -EPERM;
356 	if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
357 		return -EFAULT;
358 
359 	error = mnt_want_write_file(parfilp);
360 	if (error)
361 		return error;
362 
363 	dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
364 	if (IS_ERR(dentry)) {
365 		mnt_drop_write_file(parfilp);
366 		return PTR_ERR(dentry);
367 	}
368 
369 	if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
370 		error = -EPERM;
371 		goto out;
372 	}
373 
374 	if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
375 		error = -EFAULT;
376 		goto out;
377 	}
378 
379 	error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
380 				 fsd.fsd_dmstate);
381 
382  out:
383 	mnt_drop_write_file(parfilp);
384 	dput(dentry);
385 	return error;
386 }
387 
388 STATIC int
389 xfs_attrlist_by_handle(
390 	struct file		*parfilp,
391 	void			__user *arg)
392 {
393 	int			error = -ENOMEM;
394 	attrlist_cursor_kern_t	*cursor;
395 	struct xfs_fsop_attrlist_handlereq __user	*p = arg;
396 	xfs_fsop_attrlist_handlereq_t al_hreq;
397 	struct dentry		*dentry;
398 	char			*kbuf;
399 
400 	if (!capable(CAP_SYS_ADMIN))
401 		return -EPERM;
402 	if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
403 		return -EFAULT;
404 	if (al_hreq.buflen < sizeof(struct attrlist) ||
405 	    al_hreq.buflen > XFS_XATTR_LIST_MAX)
406 		return -EINVAL;
407 
408 	/*
409 	 * Reject flags, only allow namespaces.
410 	 */
411 	if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
412 		return -EINVAL;
413 
414 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
415 	if (IS_ERR(dentry))
416 		return PTR_ERR(dentry);
417 
418 	kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
419 	if (!kbuf)
420 		goto out_dput;
421 
422 	cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
423 	error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
424 					al_hreq.flags, cursor);
425 	if (error)
426 		goto out_kfree;
427 
428 	if (copy_to_user(&p->pos, cursor, sizeof(attrlist_cursor_kern_t))) {
429 		error = -EFAULT;
430 		goto out_kfree;
431 	}
432 
433 	if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
434 		error = -EFAULT;
435 
436 out_kfree:
437 	kmem_free(kbuf);
438 out_dput:
439 	dput(dentry);
440 	return error;
441 }
442 
443 int
444 xfs_attrmulti_attr_get(
445 	struct inode		*inode,
446 	unsigned char		*name,
447 	unsigned char		__user *ubuf,
448 	uint32_t		*len,
449 	uint32_t		flags)
450 {
451 	unsigned char		*kbuf;
452 	int			error = -EFAULT;
453 
454 	if (*len > XFS_XATTR_SIZE_MAX)
455 		return -EINVAL;
456 	kbuf = kmem_zalloc_large(*len, KM_SLEEP);
457 	if (!kbuf)
458 		return -ENOMEM;
459 
460 	error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
461 	if (error)
462 		goto out_kfree;
463 
464 	if (copy_to_user(ubuf, kbuf, *len))
465 		error = -EFAULT;
466 
467 out_kfree:
468 	kmem_free(kbuf);
469 	return error;
470 }
471 
472 int
473 xfs_attrmulti_attr_set(
474 	struct inode		*inode,
475 	unsigned char		*name,
476 	const unsigned char	__user *ubuf,
477 	uint32_t		len,
478 	uint32_t		flags)
479 {
480 	unsigned char		*kbuf;
481 	int			error;
482 
483 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
484 		return -EPERM;
485 	if (len > XFS_XATTR_SIZE_MAX)
486 		return -EINVAL;
487 
488 	kbuf = memdup_user(ubuf, len);
489 	if (IS_ERR(kbuf))
490 		return PTR_ERR(kbuf);
491 
492 	error = xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
493 	if (!error)
494 		xfs_forget_acl(inode, name, flags);
495 	kfree(kbuf);
496 	return error;
497 }
498 
499 int
500 xfs_attrmulti_attr_remove(
501 	struct inode		*inode,
502 	unsigned char		*name,
503 	uint32_t		flags)
504 {
505 	int			error;
506 
507 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
508 		return -EPERM;
509 	error = xfs_attr_remove(XFS_I(inode), name, flags);
510 	if (!error)
511 		xfs_forget_acl(inode, name, flags);
512 	return error;
513 }
514 
515 STATIC int
516 xfs_attrmulti_by_handle(
517 	struct file		*parfilp,
518 	void			__user *arg)
519 {
520 	int			error;
521 	xfs_attr_multiop_t	*ops;
522 	xfs_fsop_attrmulti_handlereq_t am_hreq;
523 	struct dentry		*dentry;
524 	unsigned int		i, size;
525 	unsigned char		*attr_name;
526 
527 	if (!capable(CAP_SYS_ADMIN))
528 		return -EPERM;
529 	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
530 		return -EFAULT;
531 
532 	/* overflow check */
533 	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
534 		return -E2BIG;
535 
536 	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
537 	if (IS_ERR(dentry))
538 		return PTR_ERR(dentry);
539 
540 	error = -E2BIG;
541 	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
542 	if (!size || size > 16 * PAGE_SIZE)
543 		goto out_dput;
544 
545 	ops = memdup_user(am_hreq.ops, size);
546 	if (IS_ERR(ops)) {
547 		error = PTR_ERR(ops);
548 		goto out_dput;
549 	}
550 
551 	error = -ENOMEM;
552 	attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
553 	if (!attr_name)
554 		goto out_kfree_ops;
555 
556 	error = 0;
557 	for (i = 0; i < am_hreq.opcount; i++) {
558 		ops[i].am_error = strncpy_from_user((char *)attr_name,
559 				ops[i].am_attrname, MAXNAMELEN);
560 		if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
561 			error = -ERANGE;
562 		if (ops[i].am_error < 0)
563 			break;
564 
565 		switch (ops[i].am_opcode) {
566 		case ATTR_OP_GET:
567 			ops[i].am_error = xfs_attrmulti_attr_get(
568 					d_inode(dentry), attr_name,
569 					ops[i].am_attrvalue, &ops[i].am_length,
570 					ops[i].am_flags);
571 			break;
572 		case ATTR_OP_SET:
573 			ops[i].am_error = mnt_want_write_file(parfilp);
574 			if (ops[i].am_error)
575 				break;
576 			ops[i].am_error = xfs_attrmulti_attr_set(
577 					d_inode(dentry), attr_name,
578 					ops[i].am_attrvalue, ops[i].am_length,
579 					ops[i].am_flags);
580 			mnt_drop_write_file(parfilp);
581 			break;
582 		case ATTR_OP_REMOVE:
583 			ops[i].am_error = mnt_want_write_file(parfilp);
584 			if (ops[i].am_error)
585 				break;
586 			ops[i].am_error = xfs_attrmulti_attr_remove(
587 					d_inode(dentry), attr_name,
588 					ops[i].am_flags);
589 			mnt_drop_write_file(parfilp);
590 			break;
591 		default:
592 			ops[i].am_error = -EINVAL;
593 		}
594 	}
595 
596 	if (copy_to_user(am_hreq.ops, ops, size))
597 		error = -EFAULT;
598 
599 	kfree(attr_name);
600  out_kfree_ops:
601 	kfree(ops);
602  out_dput:
603 	dput(dentry);
604 	return error;
605 }
606 
607 int
608 xfs_ioc_space(
609 	struct file		*filp,
610 	unsigned int		cmd,
611 	xfs_flock64_t		*bf)
612 {
613 	struct inode		*inode = file_inode(filp);
614 	struct xfs_inode	*ip = XFS_I(inode);
615 	struct iattr		iattr;
616 	enum xfs_prealloc_flags	flags = 0;
617 	uint			iolock = XFS_IOLOCK_EXCL;
618 	int			error;
619 
620 	/*
621 	 * Only allow the sys admin to reserve space unless
622 	 * unwritten extents are enabled.
