xref: /openbmc/linux/fs/xfs/xfs_ioctl.c (revision 77d84ff8)
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_sb.h"
25 #include "xfs_ag.h"
26 #include "xfs_mount.h"
27 #include "xfs_inode.h"
28 #include "xfs_ioctl.h"
29 #include "xfs_alloc.h"
30 #include "xfs_rtalloc.h"
31 #include "xfs_itable.h"
32 #include "xfs_error.h"
33 #include "xfs_attr.h"
34 #include "xfs_bmap.h"
35 #include "xfs_bmap_util.h"
36 #include "xfs_fsops.h"
37 #include "xfs_discard.h"
38 #include "xfs_quota.h"
39 #include "xfs_export.h"
40 #include "xfs_trace.h"
41 #include "xfs_icache.h"
42 #include "xfs_symlink.h"
43 #include "xfs_dinode.h"
44 #include "xfs_trans.h"
45 
46 #include <linux/capability.h>
47 #include <linux/dcache.h>
48 #include <linux/mount.h>
49 #include <linux/namei.h>
50 #include <linux/pagemap.h>
51 #include <linux/slab.h>
52 #include <linux/exportfs.h>
53 
54 /*
55  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
56  * a file or fs handle.
57  *
58  * XFS_IOC_PATH_TO_FSHANDLE
59  *    returns fs handle for a mount point or path within that mount point
60  * XFS_IOC_FD_TO_HANDLE
61  *    returns full handle for a FD opened in user space
62  * XFS_IOC_PATH_TO_HANDLE
63  *    returns full handle for a path
64  */
65 int
66 xfs_find_handle(
67 	unsigned int		cmd,
68 	xfs_fsop_handlereq_t	*hreq)
69 {
70 	int			hsize;
71 	xfs_handle_t		handle;
72 	struct inode		*inode;
73 	struct fd		f = {NULL};
74 	struct path		path;
75 	int			error;
76 	struct xfs_inode	*ip;
77 
78 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
79 		f = fdget(hreq->fd);
80 		if (!f.file)
81 			return -EBADF;
82 		inode = file_inode(f.file);
83 	} else {
84 		error = user_lpath((const char __user *)hreq->path, &path);
85 		if (error)
86 			return error;
87 		inode = path.dentry->d_inode;
88 	}
89 	ip = XFS_I(inode);
90 
91 	/*
92 	 * We can only generate handles for inodes residing on a XFS filesystem,
93 	 * and only for regular files, directories or symbolic links.
94 	 */
95 	error = -EINVAL;
96 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
97 		goto out_put;
98 
99 	error = -EBADF;
100 	if (!S_ISREG(inode->i_mode) &&
101 	    !S_ISDIR(inode->i_mode) &&
102 	    !S_ISLNK(inode->i_mode))
103 		goto out_put;
104 
105 
106 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
107 
108 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
109 		/*
110 		 * This handle only contains an fsid, zero the rest.
111 		 */
112 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
113 		hsize = sizeof(xfs_fsid_t);
114 	} else {
115 		int		lock_mode;
116 
117 		lock_mode = xfs_ilock_map_shared(ip);
118 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
119 					sizeof(handle.ha_fid.fid_len);
120 		handle.ha_fid.fid_pad = 0;
121 		handle.ha_fid.fid_gen = ip->i_d.di_gen;
122 		handle.ha_fid.fid_ino = ip->i_ino;
123 		xfs_iunlock_map_shared(ip, lock_mode);
124 
125 		hsize = XFS_HSIZE(handle);
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 -XFS_ERROR(EPERM);
215 
216 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
217 	if (IS_ERR(dentry))
218 		return PTR_ERR(dentry);
219 	inode = dentry->d_inode;
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 = -XFS_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 = -XFS_ERROR(EPERM);
236 		goto out_dput;
237 	}
238 
239 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
240 		error = -XFS_ERROR(EACCES);
241 		goto out_dput;
242 	}
243 
244 	/* Can't write directories. */
245 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
246 		error = -XFS_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 /*
279  * This is a copy from fs/namei.c:vfs_readlink(), except for removing it's
280  * unused first argument.
281  */
282 STATIC int
283 do_readlink(
284 	char __user		*buffer,
285 	int			buflen,
286 	const char		*link)
287 {
288         int len;
289 
290 	len = PTR_ERR(link);
291 	if (IS_ERR(link))
292 		goto out;
293 
294 	len = strlen(link);
295 	if (len > (unsigned) buflen)
296 		len = buflen;
297 	if (copy_to_user(buffer, link, len))
298 		len = -EFAULT;
299  out:
300 	return len;
301 }
302 
303 
304 int
305 xfs_readlink_by_handle(
306 	struct file		*parfilp,
307 	xfs_fsop_handlereq_t	*hreq)
308 {
309 	struct dentry		*dentry;
310 	__u32			olen;
311 	void			*link;
312 	int			error;
313 
314 	if (!capable(CAP_SYS_ADMIN))
315 		return -XFS_ERROR(EPERM);
316 
317 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
318 	if (IS_ERR(dentry))
319 		return PTR_ERR(dentry);
320 
321 	/* Restrict this handle operation to symlinks only. */
322 	if (!S_ISLNK(dentry->d_inode->i_mode)) {
323 		error = -XFS_ERROR(EINVAL);
324 		goto out_dput;
325 	}
326 
327 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
328 		error = -XFS_ERROR(EFAULT);
329 		goto out_dput;
330 	}
331 
332 	link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
333 	if (!link) {
334 		error = -XFS_ERROR(ENOMEM);
335 		goto out_dput;
336 	}
337 
338 	error = -xfs_readlink(XFS_I(dentry->d_inode), link);
339 	if (error)
340 		goto out_kfree;
341 	error = do_readlink(hreq->ohandle, olen, link);
342 	if (error)
343 		goto out_kfree;
344 
345  out_kfree:
346 	kfree(link);
347  out_dput:
348 	dput(dentry);
349 	return error;
350 }
351 
352 int
353 xfs_set_dmattrs(
354 	xfs_inode_t     *ip,
355 	u_int		evmask,
356 	u_int16_t	state)
357 {
358 	xfs_mount_t	*mp = ip->i_mount;
359 	xfs_trans_t	*tp;
360 	int		error;
361 
362 	if (!capable(CAP_SYS_ADMIN))
363 		return XFS_ERROR(EPERM);
364 
365 	if (XFS_FORCED_SHUTDOWN(mp))
366 		return XFS_ERROR(EIO);
367 
368 	tp = xfs_trans_alloc(mp, XFS_TRANS_SET_DMATTRS);
369 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
370 	if (error) {
371 		xfs_trans_cancel(tp, 0);
372 		return error;
373 	}
374 	xfs_ilock(ip, XFS_ILOCK_EXCL);
375 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
376 
377 	ip->i_d.di_dmevmask = evmask;
378 	ip->i_d.di_dmstate  = state;
379 
380 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
381 	error = xfs_trans_commit(tp, 0);
382 
383 	return error;
384 }
385 
386 STATIC int
387 xfs_fssetdm_by_handle(
388 	struct file		*parfilp,
389 	void			__user *arg)
390 {
391 	int			error;
392 	struct fsdmidata	fsd;
393 	xfs_fsop_setdm_handlereq_t dmhreq;
394 	struct dentry		*dentry;
395 
396 	if (!capable(CAP_MKNOD))
397 		return -XFS_ERROR(EPERM);
398 	if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
399 		return -XFS_ERROR(EFAULT);
400 
401 	error = mnt_want_write_file(parfilp);
402 	if (error)
403 		return error;
404 
405 	dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
406 	if (IS_ERR(dentry)) {
407 		mnt_drop_write_file(parfilp);
408 		return PTR_ERR(dentry);
409 	}
410 
411 	if (IS_IMMUTABLE(dentry->d_inode) || IS_APPEND(dentry->d_inode)) {
412 		error = -XFS_ERROR(EPERM);
413 		goto out;
414 	}
415 
416 	if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
417 		error = -XFS_ERROR(EFAULT);
418 		goto out;
419 	}
420 
421 	error = -xfs_set_dmattrs(XFS_I(dentry->d_inode), fsd.fsd_dmevmask,
422 				 fsd.fsd_dmstate);
423 
424  out:
425 	mnt_drop_write_file(parfilp);
426 	dput(dentry);
427 	return error;
428 }
429 
430 STATIC int
431 xfs_attrlist_by_handle(
432 	struct file		*parfilp,
433 	void			__user *arg)
434 {
435 	int			error = -ENOMEM;
436 	attrlist_cursor_kern_t	*cursor;
437 	xfs_fsop_attrlist_handlereq_t al_hreq;
438 	struct dentry		*dentry;
439 	char			*kbuf;
440 
441 	if (!capable(CAP_SYS_ADMIN))
442 		return -XFS_ERROR(EPERM);
443 	if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
444 		return -XFS_ERROR(EFAULT);
445 	if (al_hreq.buflen < sizeof(struct attrlist) ||
446 	    al_hreq.buflen > XATTR_LIST_MAX)
447 		return -XFS_ERROR(EINVAL);
448 
449 	/*
450 	 * Reject flags, only allow namespaces.
