xref: /openbmc/linux/fs/xfs/xfs_ioctl.c (revision f0702555)
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 
45 #include <linux/capability.h>
46 #include <linux/dcache.h>
47 #include <linux/mount.h>
48 #include <linux/namei.h>
49 #include <linux/pagemap.h>
50 #include <linux/slab.h>
51 #include <linux/exportfs.h>
52 
53 /*
54  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
55  * a file or fs handle.
56  *
57  * XFS_IOC_PATH_TO_FSHANDLE
58  *    returns fs handle for a mount point or path within that mount point
59  * XFS_IOC_FD_TO_HANDLE
60  *    returns full handle for a FD opened in user space
61  * XFS_IOC_PATH_TO_HANDLE
62  *    returns full handle for a path
63  */
64 int
65 xfs_find_handle(
66 	unsigned int		cmd,
67 	xfs_fsop_handlereq_t	*hreq)
68 {
69 	int			hsize;
70 	xfs_handle_t		handle;
71 	struct inode		*inode;
72 	struct fd		f = {NULL};
73 	struct path		path;
74 	int			error;
75 	struct xfs_inode	*ip;
76 
77 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
78 		f = fdget(hreq->fd);
79 		if (!f.file)
80 			return -EBADF;
81 		inode = file_inode(f.file);
82 	} else {
83 		error = user_lpath((const char __user *)hreq->path, &path);
84 		if (error)
85 			return error;
86 		inode = d_inode(path.dentry);
87 	}
88 	ip = XFS_I(inode);
89 
90 	/*
91 	 * We can only generate handles for inodes residing on a XFS filesystem,
92 	 * and only for regular files, directories or symbolic links.
93 	 */
94 	error = -EINVAL;
95 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
96 		goto out_put;
97 
98 	error = -EBADF;
99 	if (!S_ISREG(inode->i_mode) &&
100 	    !S_ISDIR(inode->i_mode) &&
101 	    !S_ISLNK(inode->i_mode))
102 		goto out_put;
103 
104 
105 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
106 
107 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
108 		/*
109 		 * This handle only contains an fsid, zero the rest.
110 		 */
111 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
112 		hsize = sizeof(xfs_fsid_t);
113 	} else {
114 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
115 					sizeof(handle.ha_fid.fid_len);
116 		handle.ha_fid.fid_pad = 0;
117 		handle.ha_fid.fid_gen = inode->i_generation;
118 		handle.ha_fid.fid_ino = ip->i_ino;
119 
120 		hsize = XFS_HSIZE(handle);
121 	}
122 
123 	error = -EFAULT;
124 	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
125 	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
126 		goto out_put;
127 
128 	error = 0;
129 
130  out_put:
131 	if (cmd == XFS_IOC_FD_TO_HANDLE)
132 		fdput(f);
133 	else
134 		path_put(&path);
135 	return error;
136 }
137 
138 /*
139  * No need to do permission checks on the various pathname components
140  * as the handle operations are privileged.
141  */
142 STATIC int
143 xfs_handle_acceptable(
144 	void			*context,
145 	struct dentry		*dentry)
146 {
147 	return 1;
148 }
149 
150 /*
151  * Convert userspace handle data into a dentry.
152  */
153 struct dentry *
154 xfs_handle_to_dentry(
155 	struct file		*parfilp,
156 	void __user		*uhandle,
157 	u32			hlen)
158 {
159 	xfs_handle_t		handle;
160 	struct xfs_fid64	fid;
161 
162 	/*
163 	 * Only allow handle opens under a directory.