623 	 */
624 	if (!xfs_sb_version_hasextflgbit(&ip->i_mount->m_sb) &&
625 	    !capable(CAP_SYS_ADMIN))
626 		return -EPERM;
627 
628 	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
629 		return -EPERM;
630 
631 	if (!(filp->f_mode & FMODE_WRITE))
632 		return -EBADF;
633 
634 	if (!S_ISREG(inode->i_mode))
635 		return -EINVAL;
636 
637 	if (filp->f_flags & O_DSYNC)
638 		flags |= XFS_PREALLOC_SYNC;
639 	if (filp->f_mode & FMODE_NOCMTIME)
640 		flags |= XFS_PREALLOC_INVISIBLE;
641 
642 	error = mnt_want_write_file(filp);
643 	if (error)
644 		return error;
645 
646 	xfs_ilock(ip, iolock);
647 	error = xfs_break_layouts(inode, &iolock);
648 	if (error)
649 		goto out_unlock;
650 
651 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
652 	iolock |= XFS_MMAPLOCK_EXCL;
653 
654 	switch (bf->l_whence) {
655 	case 0: /*SEEK_SET*/
656 		break;
657 	case 1: /*SEEK_CUR*/
658 		bf->l_start += filp->f_pos;
659 		break;
660 	case 2: /*SEEK_END*/
661 		bf->l_start += XFS_ISIZE(ip);
662 		break;
663 	default:
664 		error = -EINVAL;
665 		goto out_unlock;
666 	}
667 
668 	/*
669 	 * length of <= 0 for resv/unresv/zero is invalid.  length for
670 	 * alloc/free is ignored completely and we have no idea what userspace
671 	 * might have set it to, so set it to zero to allow range
672 	 * checks to pass.
673 	 */
674 	switch (cmd) {
675 	case XFS_IOC_ZERO_RANGE:
676 	case XFS_IOC_RESVSP:
677 	case XFS_IOC_RESVSP64:
678 	case XFS_IOC_UNRESVSP:
679 	case XFS_IOC_UNRESVSP64:
680 		if (bf->l_len <= 0) {
681 			error = -EINVAL;
682 			goto out_unlock;
683 		}
684 		break;
685 	default:
686 		bf->l_len = 0;
687 		break;
688 	}
689 
690 	if (bf->l_start < 0 ||
691 	    bf->l_start > inode->i_sb->s_maxbytes ||
692 	    bf->l_start + bf->l_len < 0 ||
693 	    bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
694 		error = -EINVAL;
695 		goto out_unlock;
696 	}
697 
698 	switch (cmd) {
699 	case XFS_IOC_ZERO_RANGE:
700 		flags |= XFS_PREALLOC_SET;
701 		error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
702 		break;
703 	case XFS_IOC_RESVSP:
704 	case XFS_IOC_RESVSP64:
705 		flags |= XFS_PREALLOC_SET;
706 		error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
707 						XFS_BMAPI_PREALLOC);
708 		break;
709 	case XFS_IOC_UNRESVSP:
710 	case XFS_IOC_UNRESVSP64:
711 		error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
712 		break;
713 	case XFS_IOC_ALLOCSP:
714 	case XFS_IOC_ALLOCSP64:
715 	case XFS_IOC_FREESP:
716 	case XFS_IOC_FREESP64:
717 		flags |= XFS_PREALLOC_CLEAR;
718 		if (bf->l_start > XFS_ISIZE(ip)) {
719 			error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
720 					bf->l_start - XFS_ISIZE(ip), 0);
721 			if (error)
722 				goto out_unlock;
723 		}
724 
725 		iattr.ia_valid = ATTR_SIZE;
726 		iattr.ia_size = bf->l_start;
727 
728 		error = xfs_vn_setattr_size(file_dentry(filp), &iattr);
729 		break;
730 	default:
731 		ASSERT(0);
732 		error = -EINVAL;
733 	}
734 
735 	if (error)
736 		goto out_unlock;
737 
738 	error = xfs_update_prealloc_flags(ip, flags);
739 
740 out_unlock:
741 	xfs_iunlock(ip, iolock);
742 	mnt_drop_write_file(filp);
743 	return error;
744 }
745 
746 STATIC int
747 xfs_ioc_bulkstat(
748 	xfs_mount_t		*mp,
749 	unsigned int		cmd,
750 	void			__user *arg)
751 {
752 	xfs_fsop_bulkreq_t	bulkreq;
753 	int			count;	/* # of records returned */
754 	xfs_ino_t		inlast;	/* last inode number */
755 	int			done;
756 	int			error;
757 
758 	/* done = 1 if there are more stats to get and if bulkstat */
759 	/* should be called again (unused here, but used in dmapi) */
760 
761 	if (!capable(CAP_SYS_ADMIN))
762 		return -EPERM;
763 
764 	if (XFS_FORCED_SHUTDOWN(mp))
765 		return -EIO;
766 
767 	if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
768 		return -EFAULT;
769 
770 	if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
771 		return -EFAULT;
772 
773 	if ((count = bulkreq.icount) <= 0)
774 		return -EINVAL;
775 
776 	if (bulkreq.ubuffer == NULL)
777 		return -EINVAL;
778 
779 	if (cmd == XFS_IOC_FSINUMBERS)
780 		error = xfs_inumbers(mp, &inlast, &count,
781 					bulkreq.ubuffer, xfs_inumbers_fmt);
782 	else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
783 		error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
784 					sizeof(xfs_bstat_t), NULL, &done);
785 	else	/* XFS_IOC_FSBULKSTAT */
786 		error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
787 				     sizeof(xfs_bstat_t), bulkreq.ubuffer,
788 				     &done);
789 
790 	if (error)
791 		return error;
792 
793 	if (bulkreq.ocount != NULL) {
794 		if (copy_to_user(bulkreq.lastip, &inlast,
795 						sizeof(xfs_ino_t)))
796 			return -EFAULT;
797 
798 		if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
799 			return -EFAULT;
800 	}
801 
802 	return 0;
803 }
804 
805 STATIC int
806 xfs_ioc_fsgeometry_v1(
807 	xfs_mount_t		*mp,
808 	void			__user *arg)
809 {
810 	xfs_fsop_geom_t         fsgeo;
811 	int			error;
812 
813 	error = xfs_fs_geometry(&mp->m_sb, &fsgeo, 3);
814 	if (error)
815 		return error;
816 
817 	/*
818 	 * Caller should have passed an argument of type
819 	 * xfs_fsop_geom_v1_t.  This is a proper subset of the
820 	 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
821 	 */
822 	if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
823 		return -EFAULT;
824 	return 0;
825 }
826 
827 STATIC int
828 xfs_ioc_fsgeometry(
829 	xfs_mount_t		*mp,
830 	void			__user *arg)
831 {
832 	xfs_fsop_geom_t		fsgeo;
833 	int			error;
834 
835 	error = xfs_fs_geometry(&mp->m_sb, &fsgeo, 4);
836 	if (error)
837 		return error;
838 
839 	if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
840 		return -EFAULT;
841 	return 0;
842 }
843 
844 /*
845  * Linux extended inode flags interface.