451 	 */
452 	if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
453 		return -XFS_ERROR(EINVAL);
454 
455 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
456 	if (IS_ERR(dentry))
457 		return PTR_ERR(dentry);
458 
459 	kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
460 	if (!kbuf)
461 		goto out_dput;
462 
463 	cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
464 	error = -xfs_attr_list(XFS_I(dentry->d_inode), kbuf, al_hreq.buflen,
465 					al_hreq.flags, cursor);
466 	if (error)
467 		goto out_kfree;
468 
469 	if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
470 		error = -EFAULT;
471 
472 out_kfree:
473 	kmem_free(kbuf);
474 out_dput:
475 	dput(dentry);
476 	return error;
477 }
478 
479 int
480 xfs_attrmulti_attr_get(
481 	struct inode		*inode,
482 	unsigned char		*name,
483 	unsigned char		__user *ubuf,
484 	__uint32_t		*len,
485 	__uint32_t		flags)
486 {
487 	unsigned char		*kbuf;
488 	int			error = EFAULT;
489 
490 	if (*len > XATTR_SIZE_MAX)
491 		return EINVAL;
492 	kbuf = kmem_zalloc_large(*len, KM_SLEEP);
493 	if (!kbuf)
494 		return ENOMEM;
495 
496 	error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
497 	if (error)
498 		goto out_kfree;
499 
500 	if (copy_to_user(ubuf, kbuf, *len))
501 		error = EFAULT;
502 
503 out_kfree:
504 	kmem_free(kbuf);
505 	return error;
506 }
507 
508 int
509 xfs_attrmulti_attr_set(
510 	struct inode		*inode,
511 	unsigned char		*name,
512 	const unsigned char	__user *ubuf,
513 	__uint32_t		len,
514 	__uint32_t		flags)
515 {
516 	unsigned char		*kbuf;
517 	int			error = EFAULT;
518 
519 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
520 		return EPERM;
521 	if (len > XATTR_SIZE_MAX)
522 		return EINVAL;
523 
524 	kbuf = memdup_user(ubuf, len);
525 	if (IS_ERR(kbuf))
526 		return PTR_ERR(kbuf);
527 
528 	error = xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
529 
530 	return error;
531 }
532 
533 int
534 xfs_attrmulti_attr_remove(
535 	struct inode		*inode,
536 	unsigned char		*name,
537 	__uint32_t		flags)
538 {
539 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
540 		return EPERM;
541 	return xfs_attr_remove(XFS_I(inode), name, flags);
542 }
543 
544 STATIC int
545 xfs_attrmulti_by_handle(
546 	struct file		*parfilp,
547 	void			__user *arg)
548 {
549 	int			error;
550 	xfs_attr_multiop_t	*ops;
551 	xfs_fsop_attrmulti_handlereq_t am_hreq;
552 	struct dentry		*dentry;
553 	unsigned int		i, size;
554 	unsigned char		*attr_name;
555 
556 	if (!capable(CAP_SYS_ADMIN))
557 		return -XFS_ERROR(EPERM);
558 	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
559 		return -XFS_ERROR(EFAULT);
560 
561 	/* overflow check */
562 	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
563 		return -E2BIG;
564 
565 	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
566 	if (IS_ERR(dentry))
567 		return PTR_ERR(dentry);
568 
569 	error = E2BIG;
570 	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
571 	if (!size || size > 16 * PAGE_SIZE)
572 		goto out_dput;
573 
574 	ops = memdup_user(am_hreq.ops, size);
575 	if (IS_ERR(ops)) {
576 		error = PTR_ERR(ops);
577 		goto out_dput;
578 	}
579 
580 	attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
581 	if (!attr_name)
582 		goto out_kfree_ops;
583 
584 	error = 0;
585 	for (i = 0; i < am_hreq.opcount; i++) {
586 		ops[i].am_error = strncpy_from_user((char *)attr_name,
587 				ops[i].am_attrname, MAXNAMELEN);
588 		if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
589 			error = -ERANGE;
590 		if (ops[i].am_error < 0)
591 			break;
592 
593 		switch (ops[i].am_opcode) {
594 		case ATTR_OP_GET:
595 			ops[i].am_error = xfs_attrmulti_attr_get(
596 					dentry->d_inode, attr_name,
597 					ops[i].am_attrvalue, &ops[i].am_length,
598 					ops[i].am_flags);
599 			break;
600 		case ATTR_OP_SET:
601 			ops[i].am_error = mnt_want_write_file(parfilp);
602 			if (ops[i].am_error)
603 				break;
604 			ops[i].am_error = xfs_attrmulti_attr_set(
605 					dentry->d_inode, attr_name,
606 					ops[i].am_attrvalue, ops[i].am_length,
607 					ops[i].am_flags);
608 			mnt_drop_write_file(parfilp);
609 			break;
610 		case ATTR_OP_REMOVE:
611 			ops[i].am_error = mnt_want_write_file(parfilp);
612 			if (ops[i].am_error)
613 				break;
614 			ops[i].am_error = xfs_attrmulti_attr_remove(
615 					dentry->d_inode, attr_name,
616 					ops[i].am_flags);
617 			mnt_drop_write_file(parfilp);
618 			break;
619 		default:
620 			ops[i].am_error = EINVAL;
621 		}
622 	}
623 
624 	if (copy_to_user(am_hreq.ops, ops, size))
625 		error = XFS_ERROR(EFAULT);
626 
627 	kfree(attr_name);
628  out_kfree_ops:
629 	kfree(ops);
630  out_dput:
631 	dput(dentry);
632 	return -error;
633 }
634 
635 int
636 xfs_ioc_space(
637 	struct xfs_inode	*ip,
638 	struct inode		*inode,
639 	struct file		*filp,
640 	int			ioflags,
641 	unsigned int		cmd,
642 	xfs_flock64_t		*bf)
643 {
644 	struct xfs_mount	*mp = ip->i_mount;
645 	struct xfs_trans	*tp;
646 	struct iattr		iattr;
647 	bool			setprealloc = false;
648 	bool			clrprealloc = false;
649 	int			error;
650 
651 	/*
652 	 * Only allow the sys admin to reserve space unless
653 	 * unwritten extents are enabled.