846  */
847 
848 STATIC unsigned int
849 xfs_merge_ioc_xflags(
850 	unsigned int	flags,
851 	unsigned int	start)
852 {
853 	unsigned int	xflags = start;
854 
855 	if (flags & FS_IMMUTABLE_FL)
856 		xflags |= FS_XFLAG_IMMUTABLE;
857 	else
858 		xflags &= ~FS_XFLAG_IMMUTABLE;
859 	if (flags & FS_APPEND_FL)
860 		xflags |= FS_XFLAG_APPEND;
861 	else
862 		xflags &= ~FS_XFLAG_APPEND;
863 	if (flags & FS_SYNC_FL)
864 		xflags |= FS_XFLAG_SYNC;
865 	else
866 		xflags &= ~FS_XFLAG_SYNC;
867 	if (flags & FS_NOATIME_FL)
868 		xflags |= FS_XFLAG_NOATIME;
869 	else
870 		xflags &= ~FS_XFLAG_NOATIME;
871 	if (flags & FS_NODUMP_FL)
872 		xflags |= FS_XFLAG_NODUMP;
873 	else
874 		xflags &= ~FS_XFLAG_NODUMP;
875 
876 	return xflags;
877 }
878 
879 STATIC unsigned int
880 xfs_di2lxflags(
881 	uint16_t	di_flags)
882 {
883 	unsigned int	flags = 0;
884 
885 	if (di_flags & XFS_DIFLAG_IMMUTABLE)
886 		flags |= FS_IMMUTABLE_FL;
887 	if (di_flags & XFS_DIFLAG_APPEND)
888 		flags |= FS_APPEND_FL;
889 	if (di_flags & XFS_DIFLAG_SYNC)
890 		flags |= FS_SYNC_FL;
891 	if (di_flags & XFS_DIFLAG_NOATIME)
892 		flags |= FS_NOATIME_FL;
893 	if (di_flags & XFS_DIFLAG_NODUMP)
894 		flags |= FS_NODUMP_FL;
895 	return flags;
896 }
897 
898 STATIC int
899 xfs_ioc_fsgetxattr(
900 	xfs_inode_t		*ip,
901 	int			attr,
902 	void			__user *arg)
903 {
904 	struct fsxattr		fa;
905 
906 	memset(&fa, 0, sizeof(struct fsxattr));
907 
908 	xfs_ilock(ip, XFS_ILOCK_SHARED);
909 	fa.fsx_xflags = xfs_ip2xflags(ip);
910 	fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
911 	fa.fsx_cowextsize = ip->i_d.di_cowextsize <<
912 			ip->i_mount->m_sb.sb_blocklog;
913 	fa.fsx_projid = xfs_get_projid(ip);
914 
915 	if (attr) {
916 		if (ip->i_afp) {
917 			if (ip->i_afp->if_flags & XFS_IFEXTENTS)
918 				fa.fsx_nextents = xfs_iext_count(ip->i_afp);
919 			else
920 				fa.fsx_nextents = ip->i_d.di_anextents;
921 		} else
922 			fa.fsx_nextents = 0;
923 	} else {
924 		if (ip->i_df.if_flags & XFS_IFEXTENTS)
925 			fa.fsx_nextents = xfs_iext_count(&ip->i_df);
926 		else
927 			fa.fsx_nextents = ip->i_d.di_nextents;
928 	}
929 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
930 
931 	if (copy_to_user(arg, &fa, sizeof(fa)))
932 		return -EFAULT;
933 	return 0;
934 }
935 
936 STATIC uint16_t
937 xfs_flags2diflags(
938 	struct xfs_inode	*ip,
939 	unsigned int		xflags)
940 {
941 	/* can't set PREALLOC this way, just preserve it */
942 	uint16_t		di_flags =
943 		(ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
944 
945 	if (xflags & FS_XFLAG_IMMUTABLE)
946 		di_flags |= XFS_DIFLAG_IMMUTABLE;
947 	if (xflags & FS_XFLAG_APPEND)
948 		di_flags |= XFS_DIFLAG_APPEND;
949 	if (xflags & FS_XFLAG_SYNC)
950 		di_flags |= XFS_DIFLAG_SYNC;
951 	if (xflags & FS_XFLAG_NOATIME)
952 		di_flags |= XFS_DIFLAG_NOATIME;
953 	if (xflags & FS_XFLAG_NODUMP)
954 		di_flags |= XFS_DIFLAG_NODUMP;
955 	if (xflags & FS_XFLAG_NODEFRAG)
956 		di_flags |= XFS_DIFLAG_NODEFRAG;
957 	if (xflags & FS_XFLAG_FILESTREAM)
958 		di_flags |= XFS_DIFLAG_FILESTREAM;
959 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
960 		if (xflags & FS_XFLAG_RTINHERIT)
961 			di_flags |= XFS_DIFLAG_RTINHERIT;
962 		if (xflags & FS_XFLAG_NOSYMLINKS)
963 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
964 		if (xflags & FS_XFLAG_EXTSZINHERIT)
965 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
966 		if (xflags & FS_XFLAG_PROJINHERIT)
967 			di_flags |= XFS_DIFLAG_PROJINHERIT;
968 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
969 		if (xflags & FS_XFLAG_REALTIME)
970 			di_flags |= XFS_DIFLAG_REALTIME;
971 		if (xflags & FS_XFLAG_EXTSIZE)
972 			di_flags |= XFS_DIFLAG_EXTSIZE;
973 	}
974 
975 	return di_flags;
976 }
977 
978 STATIC uint64_t
979 xfs_flags2diflags2(
980 	struct xfs_inode	*ip,
981 	unsigned int		xflags)
982 {
983 	uint64_t		di_flags2 =
984 		(ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK);
985 
986 	if (xflags & FS_XFLAG_DAX)
987 		di_flags2 |= XFS_DIFLAG2_DAX;
988 	if (xflags & FS_XFLAG_COWEXTSIZE)
989 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
990 
991 	return di_flags2;
992 }
993 
994 STATIC void
995 xfs_diflags_to_linux(
996 	struct xfs_inode	*ip)
997 {
998 	struct inode		*inode = VFS_I(ip);
999 	unsigned int		xflags = xfs_ip2xflags(ip);
1000 
1001 	if (xflags & FS_XFLAG_IMMUTABLE)
1002 		inode->i_flags |= S_IMMUTABLE;
1003 	else
1004 		inode->i_flags &= ~S_IMMUTABLE;
1005 	if (xflags & FS_XFLAG_APPEND)
1006 		inode->i_flags |= S_APPEND;
1007 	else
1008 		inode->i_flags &= ~S_APPEND;
1009 	if (xflags & FS_XFLAG_SYNC)
1010 		inode->i_flags |= S_SYNC;
1011 	else
1012 		inode->i_flags &= ~S_SYNC;
1013 	if (xflags & FS_XFLAG_NOATIME)
1014 		inode->i_flags |= S_NOATIME;
1015 	else
1016 		inode->i_flags &= ~S_NOATIME;
1017 #if 0	/* disabled until the flag switching races are sorted out */
1018 	if (xflags & FS_XFLAG_DAX)
1019 		inode->i_flags |= S_DAX;
1020 	else
1021 		inode->i_flags &= ~S_DAX;
1022 #endif
1023 }
1024 
1025 static int
1026 xfs_ioctl_setattr_xflags(
1027 	struct xfs_trans	*tp,
1028 	struct xfs_inode	*ip,
1029 	struct fsxattr		*fa)
1030 {
1031 	struct xfs_mount	*mp = ip->i_mount;
1032 	uint64_t		di_flags2;
1033 
1034 	/* Can't change realtime flag if any extents are allocated. */
1035 	if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1036 	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1037 		return -EINVAL;
1038 
1039 	/* If realtime flag is set then must have realtime device */
1040 	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1041 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1042 		    (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1043 			return -EINVAL;
1044 	}
1045 
1046 	/* Clear reflink if we are actually able to set the rt flag. */
1047 	if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1048 		ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
1049 
1050 	/* Don't allow us to set DAX mode for a reflinked file for now. */
1051 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1052 		return -EINVAL;
1053 
1054 	/*
1055 	 * Can't modify an immutable/append-only file unless
1056 	 * we have appropriate permission.
1057 	 */
1058 	if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1059 	     (fa->fsx_xflags & (FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))) &&
1060 	    !capable(CAP_LINUX_IMMUTABLE))
1061 		return -EPERM;
1062 
1063 	/* diflags2 only valid for v3 inodes. */
1064 	di_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1065 	if (di_flags2 && ip->i_d.di_version < 3)
1066 		return -EINVAL;
1067 
1068 	ip->i_d.di_flags = xfs_flags2diflags(ip, fa->fsx_xflags);
1069 	ip->i_d.di_flags2 = di_flags2;
1070 
1071 	xfs_diflags_to_linux(ip);
1072 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1073 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1074 	XFS_STATS_INC(mp, xs_ig_attrchg);
1075 	return 0;
1076 }
1077 
1078 /*
1079  * If we are changing DAX flags, we have to ensure the file is clean and any
1080  * cached objects in the address space are invalidated and removed. This
1081  * requires us to lock out other IO and page faults similar to a truncate
1082  * operation. The locks need to be held until the transaction has been committed
1083  * so that the cache invalidation is atomic with respect to the DAX flag
1084  * manipulation.
1085  */
1086 static int
1087 xfs_ioctl_setattr_dax_invalidate(
1088 	struct xfs_inode	*ip,
1089 	struct fsxattr		*fa,
1090 	int			*join_flags)
1091 {
1092 	struct inode		*inode = VFS_I(ip);
1093 	struct super_block	*sb = inode->i_sb;
1094 	int			error;
1095 
1096 	*join_flags = 0;
1097 
1098 	/*
1099 	 * It is only valid to set the DAX flag on regular files and
1100 	 * directories on filesystems where the block size is equal to the page
1101 	 * size. On directories it serves as an inherit hint.
1102 	 */
1103 	if (fa->fsx_xflags & FS_XFLAG_DAX) {
1104 		if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1105 			return -EINVAL;
1106 		if (!bdev_dax_supported(xfs_find_bdev_for_inode(VFS_I(ip)),
1107 				sb->s_blocksize))
1108 			return -EINVAL;
1109 	}
1110 
1111 	/* If the DAX state is not changing, we have nothing to do here. */
1112 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
1113 		return 0;
1114 	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
1115 		return 0;
1116 
1117 	/* lock, flush and invalidate mapping in preparation for flag change */
1118 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1119 	error = filemap_write_and_wait(inode->i_mapping);
1120 	if (error)
1121 		goto out_unlock;
1122 	error = invalidate_inode_pages2(inode->i_mapping);
1123 	if (error)
1124 		goto out_unlock;
1125 
1126 	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
1127 	return 0;
1128 
1129 out_unlock:
1130 	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1131 	return error;
1132 
1133 }
1134 
1135 /*
1136  * Set up the transaction structure for the setattr operation, checking that we
1137  * have permission to do so. On success, return a clean transaction and the
1138  * inode locked exclusively ready for further operation specific checks. On
1139  * failure, return an error without modifying or locking the inode.
1140  *
1141  * The inode might already be IO locked on call. If this is the case, it is
1142  * indicated in @join_flags and we take full responsibility for ensuring they
1143  * are unlocked from now on. Hence if we have an error here, we still have to
1144  * unlock them. Otherwise, once they are joined to the transaction, they will
1145  * be unlocked on commit/cancel.
1146  */
1147 static struct xfs_trans *
1148 xfs_ioctl_setattr_get_trans(
1149 	struct xfs_inode	*ip,
1150 	int			join_flags)
1151 {
1152 	struct xfs_mount	*mp = ip->i_mount;
1153 	struct xfs_trans	*tp;
1154 	int			error = -EROFS;
1155 
1156 	if (mp->m_flags & XFS_MOUNT_RDONLY)
1157 		goto out_unlock;
1158 	error = -EIO;
1159 	if (XFS_FORCED_SHUTDOWN(mp))
1160 		goto out_unlock;
1161 
1162 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
1163 	if (error)
1164 		return ERR_PTR(error);
1165 
1166 	xfs_ilock(ip, XFS_ILOCK_EXCL);
1167 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
1168 	join_flags = 0;
1169 
1170 	/*
1171 	 * CAP_FOWNER overrides the following restrictions:
1172 	 *
1173 	 * The user ID of the calling process must be equal to the file owner
1174 	 * ID, except in cases where the CAP_FSETID capability is applicable.
1175 	 */
1176 	if (!inode_owner_or_capable(VFS_I(ip))) {
1177 		error = -EPERM;
1178 		goto out_cancel;
1179 	}
1180 
1181 	if (mp->m_flags & XFS_MOUNT_WSYNC)
1182 		xfs_trans_set_sync(tp);
1183 
1184 	return tp;
1185 
1186 out_cancel:
1187 	xfs_trans_cancel(tp);
1188 out_unlock:
1189 	if (join_flags)
1190 		xfs_iunlock(ip, join_flags);
1191 	return ERR_PTR(error);
1192 }
1193 
1194 /*
1195  * extent size hint validation is somewhat cumbersome. Rules are:
1196  *
1197  * 1. extent size hint is only valid for directories and regular files
1198  * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1199  * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1200  * 4. can only be changed on regular files if no extents are allocated
1201  * 5. can be changed on directories at any time
1202  * 6. extsize hint of 0 turns off hints, clears inode flags.
1203  * 7. Extent size must be a multiple of the appropriate block size.
1204  * 8. for non-realtime files, the extent size hint must be limited
1205  *    to half the AG size to avoid alignment extending the extent beyond the
1206  *    limits of the AG.
1207  *
1208  * Please keep this function in sync with xfs_scrub_inode_extsize.
1209  */
1210 static int
1211 xfs_ioctl_setattr_check_extsize(
1212 	struct xfs_inode	*ip,
1213 	struct fsxattr		*fa)
1214 {
1215 	struct xfs_mount	*mp = ip->i_mount;
1216 
1217 	if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(VFS_I(ip)->i_mode))
1218 		return -EINVAL;
1219 
1220 	if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
1221 	    !S_ISDIR(VFS_I(ip)->i_mode))
1222 		return -EINVAL;
1223 
1224 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_d.di_nextents &&
1225 	    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1226 		return -EINVAL;
1227 
1228 	if (fa->fsx_extsize != 0) {
1229 		xfs_extlen_t    size;
1230 		xfs_fsblock_t   extsize_fsb;
1231 
1232 		extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1233 		if (extsize_fsb > MAXEXTLEN)
1234 			return -EINVAL;
1235 
1236 		if (XFS_IS_REALTIME_INODE(ip) ||
1237 		    (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1238 			size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1239 		} else {
1240 			size = mp->m_sb.sb_blocksize;
1241 			if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1242 				return -EINVAL;
1243 		}
1244 
1245 		if (fa->fsx_extsize % size)
1246 			return -EINVAL;
1247 	} else
1248 		fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
1249 
1250 	return 0;
1251 }
1252 
1253 /*
1254  * CoW extent size hint validation rules are:
1255  *
1256  * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
1257  *    The inode does not have to have any shared blocks, but it must be a v3.
1258  * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
1259  *    for a directory, the hint is propagated to new files.
1260  * 3. Can be changed on files & directories at any time.
1261  * 4. CoW extsize hint of 0 turns off hints, clears inode flags.
1262  * 5. Extent size must be a multiple of the appropriate block size.
1263  * 6. The extent size hint must be limited to half the AG size to avoid
1264  *    alignment extending the extent beyond the limits of the AG.
1265  *
1266  * Please keep this function in sync with xfs_scrub_inode_cowextsize.
1267  */
1268 static int
1269 xfs_ioctl_setattr_check_cowextsize(
1270 	struct xfs_inode	*ip,
1271 	struct fsxattr		*fa)
1272 {
1273 	struct xfs_mount	*mp = ip->i_mount;
1274 
1275 	if (!(fa->fsx_xflags & FS_XFLAG_COWEXTSIZE))
1276 		return 0;
1277 
1278 	if (!xfs_sb_version_hasreflink(&ip->i_mount->m_sb) ||
1279 	    ip->i_d.di_version != 3)
1280 		return -EINVAL;
1281 
1282 	if (!S_ISREG(VFS_I(ip)->i_mode) && !S_ISDIR(VFS_I(ip)->i_mode))
1283 		return -EINVAL;
1284 
1285 	if (fa->fsx_cowextsize != 0) {
1286 		xfs_extlen_t    size;
1287 		xfs_fsblock_t   cowextsize_fsb;
1288 
1289 		cowextsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1290 		if (cowextsize_fsb > MAXEXTLEN)
1291 			return -EINVAL;
1292 
1293 		size = mp->m_sb.sb_blocksize;
1294 		if (cowextsize_fsb > mp->m_sb.sb_agblocks / 2)
1295 			return -EINVAL;
1296 
1297 		if (fa->fsx_cowextsize % size)
1298 			return -EINVAL;
1299 	} else
1300 		fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
1301 
1302 	return 0;
1303 }
1304 
1305 static int
1306 xfs_ioctl_setattr_check_projid(
1307 	struct xfs_inode	*ip,
1308 	struct fsxattr		*fa)
1309 {
1310 	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1311 	if (fa->fsx_projid > (uint16_t)-1 &&
1312 	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1313 		return -EINVAL;
1314 
1315 	/*
1316 	 * Project Quota ID state is only allowed to change from within the init
1317 	 * namespace. Enforce that restriction only if we are trying to change
1318 	 * the quota ID state. Everything else is allowed in user namespaces.