654 	 */
655 	if (!xfs_sb_version_hasextflgbit(&ip->i_mount->m_sb) &&
656 	    !capable(CAP_SYS_ADMIN))
657 		return -XFS_ERROR(EPERM);
658 
659 	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
660 		return -XFS_ERROR(EPERM);
661 
662 	if (!(filp->f_mode & FMODE_WRITE))
663 		return -XFS_ERROR(EBADF);
664 
665 	if (!S_ISREG(inode->i_mode))
666 		return -XFS_ERROR(EINVAL);
667 
668 	error = mnt_want_write_file(filp);
669 	if (error)
670 		return error;
671 
672 	xfs_ilock(ip, XFS_IOLOCK_EXCL);
673 
674 	switch (bf->l_whence) {
675 	case 0: /*SEEK_SET*/
676 		break;
677 	case 1: /*SEEK_CUR*/
678 		bf->l_start += filp->f_pos;
679 		break;
680 	case 2: /*SEEK_END*/
681 		bf->l_start += XFS_ISIZE(ip);
682 		break;
683 	default:
684 		error = XFS_ERROR(EINVAL);
685 		goto out_unlock;
686 	}
687 
688 	/*
689 	 * length of <= 0 for resv/unresv/zero is invalid.  length for
690 	 * alloc/free is ignored completely and we have no idea what userspace
691 	 * might have set it to, so set it to zero to allow range
692 	 * checks to pass.
693 	 */
694 	switch (cmd) {
695 	case XFS_IOC_ZERO_RANGE:
696 	case XFS_IOC_RESVSP:
697 	case XFS_IOC_RESVSP64:
698 	case XFS_IOC_UNRESVSP:
699 	case XFS_IOC_UNRESVSP64:
700 		if (bf->l_len <= 0) {
701 			error = XFS_ERROR(EINVAL);
702 			goto out_unlock;
703 		}
704 		break;
705 	default:
706 		bf->l_len = 0;
707 		break;
708 	}
709 
710 	if (bf->l_start < 0 ||
711 	    bf->l_start > mp->m_super->s_maxbytes ||
712 	    bf->l_start + bf->l_len < 0 ||
713 	    bf->l_start + bf->l_len >= mp->m_super->s_maxbytes) {
714 		error = XFS_ERROR(EINVAL);
715 		goto out_unlock;
716 	}
717 
718 	switch (cmd) {
719 	case XFS_IOC_ZERO_RANGE:
720 		error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
721 		if (!error)
722 			setprealloc = true;
723 		break;
724 	case XFS_IOC_RESVSP:
725 	case XFS_IOC_RESVSP64:
726 		error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
727 						XFS_BMAPI_PREALLOC);
728 		if (!error)
729 			setprealloc = true;
730 		break;
731 	case XFS_IOC_UNRESVSP:
732 	case XFS_IOC_UNRESVSP64:
733 		error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
734 		break;
735 	case XFS_IOC_ALLOCSP:
736 	case XFS_IOC_ALLOCSP64:
737 	case XFS_IOC_FREESP:
738 	case XFS_IOC_FREESP64:
739 		if (bf->l_start > XFS_ISIZE(ip)) {
740 			error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
741 					bf->l_start - XFS_ISIZE(ip), 0);
742 			if (error)
743 				goto out_unlock;
744 		}
745 
746 		iattr.ia_valid = ATTR_SIZE;
747 		iattr.ia_size = bf->l_start;
748 
749 		error = xfs_setattr_size(ip, &iattr);
750 		if (!error)
751 			clrprealloc = true;
752 		break;
753 	default:
754 		ASSERT(0);
755 		error = XFS_ERROR(EINVAL);
756 	}
757 
758 	if (error)
759 		goto out_unlock;
760 
761 	tp = xfs_trans_alloc(mp, XFS_TRANS_WRITEID);
762 	error = xfs_trans_reserve(tp, &M_RES(mp)->tr_writeid, 0, 0);
763 	if (error) {
764 		xfs_trans_cancel(tp, 0);
765 		goto out_unlock;
766 	}
767 
768 	xfs_ilock(ip, XFS_ILOCK_EXCL);
769 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
770 
771 	if (!(ioflags & IO_INVIS)) {
772 		ip->i_d.di_mode &= ~S_ISUID;
773 		if (ip->i_d.di_mode & S_IXGRP)
774 			ip->i_d.di_mode &= ~S_ISGID;
775 		xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_MOD | XFS_ICHGTIME_CHG);
776 	}
777 
778 	if (setprealloc)
779 		ip->i_d.di_flags |= XFS_DIFLAG_PREALLOC;
780 	else if (clrprealloc)
781 		ip->i_d.di_flags &= ~XFS_DIFLAG_PREALLOC;
782 
783 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
784 	if (filp->f_flags & O_DSYNC)
785 		xfs_trans_set_sync(tp);
786 	error = xfs_trans_commit(tp, 0);
787 
788 out_unlock:
789 	xfs_iunlock(ip, XFS_IOLOCK_EXCL);
790 	mnt_drop_write_file(filp);
791 	return -error;
792 }
793 
794 STATIC int
795 xfs_ioc_bulkstat(
796 	xfs_mount_t		*mp,
797 	unsigned int		cmd,
798 	void			__user *arg)
799 {
800 	xfs_fsop_bulkreq_t	bulkreq;
801 	int			count;	/* # of records returned */
802 	xfs_ino_t		inlast;	/* last inode number */
803 	int			done;
804 	int			error;
805 
806 	/* done = 1 if there are more stats to get and if bulkstat */
807 	/* should be called again (unused here, but used in dmapi) */
808 
809 	if (!capable(CAP_SYS_ADMIN))
810 		return -EPERM;
811 
812 	if (XFS_FORCED_SHUTDOWN(mp))
813 		return -XFS_ERROR(EIO);
814 
815 	if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
816 		return -XFS_ERROR(EFAULT);
817 
818 	if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
819 		return -XFS_ERROR(EFAULT);
820 
821 	if ((count = bulkreq.icount) <= 0)
822 		return -XFS_ERROR(EINVAL);
823 
824 	if (bulkreq.ubuffer == NULL)
825 		return -XFS_ERROR(EINVAL);
826 
827 	if (cmd == XFS_IOC_FSINUMBERS)
828 		error = xfs_inumbers(mp, &inlast, &count,
829 					bulkreq.ubuffer, xfs_inumbers_fmt);
830 	else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
831 		error = xfs_bulkstat_single(mp, &inlast,
832 						bulkreq.ubuffer, &done);
833 	else	/* XFS_IOC_FSBULKSTAT */
834 		error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
835 				     sizeof(xfs_bstat_t), bulkreq.ubuffer,
836 				     &done);
837 
838 	if (error)
839 		return -error;
840 
841 	if (bulkreq.ocount != NULL) {
842 		if (copy_to_user(bulkreq.lastip, &inlast,
843 						sizeof(xfs_ino_t)))
844 			return -XFS_ERROR(EFAULT);
845 
846 		if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
847 			return -XFS_ERROR(EFAULT);
848 	}
849 
850 	return 0;
851 }
852 
853 STATIC int
854 xfs_ioc_fsgeometry_v1(
855 	xfs_mount_t		*mp,
856 	void			__user *arg)
857 {
858 	xfs_fsop_geom_t         fsgeo;
859 	int			error;
860 
861 	error = xfs_fs_geometry(mp, &fsgeo, 3);
862 	if (error)
863 		return -error;
864 
865 	/*
866 	 * Caller should have passed an argument of type
867 	 * xfs_fsop_geom_v1_t.  This is a proper subset of the
868 	 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
869 	 */
870 	if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
871 		return -XFS_ERROR(EFAULT);
872 	return 0;
873 }
874 
875 STATIC int
876 xfs_ioc_fsgeometry(
877 	xfs_mount_t		*mp,
878 	void			__user *arg)
879 {
880 	xfs_fsop_geom_t		fsgeo;
881 	int			error;
882 
883 	error = xfs_fs_geometry(mp, &fsgeo, 4);
884 	if (error)
885 		return -error;
886 
887 	if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
888 		return -XFS_ERROR(EFAULT);
889 	return 0;
890 }
891 
892 /*
893  * Linux extended inode flags interface.
894  */
895 
896 STATIC unsigned int
897 xfs_merge_ioc_xflags(
898 	unsigned int	flags,
899 	unsigned int	start)
900 {
901 	unsigned int	xflags = start;
902 
903 	if (flags & FS_IMMUTABLE_FL)
904 		xflags |= XFS_XFLAG_IMMUTABLE;
905 	else
906 		xflags &= ~XFS_XFLAG_IMMUTABLE;
907 	if (flags & FS_APPEND_FL)
908 		xflags |= XFS_XFLAG_APPEND;
909 	else
910 		xflags &= ~XFS_XFLAG_APPEND;
911 	if (flags & FS_SYNC_FL)
912 		xflags |= XFS_XFLAG_SYNC;
913 	else
914 		xflags &= ~XFS_XFLAG_SYNC;
915 	if (flags & FS_NOATIME_FL)
916 		xflags |= XFS_XFLAG_NOATIME;
917 	else
918 		xflags &= ~XFS_XFLAG_NOATIME;
919 	if (flags & FS_NODUMP_FL)
920 		xflags |= XFS_XFLAG_NODUMP;
921 	else
922 		xflags &= ~XFS_XFLAG_NODUMP;
923 
924 	return xflags;
925 }
926 
927 STATIC unsigned int
928 xfs_di2lxflags(
929 	__uint16_t	di_flags)
930 {
931 	unsigned int	flags = 0;
932 
933 	if (di_flags & XFS_DIFLAG_IMMUTABLE)
934 		flags |= FS_IMMUTABLE_FL;
935 	if (di_flags & XFS_DIFLAG_APPEND)
936 		flags |= FS_APPEND_FL;
937 	if (di_flags & XFS_DIFLAG_SYNC)
938 		flags |= FS_SYNC_FL;
939 	if (di_flags & XFS_DIFLAG_NOATIME)
940 		flags |= FS_NOATIME_FL;
941 	if (di_flags & XFS_DIFLAG_NODUMP)
942 		flags |= FS_NODUMP_FL;
943 	return flags;
944 }
945 
946 STATIC int
947 xfs_ioc_fsgetxattr(
948 	xfs_inode_t		*ip,
949 	int			attr,
950 	void			__user *arg)
951 {
952 	struct fsxattr		fa;
953 
954 	memset(&fa, 0, sizeof(struct fsxattr));
955 
956 	xfs_ilock(ip, XFS_ILOCK_SHARED);
957 	fa.fsx_xflags = xfs_ip2xflags(ip);
958 	fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
959 	fa.fsx_projid = xfs_get_projid(ip);
960 
961 	if (attr) {
962 		if (ip->i_afp) {
963 			if (ip->i_afp->if_flags & XFS_IFEXTENTS)
964 				fa.fsx_nextents = ip->i_afp->if_bytes /
965 							sizeof(xfs_bmbt_rec_t);
966 			else
967 				fa.fsx_nextents = ip->i_d.di_anextents;
968 		} else
969 			fa.fsx_nextents = 0;
970 	} else {
971 		if (ip->i_df.if_flags & XFS_IFEXTENTS)
972 			fa.fsx_nextents = ip->i_df.if_bytes /
973 						sizeof(xfs_bmbt_rec_t);
974 		else
975 			fa.fsx_nextents = ip->i_d.di_nextents;
976 	}
977 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
978 
979 	if (copy_to_user(arg, &fa, sizeof(fa)))
980 		return -EFAULT;
981 	return 0;
982 }
983 
984 STATIC void
985 xfs_set_diflags(
986 	struct xfs_inode	*ip,
987 	unsigned int		xflags)
988 {
989 	unsigned int		di_flags;
990 
991 	/* can't set PREALLOC this way, just preserve it */
992 	di_flags = (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
993 	if (xflags & XFS_XFLAG_IMMUTABLE)
994 		di_flags |= XFS_DIFLAG_IMMUTABLE;
995 	if (xflags & XFS_XFLAG_APPEND)
996 		di_flags |= XFS_DIFLAG_APPEND;
997 	if (xflags & XFS_XFLAG_SYNC)
998 		di_flags |= XFS_DIFLAG_SYNC;
999 	if (xflags & XFS_XFLAG_NOATIME)
1000 		di_flags |= XFS_DIFLAG_NOATIME;
1001 	if (xflags & XFS_XFLAG_NODUMP)
1002 		di_flags |= XFS_DIFLAG_NODUMP;
1003 	if (xflags & XFS_XFLAG_PROJINHERIT)
1004 		di_flags |= XFS_DIFLAG_PROJINHERIT;
1005 	if (xflags & XFS_XFLAG_NODEFRAG)
1006 		di_flags |= XFS_DIFLAG_NODEFRAG;
1007 	if (xflags & XFS_XFLAG_FILESTREAM)
1008 		di_flags |= XFS_DIFLAG_FILESTREAM;
1009 	if (S_ISDIR(ip->i_d.di_mode)) {
1010 		if (xflags & XFS_XFLAG_RTINHERIT)
1011 			di_flags |= XFS_DIFLAG_RTINHERIT;
1012 		if (xflags & XFS_XFLAG_NOSYMLINKS)
1013 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
1014 		if (xflags & XFS_XFLAG_EXTSZINHERIT)
1015 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
1016 	} else if (S_ISREG(ip->i_d.di_mode)) {
1017 		if (xflags & XFS_XFLAG_REALTIME)
1018 			di_flags |= XFS_DIFLAG_REALTIME;
1019 		if (xflags & XFS_XFLAG_EXTSIZE)
1020 			di_flags |= XFS_DIFLAG_EXTSIZE;
1021 	}
1022 
1023 	ip->i_d.di_flags = di_flags;
1024 }
1025 
1026 STATIC void
1027 xfs_diflags_to_linux(
1028 	struct xfs_inode	*ip)
1029 {
1030 	struct inode		*inode = VFS_I(ip);
1031 	unsigned int		xflags = xfs_ip2xflags(ip);
1032 
1033 	if (xflags & XFS_XFLAG_IMMUTABLE)
1034 		inode->i_flags |= S_IMMUTABLE;
1035 	else
1036 		inode->i_flags &= ~S_IMMUTABLE;
1037 	if (xflags & XFS_XFLAG_APPEND)
1038 		inode->i_flags |= S_APPEND;
1039 	else
1040 		inode->i_flags &= ~S_APPEND;
1041 	if (xflags & XFS_XFLAG_SYNC)
1042 		inode->i_flags |= S_SYNC;
1043 	else
1044 		inode->i_flags &= ~S_SYNC;
1045 	if (xflags & XFS_XFLAG_NOATIME)
1046 		inode->i_flags |= S_NOATIME;
1047 	else
1048 		inode->i_flags &= ~S_NOATIME;
1049 }
1050 
1051 #define FSX_PROJID	1
1052 #define FSX_EXTSIZE	2
1053 #define FSX_XFLAGS	4
1054 #define FSX_NONBLOCK	8
1055 
1056 STATIC int
1057 xfs_ioctl_setattr(
1058 	xfs_inode_t		*ip,
1059 	struct fsxattr		*fa,
1060 	int			mask)
1061 {
1062 	struct xfs_mount	*mp = ip->i_mount;
1063 	struct xfs_trans	*tp;
1064 	unsigned int		lock_flags = 0;
1065 	struct xfs_dquot	*udqp = NULL;
1066 	struct xfs_dquot	*pdqp = NULL;
1067 	struct xfs_dquot	*olddquot = NULL;
1068 	int			code;
1069 
1070 	trace_xfs_ioctl_setattr(ip);
1071 
1072 	if (mp->m_flags & XFS_MOUNT_RDONLY)
1073 		return XFS_ERROR(EROFS);
1074 	if (XFS_FORCED_SHUTDOWN(mp))
1075 		return XFS_ERROR(EIO);
1076 
1077 	/*
1078 	 * Disallow 32bit project ids when projid32bit feature is not enabled.