1319 	 */
1320 	if (current_user_ns() == &init_user_ns)
1321 		return 0;
1322 
1323 	if (xfs_get_projid(ip) != fa->fsx_projid)
1324 		return -EINVAL;
1325 	if ((fa->fsx_xflags & FS_XFLAG_PROJINHERIT) !=
1326 	    (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1327 		return -EINVAL;
1328 
1329 	return 0;
1330 }
1331 
1332 STATIC int
1333 xfs_ioctl_setattr(
1334 	xfs_inode_t		*ip,
1335 	struct fsxattr		*fa)
1336 {
1337 	struct xfs_mount	*mp = ip->i_mount;
1338 	struct xfs_trans	*tp;
1339 	struct xfs_dquot	*udqp = NULL;
1340 	struct xfs_dquot	*pdqp = NULL;
1341 	struct xfs_dquot	*olddquot = NULL;
1342 	int			code;
1343 	int			join_flags = 0;
1344 
1345 	trace_xfs_ioctl_setattr(ip);
1346 
1347 	code = xfs_ioctl_setattr_check_projid(ip, fa);
1348 	if (code)
1349 		return code;
1350 
1351 	/*
1352 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1353 	 * before we start any other transactions. Trying to do this later
1354 	 * is messy. We don't care to take a readlock to look at the ids
1355 	 * in inode here, because we can't hold it across the trans_reserve.
1356 	 * If the IDs do change before we take the ilock, we're covered
1357 	 * because the i_*dquot fields will get updated anyway.
1358 	 */
1359 	if (XFS_IS_QUOTA_ON(mp)) {
1360 		code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1361 					 ip->i_d.di_gid, fa->fsx_projid,
1362 					 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1363 		if (code)
1364 			return code;
1365 	}
1366 
1367 	/*
1368 	 * Changing DAX config may require inode locking for mapping
1369 	 * invalidation. These need to be held all the way to transaction commit
1370 	 * or cancel time, so need to be passed through to
1371 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1372 	 * appropriately.
1373 	 */
1374 	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
1375 	if (code)
1376 		goto error_free_dquots;
1377 
1378 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1379 	if (IS_ERR(tp)) {
1380 		code = PTR_ERR(tp);
1381 		goto error_free_dquots;
1382 	}
1383 
1384 
1385 	if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1386 	    xfs_get_projid(ip) != fa->fsx_projid) {
1387 		code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1388 				capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
1389 		if (code)	/* out of quota */
1390 			goto error_trans_cancel;
1391 	}
1392 
1393 	code = xfs_ioctl_setattr_check_extsize(ip, fa);
1394 	if (code)
1395 		goto error_trans_cancel;
1396 
1397 	code = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1398 	if (code)
1399 		goto error_trans_cancel;
1400 
1401 	code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1402 	if (code)
1403 		goto error_trans_cancel;
1404 
1405 	/*
1406 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1407 	 * overrides the following restrictions:
1408 	 *
1409 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1410 	 * successful return from chown()
1411 	 */
1412 
1413 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1414 	    !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1415 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1416 
1417 	/* Change the ownerships and register project quota modifications */
1418 	if (xfs_get_projid(ip) != fa->fsx_projid) {
1419 		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1420 			olddquot = xfs_qm_vop_chown(tp, ip,
1421 						&ip->i_pdquot, pdqp);
1422 		}
1423 		ASSERT(ip->i_d.di_version > 1);
1424 		xfs_set_projid(ip, fa->fsx_projid);
1425 	}
1426 
1427 	/*
1428 	 * Only set the extent size hint if we've already determined that the
1429 	 * extent size hint should be set on the inode. If no extent size flags
1430 	 * are set on the inode then unconditionally clear the extent size hint.
1431 	 */
1432 	if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1433 		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1434 	else
1435 		ip->i_d.di_extsize = 0;
1436 	if (ip->i_d.di_version == 3 &&
1437 	    (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
1438 		ip->i_d.di_cowextsize = fa->fsx_cowextsize >>
1439 				mp->m_sb.sb_blocklog;
1440 	else
1441 		ip->i_d.di_cowextsize = 0;
1442 
1443 	code = xfs_trans_commit(tp);
1444 
1445 	/*
1446 	 * Release any dquot(s) the inode had kept before chown.
1447 	 */
1448 	xfs_qm_dqrele(olddquot);
1449 	xfs_qm_dqrele(udqp);
1450 	xfs_qm_dqrele(pdqp);
1451 
1452 	return code;
1453 
1454 error_trans_cancel:
1455 	xfs_trans_cancel(tp);
1456 error_free_dquots:
1457 	xfs_qm_dqrele(udqp);
1458 	xfs_qm_dqrele(pdqp);
1459 	return code;
1460 }
1461 
1462 STATIC int
1463 xfs_ioc_fssetxattr(
1464 	xfs_inode_t		*ip,
1465 	struct file		*filp,
1466 	void			__user *arg)
1467 {
1468 	struct fsxattr		fa;
1469 	int error;
1470 
1471 	if (copy_from_user(&fa, arg, sizeof(fa)))
1472 		return -EFAULT;
1473 
1474 	error = mnt_want_write_file(filp);
1475 	if (error)
1476 		return error;
1477 	error = xfs_ioctl_setattr(ip, &fa);
1478 	mnt_drop_write_file(filp);
1479 	return error;
1480 }
1481 
1482 STATIC int
1483 xfs_ioc_getxflags(
1484 	xfs_inode_t		*ip,
1485 	void			__user *arg)
1486 {
1487 	unsigned int		flags;
1488 
1489 	flags = xfs_di2lxflags(ip->i_d.di_flags);
1490 	if (copy_to_user(arg, &flags, sizeof(flags)))
1491 		return -EFAULT;
1492 	return 0;
1493 }
1494 
1495 STATIC int
1496 xfs_ioc_setxflags(
1497 	struct xfs_inode	*ip,
1498 	struct file		*filp,
1499 	void			__user *arg)
1500 {
1501 	struct xfs_trans	*tp;
1502 	struct fsxattr		fa;
1503 	unsigned int		flags;
1504 	int			join_flags = 0;
1505 	int			error;
1506 
1507 	if (copy_from_user(&flags, arg, sizeof(flags)))
1508 		return -EFAULT;
1509 
1510 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1511 		      FS_NOATIME_FL | FS_NODUMP_FL | \
1512 		      FS_SYNC_FL))
1513 		return -EOPNOTSUPP;
1514 
1515 	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1516 
1517 	error = mnt_want_write_file(filp);
1518 	if (error)
1519 		return error;
1520 
1521 	/*
1522 	 * Changing DAX config may require inode locking for mapping
1523 	 * invalidation. These need to be held all the way to transaction commit
1524 	 * or cancel time, so need to be passed through to
1525 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1526 	 * appropriately.