1079 	 */
1080 	if ((mask & FSX_PROJID) && (fa->fsx_projid > (__uint16_t)-1) &&
1081 			!xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1082 		return XFS_ERROR(EINVAL);
1083 
1084 	/*
1085 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1086 	 * before we start any other transactions. Trying to do this later
1087 	 * is messy. We don't care to take a readlock to look at the ids
1088 	 * in inode here, because we can't hold it across the trans_reserve.
1089 	 * If the IDs do change before we take the ilock, we're covered
1090 	 * because the i_*dquot fields will get updated anyway.
1091 	 */
1092 	if (XFS_IS_QUOTA_ON(mp) && (mask & FSX_PROJID)) {
1093 		code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1094 					 ip->i_d.di_gid, fa->fsx_projid,
1095 					 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1096 		if (code)
1097 			return code;
1098 	}
1099 
1100 	/*
1101 	 * For the other attributes, we acquire the inode lock and
1102 	 * first do an error checking pass.
1103 	 */
1104 	tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
1105 	code = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
1106 	if (code)
1107 		goto error_return;
1108 
1109 	lock_flags = XFS_ILOCK_EXCL;
1110 	xfs_ilock(ip, lock_flags);
1111 
1112 	/*
1113 	 * CAP_FOWNER overrides the following restrictions:
1114 	 *
1115 	 * The user ID of the calling process must be equal
1116 	 * to the file owner ID, except in cases where the
1117 	 * CAP_FSETID capability is applicable.
1118 	 */
1119 	if (!inode_owner_or_capable(VFS_I(ip))) {
1120 		code = XFS_ERROR(EPERM);
1121 		goto error_return;
1122 	}
1123 
1124 	/*
1125 	 * Do a quota reservation only if projid is actually going to change.
1126 	 * Only allow changing of projid from init_user_ns since it is a
1127 	 * non user namespace aware identifier.
1128 	 */
1129 	if (mask & FSX_PROJID) {
1130 		if (current_user_ns() != &init_user_ns) {
1131 			code = XFS_ERROR(EINVAL);
1132 			goto error_return;
1133 		}
1134 
1135 		if (XFS_IS_QUOTA_RUNNING(mp) &&
1136 		    XFS_IS_PQUOTA_ON(mp) &&
1137 		    xfs_get_projid(ip) != fa->fsx_projid) {
1138 			ASSERT(tp);
1139 			code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL,
1140 						pdqp, capable(CAP_FOWNER) ?
1141 						XFS_QMOPT_FORCE_RES : 0);
1142 			if (code)	/* out of quota */
1143 				goto error_return;
1144 		}
1145 	}
1146 
1147 	if (mask & FSX_EXTSIZE) {
1148 		/*
1149 		 * Can't change extent size if any extents are allocated.
1150 		 */
1151 		if (ip->i_d.di_nextents &&
1152 		    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) !=
1153 		     fa->fsx_extsize)) {
1154 			code = XFS_ERROR(EINVAL);	/* EFBIG? */
1155 			goto error_return;
1156 		}
1157 
1158 		/*
1159 		 * Extent size must be a multiple of the appropriate block
1160 		 * size, if set at all. It must also be smaller than the
1161 		 * maximum extent size supported by the filesystem.
1162 		 *
1163 		 * Also, for non-realtime files, limit the extent size hint to
1164 		 * half the size of the AGs in the filesystem so alignment
1165 		 * doesn't result in extents larger than an AG.
1166 		 */
1167 		if (fa->fsx_extsize != 0) {
1168 			xfs_extlen_t    size;
1169 			xfs_fsblock_t   extsize_fsb;
1170 
1171 			extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1172 			if (extsize_fsb > MAXEXTLEN) {
1173 				code = XFS_ERROR(EINVAL);
1174 				goto error_return;
1175 			}
1176 
1177 			if (XFS_IS_REALTIME_INODE(ip) ||
1178 			    ((mask & FSX_XFLAGS) &&
1179 			    (fa->fsx_xflags & XFS_XFLAG_REALTIME))) {
1180 				size = mp->m_sb.sb_rextsize <<
1181 				       mp->m_sb.sb_blocklog;
1182 			} else {
1183 				size = mp->m_sb.sb_blocksize;
1184 				if (extsize_fsb > mp->m_sb.sb_agblocks / 2) {
1185 					code = XFS_ERROR(EINVAL);
1186 					goto error_return;
1187 				}
1188 			}
1189 
1190 			if (fa->fsx_extsize % size) {
1191 				code = XFS_ERROR(EINVAL);
1192 				goto error_return;
1193 			}
1194 		}
1195 	}
1196 
1197 
1198 	if (mask & FSX_XFLAGS) {
1199 		/*
1200 		 * Can't change realtime flag if any extents are allocated.
1201 		 */
1202 		if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1203 		    (XFS_IS_REALTIME_INODE(ip)) !=
1204 		    (fa->fsx_xflags & XFS_XFLAG_REALTIME)) {
1205 			code = XFS_ERROR(EINVAL);	/* EFBIG? */
1206 			goto error_return;
1207 		}
1208 
1209 		/*
1210 		 * If realtime flag is set then must have realtime data.
1211 		 */
1212 		if ((fa->fsx_xflags & XFS_XFLAG_REALTIME)) {
1213 			if ((mp->m_sb.sb_rblocks == 0) ||
1214 			    (mp->m_sb.sb_rextsize == 0) ||
1215 			    (ip->i_d.di_extsize % mp->m_sb.sb_rextsize)) {
1216 				code = XFS_ERROR(EINVAL);
1217 				goto error_return;
1218 			}
1219 		}
1220 
1221 		/*
1222 		 * Can't modify an immutable/append-only file unless
1223 		 * we have appropriate permission.
1224 		 */
1225 		if ((ip->i_d.di_flags &
1226 				(XFS_DIFLAG_IMMUTABLE|XFS_DIFLAG_APPEND) ||
1227 		     (fa->fsx_xflags &
1228 				(XFS_XFLAG_IMMUTABLE | XFS_XFLAG_APPEND))) &&
1229 		    !capable(CAP_LINUX_IMMUTABLE)) {
1230 			code = XFS_ERROR(EPERM);
1231 			goto error_return;
1232 		}
1233 	}
1234 
1235 	xfs_trans_ijoin(tp, ip, 0);
1236 
1237 	/*
1238 	 * Change file ownership.  Must be the owner or privileged.
1239 	 */
1240 	if (mask & FSX_PROJID) {
1241 		/*
1242 		 * CAP_FSETID overrides the following restrictions:
1243 		 *
1244 		 * The set-user-ID and set-group-ID bits of a file will be
1245 		 * cleared upon successful return from chown()
1246 		 */
1247 		if ((ip->i_d.di_mode & (S_ISUID|S_ISGID)) &&
1248 		    !inode_capable(VFS_I(ip), CAP_FSETID))
1249 			ip->i_d.di_mode &= ~(S_ISUID|S_ISGID);
1250 
1251 		/*
1252 		 * Change the ownerships and register quota modifications
1253 		 * in the transaction.
1254 		 */
1255 		if (xfs_get_projid(ip) != fa->fsx_projid) {
1256 			if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1257 				olddquot = xfs_qm_vop_chown(tp, ip,
1258 							&ip->i_pdquot, pdqp);
1259 			}
1260 			xfs_set_projid(ip, fa->fsx_projid);
1261 
1262 			/*
1263 			 * We may have to rev the inode as well as
1264 			 * the superblock version number since projids didn't
1265 			 * exist before DINODE_VERSION_2 and SB_VERSION_NLINK.
1266 			 */
1267 			if (ip->i_d.di_version == 1)
1268 				xfs_bump_ino_vers2(tp, ip);
1269 		}
1270 
1271 	}
1272 
1273 	if (mask & FSX_EXTSIZE)
1274 		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1275 	if (mask & FSX_XFLAGS) {
1276 		xfs_set_diflags(ip, fa->fsx_xflags);
1277 		xfs_diflags_to_linux(ip);
1278 	}
1279 
1280 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1281 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1282 
1283 	XFS_STATS_INC(xs_ig_attrchg);
1284 
1285 	/*
1286 	 * If this is a synchronous mount, make sure that the
1287 	 * transaction goes to disk before returning to the user.
1288 	 * This is slightly sub-optimal in that truncates require
1289 	 * two sync transactions instead of one for wsync filesystems.
1290 	 * One for the truncate and one for the timestamps since we
1291 	 * don't want to change the timestamps unless we're sure the
1292 	 * truncate worked.  Truncates are less than 1% of the laddis
1293 	 * mix so this probably isn't worth the trouble to optimize.
1294 	 */
1295 	if (mp->m_flags & XFS_MOUNT_WSYNC)
1296 		xfs_trans_set_sync(tp);
1297 	code = xfs_trans_commit(tp, 0);
1298 	xfs_iunlock(ip, lock_flags);
1299 
1300 	/*
1301 	 * Release any dquot(s) the inode had kept before chown.
1302 	 */
1303 	xfs_qm_dqrele(olddquot);
1304 	xfs_qm_dqrele(udqp);
1305 	xfs_qm_dqrele(pdqp);
1306 
1307 	return code;
1308 
1309  error_return:
1310 	xfs_qm_dqrele(udqp);
1311 	xfs_qm_dqrele(pdqp);
1312 	xfs_trans_cancel(tp, 0);
1313 	if (lock_flags)
1314 		xfs_iunlock(ip, lock_flags);
1315 	return code;
1316 }
1317 
1318 STATIC int
1319 xfs_ioc_fssetxattr(
1320 	xfs_inode_t		*ip,
1321 	struct file		*filp,
1322 	void			__user *arg)
1323 {
1324 	struct fsxattr		fa;
1325 	unsigned int		mask;
1326 	int error;
1327 
1328 	if (copy_from_user(&fa, arg, sizeof(fa)))
1329 		return -EFAULT;
1330 
1331 	mask = FSX_XFLAGS | FSX_EXTSIZE | FSX_PROJID;
1332 	if (filp->f_flags & (O_NDELAY|O_NONBLOCK))
1333 		mask |= FSX_NONBLOCK;
1334 
1335 	error = mnt_want_write_file(filp);
1336 	if (error)
1337 		return error;
1338 	error = xfs_ioctl_setattr(ip, &fa, mask);
1339 	mnt_drop_write_file(filp);
1340 	return -error;
1341 }
1342 
1343 STATIC int
1344 xfs_ioc_getxflags(
1345 	xfs_inode_t		*ip,
1346 	void			__user *arg)
1347 {
1348 	unsigned int		flags;
1349 
1350 	flags = xfs_di2lxflags(ip->i_d.di_flags);
1351 	if (copy_to_user(arg, &flags, sizeof(flags)))
1352 		return -EFAULT;
1353 	return 0;
1354 }
1355 
1356 STATIC int
1357 xfs_ioc_setxflags(
1358 	xfs_inode_t		*ip,
1359 	struct file		*filp,
1360 	void			__user *arg)
1361 {
1362 	struct fsxattr		fa;
1363 	unsigned int		flags;
1364 	unsigned int		mask;
1365 	int error;
1366 
1367 	if (copy_from_user(&flags, arg, sizeof(flags)))
1368 		return -EFAULT;
1369 
1370 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1371 		      FS_NOATIME_FL | FS_NODUMP_FL | \
1372 		      FS_SYNC_FL))
1373 		return -EOPNOTSUPP;
1374 
1375 	mask = FSX_XFLAGS;
1376 	if (filp->f_flags & (O_NDELAY|O_NONBLOCK))
1377 		mask |= FSX_NONBLOCK;
1378 	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1379 
1380 	error = mnt_want_write_file(filp);
1381 	if (error)
1382 		return error;
1383 	error = xfs_ioctl_setattr(ip, &fa, mask);
1384 	mnt_drop_write_file(filp);
1385 	return -error;
1386 }
1387 
1388 STATIC int
1389 xfs_getbmap_format(void **ap, struct getbmapx *bmv, int *full)
1390 {
1391 	struct getbmap __user	*base = *ap;
1392 
1393 	/* copy only getbmap portion (not getbmapx) */
1394 	if (copy_to_user(base, bmv, sizeof(struct getbmap)))
1395 		return XFS_ERROR(EFAULT);
1396 
1397 	*ap += sizeof(struct getbmap);
1398 	return 0;
1399 }
1400 
1401 STATIC int
1402 xfs_ioc_getbmap(
1403 	struct xfs_inode	*ip,
1404 	int			ioflags,
1405 	unsigned int		cmd,