1527 	 */
1528 	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
1529 	if (error)
1530 		goto out_drop_write;
1531 
1532 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1533 	if (IS_ERR(tp)) {
1534 		error = PTR_ERR(tp);
1535 		goto out_drop_write;
1536 	}
1537 
1538 	error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1539 	if (error) {
1540 		xfs_trans_cancel(tp);
1541 		goto out_drop_write;
1542 	}
1543 
1544 	error = xfs_trans_commit(tp);
1545 out_drop_write:
1546 	mnt_drop_write_file(filp);
1547 	return error;
1548 }
1549 
1550 static bool
1551 xfs_getbmap_format(
1552 	struct kgetbmap		*p,
1553 	struct getbmapx __user	*u,
1554 	size_t			recsize)
1555 {
1556 	if (put_user(p->bmv_offset, &u->bmv_offset) ||
1557 	    put_user(p->bmv_block, &u->bmv_block) ||
1558 	    put_user(p->bmv_length, &u->bmv_length) ||
1559 	    put_user(0, &u->bmv_count) ||
1560 	    put_user(0, &u->bmv_entries))
1561 		return false;
1562 	if (recsize < sizeof(struct getbmapx))
1563 		return true;
1564 	if (put_user(0, &u->bmv_iflags) ||
1565 	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
1566 	    put_user(0, &u->bmv_unused1) ||
1567 	    put_user(0, &u->bmv_unused2))
1568 		return false;
1569 	return true;
1570 }
1571 
1572 STATIC int
1573 xfs_ioc_getbmap(
1574 	struct file		*file,
1575 	unsigned int		cmd,
1576 	void			__user *arg)
1577 {
1578 	struct getbmapx		bmx = { 0 };
1579 	struct kgetbmap		*buf;
1580 	size_t			recsize;
1581 	int			error, i;
1582 
1583 	switch (cmd) {
1584 	case XFS_IOC_GETBMAPA:
1585 		bmx.bmv_iflags = BMV_IF_ATTRFORK;
1586 		/*FALLTHRU*/
1587 	case XFS_IOC_GETBMAP:
1588 		if (file->f_mode & FMODE_NOCMTIME)
1589 			bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1590 		/* struct getbmap is a strict subset of struct getbmapx. */
1591 		recsize = sizeof(struct getbmap);
1592 		break;
1593 	case XFS_IOC_GETBMAPX:
1594 		recsize = sizeof(struct getbmapx);
1595 		break;
1596 	default:
1597 		return -EINVAL;
1598 	}
1599 
1600 	if (copy_from_user(&bmx, arg, recsize))
1601 		return -EFAULT;
1602 
1603 	if (bmx.bmv_count < 2)
1604 		return -EINVAL;
1605 	if (bmx.bmv_count > ULONG_MAX / recsize)
1606 		return -ENOMEM;
1607 
1608 	buf = kmem_zalloc_large(bmx.bmv_count * sizeof(*buf), 0);
1609 	if (!buf)
1610 		return -ENOMEM;
1611 
1612 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1613 	if (error)
1614 		goto out_free_buf;
1615 
1616 	error = -EFAULT;
1617 	if (copy_to_user(arg, &bmx, recsize))
1618 		goto out_free_buf;
1619 	arg += recsize;
1620 
1621 	for (i = 0; i < bmx.bmv_entries; i++) {
1622 		if (!xfs_getbmap_format(buf + i, arg, recsize))
1623 			goto out_free_buf;
1624 		arg += recsize;
1625 	}
1626 
1627 	error = 0;
1628 out_free_buf:
1629 	kmem_free(buf);
1630 	return 0;
1631 }
1632 
1633 struct getfsmap_info {
1634 	struct xfs_mount	*mp;
1635 	struct fsmap_head __user *data;
1636 	unsigned int		idx;
1637 	__u32			last_flags;
1638 };
1639 
1640 STATIC int
1641 xfs_getfsmap_format(struct xfs_fsmap *xfm, void *priv)
1642 {
1643 	struct getfsmap_info	*info = priv;
1644 	struct fsmap		fm;
1645 
1646 	trace_xfs_getfsmap_mapping(info->mp, xfm);
1647 
1648 	info->last_flags = xfm->fmr_flags;
1649 	xfs_fsmap_from_internal(&fm, xfm);
1650 	if (copy_to_user(&info->data->fmh_recs[info->idx++], &fm,
1651 			sizeof(struct fsmap)))
1652 		return -EFAULT;
1653 
1654 	return 0;
1655 }
1656 
1657 STATIC int
1658 xfs_ioc_getfsmap(
1659 	struct xfs_inode	*ip,
1660 	struct fsmap_head	__user *arg)
1661 {
1662 	struct getfsmap_info	info = { NULL };
1663 	struct xfs_fsmap_head	xhead = {0};
1664 	struct fsmap_head	head;
1665 	bool			aborted = false;
1666 	int			error;
1667 
1668 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1669 		return -EFAULT;
1670 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1671 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1672 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1673 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1674 		       sizeof(head.fmh_keys[1].fmr_reserved)))
1675 		return -EINVAL;
1676 
1677 	xhead.fmh_iflags = head.fmh_iflags;
1678 	xhead.fmh_count = head.fmh_count;
1679 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1680 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1681 
1682 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1683 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1684 
1685 	info.mp = ip->i_mount;
1686 	info.data = arg;
1687 	error = xfs_getfsmap(ip->i_mount, &xhead, xfs_getfsmap_format, &info);
1688 	if (error == XFS_BTREE_QUERY_RANGE_ABORT) {
1689 		error = 0;
1690 		aborted = true;
1691 	} else if (error)
1692 		return error;
1693 
1694 	/* If we didn't abort, set the "last" flag in the last fmx */
1695 	if (!aborted && info.idx) {
1696 		info.last_flags |= FMR_OF_LAST;
1697 		if (copy_to_user(&info.data->fmh_recs[info.idx - 1].fmr_flags,
1698 				&info.last_flags, sizeof(info.last_flags)))
1699 			return -EFAULT;
1700 	}
1701 
1702 	/* copy back header */
1703 	head.fmh_entries = xhead.fmh_entries;
1704 	head.fmh_oflags = xhead.fmh_oflags;
1705 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
1706 		return -EFAULT;
1707 
1708 	return 0;
1709 }
1710 
1711 STATIC int
1712 xfs_ioc_scrub_metadata(
1713 	struct xfs_inode		*ip,
1714 	void				__user *arg)
1715 {
1716 	struct xfs_scrub_metadata	scrub;
1717 	int				error;
1718 
1719 	if (!capable(CAP_SYS_ADMIN))
1720 		return -EPERM;
1721 
1722 	if (copy_from_user(&scrub, arg, sizeof(scrub)))
1723 		return -EFAULT;
1724 
1725 	error = xfs_scrub_metadata(ip, &scrub);
1726 	if (error)
1727 		return error;
1728 
1729 	if (copy_to_user(arg, &scrub, sizeof(scrub)))
1730 		return -EFAULT;
1731 
1732 	return 0;
1733 }
1734 
1735 int
1736 xfs_ioc_swapext(
1737 	xfs_swapext_t	*sxp)
1738 {
1739 	xfs_inode_t     *ip, *tip;
1740 	struct fd	f, tmp;
1741 	int		error = 0;
1742 
1743 	/* Pull information for the target fd */
1744 	f = fdget((int)sxp->sx_fdtarget);
1745 	if (!f.file) {
1746 		error = -EINVAL;
1747 		goto out;
1748 	}
1749 
1750 	if (!(f.file->f_mode & FMODE_WRITE) ||
1751 	    !(f.file->f_mode & FMODE_READ) ||
1752 	    (f.file->f_flags & O_APPEND)) {
1753 		error = -EBADF;
1754 		goto out_put_file;
1755 	}
1756 
1757 	tmp = fdget((int)sxp->sx_fdtmp);
1758 	if (!tmp.file) {
1759 		error = -EINVAL;
1760 		goto out_put_file;
1761 	}
1762 
1763 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1764 	    !(tmp.file->f_mode & FMODE_READ) ||
1765 	    (tmp.file->f_flags & O_APPEND)) {
1766 		error = -EBADF;
1767 		goto out_put_tmp_file;
1768 	}
1769 
1770 	if (IS_SWAPFILE(file_inode(f.file)) ||
1771 	    IS_SWAPFILE(file_inode(tmp.file))) {
1772 		error = -EINVAL;
1773 		goto out_put_tmp_file;
1774 	}
1775 
1776 	/*
1777 	 * We need to ensure that the fds passed in point to XFS inodes
1778 	 * before we cast and access them as XFS structures as we have no
1779 	 * control over what the user passes us here.