1406 	void			__user *arg)
1407 {
1408 	struct getbmapx		bmx;
1409 	int			error;
1410 
1411 	if (copy_from_user(&bmx, arg, sizeof(struct getbmapx)))
1412 		return -XFS_ERROR(EFAULT);
1413 
1414 	if (bmx.bmv_count < 2)
1415 		return -XFS_ERROR(EINVAL);
1416 
1417 	bmx.bmv_iflags = (cmd == XFS_IOC_GETBMAPA ? BMV_IF_ATTRFORK : 0);
1418 	if (ioflags & IO_INVIS)
1419 		bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1420 
1421 	error = xfs_getbmap(ip, &bmx, xfs_getbmap_format,
1422 			    (struct getbmap *)arg+1);
1423 	if (error)
1424 		return -error;
1425 
1426 	/* copy back header - only size of getbmap */
1427 	if (copy_to_user(arg, &bmx, sizeof(struct getbmap)))
1428 		return -XFS_ERROR(EFAULT);
1429 	return 0;
1430 }
1431 
1432 STATIC int
1433 xfs_getbmapx_format(void **ap, struct getbmapx *bmv, int *full)
1434 {
1435 	struct getbmapx __user	*base = *ap;
1436 
1437 	if (copy_to_user(base, bmv, sizeof(struct getbmapx)))
1438 		return XFS_ERROR(EFAULT);
1439 
1440 	*ap += sizeof(struct getbmapx);
1441 	return 0;
1442 }
1443 
1444 STATIC int
1445 xfs_ioc_getbmapx(
1446 	struct xfs_inode	*ip,
1447 	void			__user *arg)
1448 {
1449 	struct getbmapx		bmx;
1450 	int			error;
1451 
1452 	if (copy_from_user(&bmx, arg, sizeof(bmx)))
1453 		return -XFS_ERROR(EFAULT);
1454 
1455 	if (bmx.bmv_count < 2)
1456 		return -XFS_ERROR(EINVAL);
1457 
1458 	if (bmx.bmv_iflags & (~BMV_IF_VALID))
1459 		return -XFS_ERROR(EINVAL);
1460 
1461 	error = xfs_getbmap(ip, &bmx, xfs_getbmapx_format,
1462 			    (struct getbmapx *)arg+1);
1463 	if (error)
1464 		return -error;
1465 
1466 	/* copy back header */
1467 	if (copy_to_user(arg, &bmx, sizeof(struct getbmapx)))
1468 		return -XFS_ERROR(EFAULT);
1469 
1470 	return 0;
1471 }
1472 
1473 int
1474 xfs_ioc_swapext(
1475 	xfs_swapext_t	*sxp)
1476 {
1477 	xfs_inode_t     *ip, *tip;
1478 	struct fd	f, tmp;
1479 	int		error = 0;
1480 
1481 	/* Pull information for the target fd */
1482 	f = fdget((int)sxp->sx_fdtarget);
1483 	if (!f.file) {
1484 		error = XFS_ERROR(EINVAL);
1485 		goto out;
1486 	}
1487 
1488 	if (!(f.file->f_mode & FMODE_WRITE) ||
1489 	    !(f.file->f_mode & FMODE_READ) ||
1490 	    (f.file->f_flags & O_APPEND)) {
1491 		error = XFS_ERROR(EBADF);
1492 		goto out_put_file;
1493 	}
1494 
1495 	tmp = fdget((int)sxp->sx_fdtmp);
1496 	if (!tmp.file) {
1497 		error = XFS_ERROR(EINVAL);
1498 		goto out_put_file;
1499 	}
1500 
1501 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1502 	    !(tmp.file->f_mode & FMODE_READ) ||
1503 	    (tmp.file->f_flags & O_APPEND)) {
1504 		error = XFS_ERROR(EBADF);
1505 		goto out_put_tmp_file;
1506 	}
1507 
1508 	if (IS_SWAPFILE(file_inode(f.file)) ||
1509 	    IS_SWAPFILE(file_inode(tmp.file))) {
1510 		error = XFS_ERROR(EINVAL);
1511 		goto out_put_tmp_file;
1512 	}
1513 
1514 	ip = XFS_I(file_inode(f.file));
1515 	tip = XFS_I(file_inode(tmp.file));
1516 
1517 	if (ip->i_mount != tip->i_mount) {
1518 		error = XFS_ERROR(EINVAL);
1519 		goto out_put_tmp_file;
1520 	}
1521 
1522 	if (ip->i_ino == tip->i_ino) {
1523 		error = XFS_ERROR(EINVAL);
1524 		goto out_put_tmp_file;
1525 	}
1526 
1527 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1528 		error = XFS_ERROR(EIO);
1529 		goto out_put_tmp_file;
1530 	}
1531 
1532 	error = xfs_swap_extents(ip, tip, sxp);
1533 
1534  out_put_tmp_file:
1535 	fdput(tmp);
1536  out_put_file:
1537 	fdput(f);
1538  out:
1539 	return error;
1540 }
1541 
1542 /*
1543  * Note: some of the ioctl's return positive numbers as a
1544  * byte count indicating success, such as readlink_by_handle.
1545  * So we don't "sign flip" like most other routines.  This means
1546  * true errors need to be returned as a negative value.
1547  */
1548 long
1549 xfs_file_ioctl(
1550 	struct file		*filp,
1551 	unsigned int		cmd,
1552 	unsigned long		p)
1553 {
1554 	struct inode		*inode = file_inode(filp);
1555 	struct xfs_inode	*ip = XFS_I(inode);
1556 	struct xfs_mount	*mp = ip->i_mount;
1557 	void			__user *arg = (void __user *)p;
1558 	int			ioflags = 0;
1559 	int			error;
1560 
1561 	if (filp->f_mode & FMODE_NOCMTIME)
1562 		ioflags |= IO_INVIS;
1563 
1564 	trace_xfs_file_ioctl(ip);
1565 
1566 	switch (cmd) {
1567 	case FITRIM:
1568 		return xfs_ioc_trim(mp, arg);
1569 	case XFS_IOC_ALLOCSP:
1570 	case XFS_IOC_FREESP:
1571 	case XFS_IOC_RESVSP:
1572 	case XFS_IOC_UNRESVSP:
1573 	case XFS_IOC_ALLOCSP64:
1574 	case XFS_IOC_FREESP64:
1575 	case XFS_IOC_RESVSP64:
1576 	case XFS_IOC_UNRESVSP64:
1577 	case XFS_IOC_ZERO_RANGE: {
1578 		xfs_flock64_t		bf;
1579 
1580 		if (copy_from_user(&bf, arg, sizeof(bf)))
1581 			return -XFS_ERROR(EFAULT);
1582 		return xfs_ioc_space(ip, inode, filp, ioflags, cmd, &bf);
1583 	}
1584 	case XFS_IOC_DIOINFO: {
1585 		struct dioattr	da;
1586 		xfs_buftarg_t	*target =
1587 			XFS_IS_REALTIME_INODE(ip) ?
1588 			mp->m_rtdev_targp : mp->m_ddev_targp;
1589 
1590 		da.d_mem = da.d_miniosz = 1 << target->bt_sshift;
1591 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1592 
1593 		if (copy_to_user(arg, &da, sizeof(da)))
1594 			return -XFS_ERROR(EFAULT);
1595 		return 0;
1596 	}
1597 
1598 	case XFS_IOC_FSBULKSTAT_SINGLE:
1599 	case XFS_IOC_FSBULKSTAT:
1600 	case XFS_IOC_FSINUMBERS:
1601 		return xfs_ioc_bulkstat(mp, cmd, arg);
1602 
1603 	case XFS_IOC_FSGEOMETRY_V1:
1604 		return xfs_ioc_fsgeometry_v1(mp, arg);
1605 
1606 	case XFS_IOC_FSGEOMETRY:
1607 		return xfs_ioc_fsgeometry(mp, arg);
1608 
1609 	case XFS_IOC_GETVERSION:
1610 		return put_user(inode->i_generation, (int __user *)arg);
1611 
1612 	case XFS_IOC_FSGETXATTR:
1613 		return xfs_ioc_fsgetxattr(ip, 0, arg);
1614 	case XFS_IOC_FSGETXATTRA:
1615 		return xfs_ioc_fsgetxattr(ip, 1, arg);
1616 	case XFS_IOC_FSSETXATTR:
1617 		return xfs_ioc_fssetxattr(ip, filp, arg);
1618 	case XFS_IOC_GETXFLAGS:
1619 		return xfs_ioc_getxflags(ip, arg);
1620 	case XFS_IOC_SETXFLAGS:
1621 		return xfs_ioc_setxflags(ip, filp, arg);
1622 
1623 	case XFS_IOC_FSSETDM: {
1624 		struct fsdmidata	dmi;
1625 
1626 		if (copy_from_user(&dmi, arg, sizeof(dmi)))
1627 			return -XFS_ERROR(EFAULT);
1628 
1629 		error = mnt_want_write_file(filp);
1630 		if (error)
1631 			return error;
1632 
1633 		error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1634 				dmi.fsd_dmstate);
1635 		mnt_drop_write_file(filp);
1636 		return -error;
1637 	}
1638 
1639 	case XFS_IOC_GETBMAP:
1640 	case XFS_IOC_GETBMAPA:
1641 		return xfs_ioc_getbmap(ip, ioflags, cmd, arg);
1642 
1643 	case XFS_IOC_GETBMAPX:
1644 		return xfs_ioc_getbmapx(ip, arg);
1645 
1646 	case XFS_IOC_FD_TO_HANDLE:
1647 	case XFS_IOC_PATH_TO_HANDLE:
1648 	case XFS_IOC_PATH_TO_FSHANDLE: {
1649 		xfs_fsop_handlereq_t	hreq;
1650 
1651 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1652 			return -XFS_ERROR(EFAULT);
1653 		return xfs_find_handle(cmd, &hreq);
1654 	}
1655 	case XFS_IOC_OPEN_BY_HANDLE: {
1656 		xfs_fsop_handlereq_t	hreq;
1657 
1658 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1659 			return -XFS_ERROR(EFAULT);
1660 		return xfs_open_by_handle(filp, &hreq);
1661 	}
1662 	case XFS_IOC_FSSETDM_BY_HANDLE:
1663 		return xfs_fssetdm_by_handle(filp, arg);
1664 
1665 	case XFS_IOC_READLINK_BY_HANDLE: {
1666 		xfs_fsop_handlereq_t	hreq;
1667 
1668 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1669 			return -XFS_ERROR(EFAULT);
1670 		return xfs_readlink_by_handle(filp, &hreq);
1671 	}
1672 	case XFS_IOC_ATTRLIST_BY_HANDLE:
1673 		return xfs_attrlist_by_handle(filp, arg);
1674 
1675 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
1676 		return xfs_attrmulti_by_handle(filp, arg);
1677 
1678 	case XFS_IOC_SWAPEXT: {
1679 		struct xfs_swapext	sxp;
1680 
1681 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1682 			return -XFS_ERROR(EFAULT);
1683 		error = mnt_want_write_file(filp);
1684 		if (error)
1685 			return error;
1686 		error = xfs_ioc_swapext(&sxp);
1687 		mnt_drop_write_file(filp);
1688 		return -error;
1689 	}
1690 
1691 	case XFS_IOC_FSCOUNTS: {
1692 		xfs_fsop_counts_t out;
1693 
1694 		error = xfs_fs_counts(mp, &out);
1695 		if (error)
1696 			return -error;
1697 
1698 		if (copy_to_user(arg, &out, sizeof(out)))
1699 			return -XFS_ERROR(EFAULT);
1700 		return 0;
1701 	}
1702 
1703 	case XFS_IOC_SET_RESBLKS: {
1704 		xfs_fsop_resblks_t inout;
1705 		__uint64_t	   in;
1706 
1707 		if (!capable(CAP_SYS_ADMIN))
1708 			return -EPERM;
1709 
1710 		if (mp->m_flags & XFS_MOUNT_RDONLY)
1711 			return -XFS_ERROR(EROFS);
1712 
1713 		if (copy_from_user(&inout, arg, sizeof(inout)))
1714 			return -XFS_ERROR(EFAULT);
1715 
1716 		error = mnt_want_write_file(filp);
1717 		if (error)
1718 			return error;
1719 
1720 		/* input parameter is passed in resblks field of structure */
1721 		in = inout.resblks;
1722 		error = xfs_reserve_blocks(mp, &in, &inout);
1723 		mnt_drop_write_file(filp);
1724 		if (error)
1725 			return -error;
1726 
1727 		if (copy_to_user(arg, &inout, sizeof(inout)))
1728 			return -XFS_ERROR(EFAULT);
1729 		return 0;
1730 	}
1731 
1732 	case XFS_IOC_GET_RESBLKS: {
1733 		xfs_fsop_resblks_t out;
1734 
1735 		if (!capable(CAP_SYS_ADMIN))
1736 			return -EPERM;
1737 
1738 		error = xfs_reserve_blocks(mp, NULL, &out);
1739 		if (error)
1740 			return -error;
1741 
1742 		if (copy_to_user(arg, &out, sizeof(out)))
1743 			return -XFS_ERROR(EFAULT);
1744 
1745 		return 0;
1746 	}
1747 
1748 	case XFS_IOC_FSGROWFSDATA: {
1749 		xfs_growfs_data_t in;
1750 
1751 		if (copy_from_user(&in, arg, sizeof(in)))
1752 			return -XFS_ERROR(EFAULT);
1753 
1754 		error = mnt_want_write_file(filp);
1755 		if (error)
1756 			return error;
1757 		error = xfs_growfs_data(mp, &in);
1758 		mnt_drop_write_file(filp);
1759 		return -error;
1760 	}
1761 
1762 	case XFS_IOC_FSGROWFSLOG: {
1763 		xfs_growfs_log_t in;
1764 
1765 		if (copy_from_user(&in, arg, sizeof(in)))
1766 			return -XFS_ERROR(EFAULT);
1767 
1768 		error = mnt_want_write_file(filp);
1769 		if (error)
1770 			return error;
1771 		error = xfs_growfs_log(mp, &in);
1772 		mnt_drop_write_file(filp);
1773 		return -error;
1774 	}
1775 
1776 	case XFS_IOC_FSGROWFSRT: {
1777 		xfs_growfs_rt_t in;
1778 
1779 		if (copy_from_user(&in, arg, sizeof(in)))
1780 			return -XFS_ERROR(EFAULT);
1781 
1782 		error = mnt_want_write_file(filp);
1783 		if (error)
1784 			return error;
1785 		error = xfs_growfs_rt(mp, &in);
1786 		mnt_drop_write_file(filp);
1787 		return -error;
1788 	}
1789 
1790 	case XFS_IOC_GOINGDOWN: {
1791 		__uint32_t in;
1792 
1793 		if (!capable(CAP_SYS_ADMIN))
1794 			return -EPERM;
1795 
1796 		if (get_user(in, (__uint32_t __user *)arg))
1797 			return -XFS_ERROR(EFAULT);
1798 
1799 		error = xfs_fs_goingdown(mp, in);
1800 		return -error;
1801 	}
1802 
1803 	case XFS_IOC_ERROR_INJECTION: {
1804 		xfs_error_injection_t in;
1805 
1806 		if (!capable(CAP_SYS_ADMIN))
1807 			return -EPERM;
1808 
1809 		if (copy_from_user(&in, arg, sizeof(in)))
1810 			return -XFS_ERROR(EFAULT);
1811 
1812 		error = xfs_errortag_add(in.errtag, mp);
1813 		return -error;
1814 	}
1815 
1816 	case XFS_IOC_ERROR_CLEARALL:
1817 		if (!capable(CAP_SYS_ADMIN))
1818 			return -EPERM;
1819 
1820 		error = xfs_errortag_clearall(mp, 1);
1821 		return -error;
1822 
1823 	case XFS_IOC_FREE_EOFBLOCKS: {
1824 		struct xfs_fs_eofblocks eofb;
1825 		struct xfs_eofblocks keofb;
1826 
1827 		if (!capable(CAP_SYS_ADMIN))
1828 			return -EPERM;
1829 
1830 		if (mp->m_flags & XFS_MOUNT_RDONLY)
1831 			return -XFS_ERROR(EROFS);
1832 
1833 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
1834 			return -XFS_ERROR(EFAULT);
1835 
1836 		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
1837 		if (error)
1838 			return -error;
1839 
1840 		return -xfs_icache_free_eofblocks(mp, &keofb);
1841 	}
1842 
1843 	default:
1844 		return -ENOTTY;
1845 	}
1846 }
1847