1780 	 */
1781 	if (f.file->f_op != &xfs_file_operations ||
1782 	    tmp.file->f_op != &xfs_file_operations) {
1783 		error = -EINVAL;
1784 		goto out_put_tmp_file;
1785 	}
1786 
1787 	ip = XFS_I(file_inode(f.file));
1788 	tip = XFS_I(file_inode(tmp.file));
1789 
1790 	if (ip->i_mount != tip->i_mount) {
1791 		error = -EINVAL;
1792 		goto out_put_tmp_file;
1793 	}
1794 
1795 	if (ip->i_ino == tip->i_ino) {
1796 		error = -EINVAL;
1797 		goto out_put_tmp_file;
1798 	}
1799 
1800 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1801 		error = -EIO;
1802 		goto out_put_tmp_file;
1803 	}
1804 
1805 	error = xfs_swap_extents(ip, tip, sxp);
1806 
1807  out_put_tmp_file:
1808 	fdput(tmp);
1809  out_put_file:
1810 	fdput(f);
1811  out:
1812 	return error;
1813 }
1814 
1815 static int
1816 xfs_ioc_getlabel(
1817 	struct xfs_mount	*mp,
1818 	char			__user *user_label)
1819 {
1820 	struct xfs_sb		*sbp = &mp->m_sb;
1821 	char			label[XFSLABEL_MAX + 1];
1822 
1823 	/* Paranoia */
1824 	BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1825 
1826 	spin_lock(&mp->m_sb_lock);
1827 	strncpy(label, sbp->sb_fname, sizeof(sbp->sb_fname));
1828 	spin_unlock(&mp->m_sb_lock);
1829 
1830 	/* xfs on-disk label is 12 chars, be sure we send a null to user */
1831 	label[XFSLABEL_MAX] = '\0';
1832 	if (copy_to_user(user_label, label, sizeof(sbp->sb_fname)))
1833 		return -EFAULT;
1834 	return 0;
1835 }
1836 
1837 static int
1838 xfs_ioc_setlabel(
1839 	struct file		*filp,
1840 	struct xfs_mount	*mp,
1841 	char			__user *newlabel)
1842 {
1843 	struct xfs_sb		*sbp = &mp->m_sb;
1844 	char			label[XFSLABEL_MAX + 1];
1845 	size_t			len;
1846 	int			error;
1847 
1848 	if (!capable(CAP_SYS_ADMIN))
1849 		return -EPERM;
1850 	/*
1851 	 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1852 	 * smaller, at 12 bytes.  We copy one more to be sure we find the
1853 	 * (required) NULL character to test the incoming label length.
1854 	 * NB: The on disk label doesn't need to be null terminated.
1855 	 */
1856 	if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1857 		return -EFAULT;
1858 	len = strnlen(label, XFSLABEL_MAX + 1);
1859 	if (len > sizeof(sbp->sb_fname))
1860 		return -EINVAL;
1861 
1862 	error = mnt_want_write_file(filp);
1863 	if (error)
1864 		return error;
1865 
1866 	spin_lock(&mp->m_sb_lock);
1867 	memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1868 	strncpy(sbp->sb_fname, label, sizeof(sbp->sb_fname));
1869 	spin_unlock(&mp->m_sb_lock);
1870 
1871 	/*
1872 	 * Now we do several things to satisfy userspace.
1873 	 * In addition to normal logging of the primary superblock, we also
1874 	 * immediately write these changes to sector zero for the primary, then
1875 	 * update all backup supers (as xfs_db does for a label change), then
1876 	 * invalidate the block device page cache.  This is so that any prior
1877 	 * buffered reads from userspace (i.e. from blkid) are invalidated,
1878 	 * and userspace will see the newly-written label.
1879 	 */
1880 	error = xfs_sync_sb_buf(mp);
1881 	if (error)
1882 		goto out;
1883 	/*
1884 	 * growfs also updates backup supers so lock against that.
1885 	 */
1886 	mutex_lock(&mp->m_growlock);
1887 	error = xfs_update_secondary_sbs(mp);
1888 	mutex_unlock(&mp->m_growlock);
1889 
1890 	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1891 
1892 out:
1893 	mnt_drop_write_file(filp);
1894 	return error;
1895 }
1896 
1897 /*
1898  * Note: some of the ioctl's return positive numbers as a
1899  * byte count indicating success, such as readlink_by_handle.
1900  * So we don't "sign flip" like most other routines.  This means
1901  * true errors need to be returned as a negative value.
1902  */
1903 long
1904 xfs_file_ioctl(
1905 	struct file		*filp,
1906 	unsigned int		cmd,
1907 	unsigned long		p)
1908 {
1909 	struct inode		*inode = file_inode(filp);
1910 	struct xfs_inode	*ip = XFS_I(inode);
1911 	struct xfs_mount	*mp = ip->i_mount;
1912 	void			__user *arg = (void __user *)p;
1913 	int			error;
1914 
1915 	trace_xfs_file_ioctl(ip);
1916 
1917 	switch (cmd) {
1918 	case FITRIM:
1919 		return xfs_ioc_trim(mp, arg);
1920 	case FS_IOC_GETFSLABEL:
1921 		return xfs_ioc_getlabel(mp, arg);
1922 	case FS_IOC_SETFSLABEL:
1923 		return xfs_ioc_setlabel(filp, mp, arg);
1924 	case XFS_IOC_ALLOCSP:
1925 	case XFS_IOC_FREESP:
1926 	case XFS_IOC_RESVSP:
1927 	case XFS_IOC_UNRESVSP:
1928 	case XFS_IOC_ALLOCSP64:
1929 	case XFS_IOC_FREESP64:
1930 	case XFS_IOC_RESVSP64:
1931 	case XFS_IOC_UNRESVSP64:
1932 	case XFS_IOC_ZERO_RANGE: {
1933 		xfs_flock64_t		bf;
1934 
1935 		if (copy_from_user(&bf, arg, sizeof(bf)))
1936 			return -EFAULT;
1937 		return xfs_ioc_space(filp, cmd, &bf);
1938 	}
1939 	case XFS_IOC_DIOINFO: {
1940 		struct dioattr	da;
1941 		xfs_buftarg_t	*target =
1942 			XFS_IS_REALTIME_INODE(ip) ?
1943 			mp->m_rtdev_targp : mp->m_ddev_targp;
1944 
1945 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1946 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1947 
1948 		if (copy_to_user(arg, &da, sizeof(da)))
1949 			return -EFAULT;
1950 		return 0;
1951 	}
1952 
1953 	case XFS_IOC_FSBULKSTAT_SINGLE:
1954 	case XFS_IOC_FSBULKSTAT:
1955 	case XFS_IOC_FSINUMBERS:
1956 		return xfs_ioc_bulkstat(mp, cmd, arg);
1957 
1958 	case XFS_IOC_FSGEOMETRY_V1:
1959 		return xfs_ioc_fsgeometry_v1(mp, arg);
1960 
1961 	case XFS_IOC_FSGEOMETRY:
1962 		return xfs_ioc_fsgeometry(mp, arg);
1963 
1964 	case XFS_IOC_GETVERSION:
1965 		return put_user(inode->i_generation, (int __user *)arg);
1966 
1967 	case XFS_IOC_FSGETXATTR:
1968 		return xfs_ioc_fsgetxattr(ip, 0, arg);
1969 	case XFS_IOC_FSGETXATTRA:
1970 		return xfs_ioc_fsgetxattr(ip, 1, arg);
1971 	case XFS_IOC_FSSETXATTR:
1972 		return xfs_ioc_fssetxattr(ip, filp, arg);
1973 	case XFS_IOC_GETXFLAGS:
1974 		return xfs_ioc_getxflags(ip, arg);
1975 	case XFS_IOC_SETXFLAGS:
1976 		return xfs_ioc_setxflags(ip, filp, arg);
1977 
1978 	case XFS_IOC_FSSETDM: {
1979 		struct fsdmidata	dmi;
1980 
1981 		if (copy_from_user(&dmi, arg, sizeof(dmi)))
1982 			return -EFAULT;
1983 
1984 		error = mnt_want_write_file(filp);
1985 		if (error)
1986 			return error;
1987 
1988 		error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1989 				dmi.fsd_dmstate);
1990 		mnt_drop_write_file(filp);
1991 		return error;
1992 	}
1993 
1994 	case XFS_IOC_GETBMAP:
1995 	case XFS_IOC_GETBMAPA:
1996 	case XFS_IOC_GETBMAPX:
1997 		return xfs_ioc_getbmap(filp, cmd, arg);
1998 
1999 	case FS_IOC_GETFSMAP:
2000 		return xfs_ioc_getfsmap(ip, arg);
2001 
2002 	case XFS_IOC_SCRUB_METADATA:
2003 		return xfs_ioc_scrub_metadata(ip, arg);
2004 
2005 	case XFS_IOC_FD_TO_HANDLE:
2006 	case XFS_IOC_PATH_TO_HANDLE:
2007 	case XFS_IOC_PATH_TO_FSHANDLE: {
2008 		xfs_fsop_handlereq_t	hreq;
2009 
2010 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
2011 			return -EFAULT;
2012 		return xfs_find_handle(cmd, &hreq);
2013 	}
2014 	case XFS_IOC_OPEN_BY_HANDLE: {
2015 		xfs_fsop_handlereq_t	hreq;
2016 
2017 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2018 			return -EFAULT;
2019 		return xfs_open_by_handle(filp, &hreq);
2020 	}
2021 	case XFS_IOC_FSSETDM_BY_HANDLE:
2022 		return xfs_fssetdm_by_handle(filp, arg);
2023 
2024 	case XFS_IOC_READLINK_BY_HANDLE: {
2025 		xfs_fsop_handlereq_t	hreq;
2026 
2027 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2028 			return -EFAULT;
2029 		return xfs_readlink_by_handle(filp, &hreq);
2030 	}
2031 	case XFS_IOC_ATTRLIST_BY_HANDLE:
2032 		return xfs_attrlist_by_handle(filp, arg);
2033 
2034 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
2035 		return xfs_attrmulti_by_handle(filp, arg);
2036 
2037 	case XFS_IOC_SWAPEXT: {
2038 		struct xfs_swapext	sxp;
2039 
2040 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
2041 			return -EFAULT;
2042 		error = mnt_want_write_file(filp);
2043 		if (error)
2044 			return error;
2045 		error = xfs_ioc_swapext(&sxp);
2046 		mnt_drop_write_file(filp);
2047 		return error;
2048 	}
2049 
2050 	case XFS_IOC_FSCOUNTS: {
2051 		xfs_fsop_counts_t out;
2052 
2053 		error = xfs_fs_counts(mp, &out);
2054 		if (error)
2055 			return error;
2056 
2057 		if (copy_to_user(arg, &out, sizeof(out)))
2058 			return -EFAULT;
2059 		return 0;
2060 	}
2061 
2062 	case XFS_IOC_SET_RESBLKS: {
2063 		xfs_fsop_resblks_t inout;
2064 		uint64_t	   in;
2065 
2066 		if (!capable(CAP_SYS_ADMIN))
2067 			return -EPERM;
2068 
2069 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2070 			return -EROFS;
2071 
2072 		if (copy_from_user(&inout, arg, sizeof(inout)))
2073 			return -EFAULT;
2074 
2075 		error = mnt_want_write_file(filp);
2076 		if (error)
2077 			return error;
2078 
2079 		/* input parameter is passed in resblks field of structure */
2080 		in = inout.resblks;
2081 		error = xfs_reserve_blocks(mp, &in, &inout);
2082 		mnt_drop_write_file(filp);
2083 		if (error)
2084 			return error;
2085 
2086 		if (copy_to_user(arg, &inout, sizeof(inout)))
2087 			return -EFAULT;
2088 		return 0;
2089 	}
2090 
2091 	case XFS_IOC_GET_RESBLKS: {
2092 		xfs_fsop_resblks_t out;
2093 
2094 		if (!capable(CAP_SYS_ADMIN))
2095 			return -EPERM;
2096 
2097 		error = xfs_reserve_blocks(mp, NULL, &out);
2098 		if (error)
2099 			return error;
2100 
2101 		if (copy_to_user(arg, &out, sizeof(out)))
2102 			return -EFAULT;
2103 
2104 		return 0;
2105 	}
2106 
2107 	case XFS_IOC_FSGROWFSDATA: {
2108 		xfs_growfs_data_t in;
2109 
2110 		if (copy_from_user(&in, arg, sizeof(in)))
2111 			return -EFAULT;
2112 
2113 		error = mnt_want_write_file(filp);
2114 		if (error)
2115 			return error;
2116 		error = xfs_growfs_data(mp, &in);
2117 		mnt_drop_write_file(filp);
2118 		return error;
2119 	}
2120 
2121 	case XFS_IOC_FSGROWFSLOG: {
2122 		xfs_growfs_log_t in;
2123 
2124 		if (copy_from_user(&in, arg, sizeof(in)))
2125 			return -EFAULT;
2126 
2127 		error = mnt_want_write_file(filp);
2128 		if (error)
2129 			return error;
2130 		error = xfs_growfs_log(mp, &in);
2131 		mnt_drop_write_file(filp);
2132 		return error;
2133 	}
2134 
2135 	case XFS_IOC_FSGROWFSRT: {
2136 		xfs_growfs_rt_t in;
2137 
2138 		if (copy_from_user(&in, arg, sizeof(in)))
2139 			return -EFAULT;
2140 
2141 		error = mnt_want_write_file(filp);
2142 		if (error)
2143 			return error;
2144 		error = xfs_growfs_rt(mp, &in);
2145 		mnt_drop_write_file(filp);
2146 		return error;
2147 	}
2148 
2149 	case XFS_IOC_GOINGDOWN: {
2150 		uint32_t in;
2151 
2152 		if (!capable(CAP_SYS_ADMIN))
2153 			return -EPERM;
2154 
2155 		if (get_user(in, (uint32_t __user *)arg))
2156 			return -EFAULT;
2157 
2158 		return xfs_fs_goingdown(mp, in);
2159 	}
2160 
2161 	case XFS_IOC_ERROR_INJECTION: {
2162 		xfs_error_injection_t in;
2163 
2164 		if (!capable(CAP_SYS_ADMIN))
2165 			return -EPERM;
2166 
2167 		if (copy_from_user(&in, arg, sizeof(in)))
2168 			return -EFAULT;
2169 
2170 		return xfs_errortag_add(mp, in.errtag);
2171 	}
2172 
2173 	case XFS_IOC_ERROR_CLEARALL:
2174 		if (!capable(CAP_SYS_ADMIN))
2175 			return -EPERM;
2176 
2177 		return xfs_errortag_clearall(mp);
2178 
2179 	case XFS_IOC_FREE_EOFBLOCKS: {
2180 		struct xfs_fs_eofblocks eofb;
2181 		struct xfs_eofblocks keofb;
2182 
2183 		if (!capable(CAP_SYS_ADMIN))
2184 			return -EPERM;
2185 
2186 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2187 			return -EROFS;
2188 
2189 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2190 			return -EFAULT;
2191 
2192 		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2193 		if (error)
2194 			return error;
2195 
2196 		return xfs_icache_free_eofblocks(mp, &keofb);
2197 	}
2198 
2199 	default:
2200 		return -ENOTTY;
2201 	}
2202 }
2203