164 	 */
165 	if (!S_ISDIR(file_inode(parfilp)->i_mode))
166 		return ERR_PTR(-ENOTDIR);
167 
168 	if (hlen != sizeof(xfs_handle_t))
169 		return ERR_PTR(-EINVAL);
170 	if (copy_from_user(&handle, uhandle, hlen))
171 		return ERR_PTR(-EFAULT);
172 	if (handle.ha_fid.fid_len !=
173 	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
174 		return ERR_PTR(-EINVAL);
175 
176 	memset(&fid, 0, sizeof(struct fid));
177 	fid.ino = handle.ha_fid.fid_ino;
178 	fid.gen = handle.ha_fid.fid_gen;
179 
180 	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
181 			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
182 			xfs_handle_acceptable, NULL);
183 }
184 
185 STATIC struct dentry *
186 xfs_handlereq_to_dentry(
187 	struct file		*parfilp,
188 	xfs_fsop_handlereq_t	*hreq)
189 {
190 	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
191 }
192 
193 int
194 xfs_open_by_handle(
195 	struct file		*parfilp,
196 	xfs_fsop_handlereq_t	*hreq)
197 {
198 	const struct cred	*cred = current_cred();
199 	int			error;
200 	int			fd;
201 	int			permflag;
202 	struct file		*filp;
203 	struct inode		*inode;
204 	struct dentry		*dentry;
205 	fmode_t			fmode;
206 	struct path		path;
207 
208 	if (!capable(CAP_SYS_ADMIN))
209 		return -EPERM;
210 
211 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
212 	if (IS_ERR(dentry))
213 		return PTR_ERR(dentry);
214 	inode = d_inode(dentry);
215 
216 	/* Restrict xfs_open_by_handle to directories & regular files. */
217 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
218 		error = -EPERM;
219 		goto out_dput;
220 	}
221 
222 #if BITS_PER_LONG != 32
223 	hreq->oflags |= O_LARGEFILE;
224 #endif
225 
226 	permflag = hreq->oflags;
227 	fmode = OPEN_FMODE(permflag);
228 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
229 	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
230 		error = -EPERM;
231 		goto out_dput;
232 	}
233 
234 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
235 		error = -EACCES;
236 		goto out_dput;
237 	}
238 
239 	/* Can't write directories. */
240 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
241 		error = -EISDIR;
242 		goto out_dput;
243 	}
244 
245 	fd = get_unused_fd_flags(0);
246 	if (fd < 0) {
247 		error = fd;
248 		goto out_dput;
249 	}
250 
251 	path.mnt = parfilp->f_path.mnt;
252 	path.dentry = dentry;
253 	filp = dentry_open(&path, hreq->oflags, cred);
254 	dput(dentry);
255 	if (IS_ERR(filp)) {
256 		put_unused_fd(fd);
257 		return PTR_ERR(filp);
258 	}
259 
260 	if (S_ISREG(inode->i_mode)) {
261 		filp->f_flags |= O_NOATIME;
262 		filp->f_mode |= FMODE_NOCMTIME;
263 	}
264 
265 	fd_install(fd, filp);
266 	return fd;
267 
268  out_dput:
269 	dput(dentry);
270 	return error;
271 }
272 
273 int
274 xfs_readlink_by_handle(
275 	struct file		*parfilp,
276 	xfs_fsop_handlereq_t	*hreq)
277 {
278 	struct dentry		*dentry;
279 	__u32			olen;
280 	int			error;
281 
282 	if (!capable(CAP_SYS_ADMIN))
283 		return -EPERM;
284 
285 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
286 	if (IS_ERR(dentry))
287 		return PTR_ERR(dentry);
288 
289 	/* Restrict this handle operation to symlinks only. */
290 	if (!d_inode(dentry)->i_op->readlink) {
291 		error = -EINVAL;
292 		goto out_dput;
293 	}
294 
295 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
296 		error = -EFAULT;
297 		goto out_dput;
298 	}
299 
300 	error = d_inode(dentry)->i_op->readlink(dentry, hreq->ohandle, olen);
301 
302  out_dput:
303 	dput(dentry);
304 	return error;
305 }
306 
307 int
308 xfs_set_dmattrs(
309 	xfs_inode_t     *ip,
310 	u_int		evmask,
311 	u_int16_t	state)
312 {
313 	xfs_mount_t	*mp = ip->i_mount;
314 	xfs_trans_t	*tp;
315 	int		error;
316 
317 	if (!capable(CAP_SYS_ADMIN))
318 		return -EPERM;
319 
320 	if (XFS_FORCED_SHUTDOWN(mp))
321 		return -EIO;
322 
323 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
324 	if (error)
325 		return error;
326 
327 	xfs_ilock(ip, XFS_ILOCK_EXCL);
328 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
329 
330 	ip->i_d.di_dmevmask = evmask;
331 	ip->i_d.di_dmstate  = state;
332 
333 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
334 	error = xfs_trans_commit(tp);
335 
336 	return error;
337 }
338 
339 STATIC int
340 xfs_fssetdm_by_handle(
341 	struct file		*parfilp,
342 	void			__user *arg)
343 {
344 	int			error;
345 	struct fsdmidata	fsd;
346 	xfs_fsop_setdm_handlereq_t dmhreq;
347 	struct dentry		*dentry;
348 
349 	if (!capable(CAP_MKNOD))
350 		return -EPERM;
351 	if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
352 		return -EFAULT;
353 
354 	error = mnt_want_write_file(parfilp);
355 	if (error)
356 		return error;
357 
358 	dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
359 	if (IS_ERR(dentry)) {
360 		mnt_drop_write_file(parfilp);
361 		return PTR_ERR(dentry);
362 	}
363 
364 	if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
365 		error = -EPERM;
366 		goto out;
367 	}
368 
369 	if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
370 		error = -EFAULT;
371 		goto out;
372 	}
373 
374 	error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
375 				 fsd.fsd_dmstate);
376 
377  out:
378 	mnt_drop_write_file(parfilp);
379 	dput(dentry);
380 	return error;
381 }
382 
383 STATIC int
384 xfs_attrlist_by_handle(
385 	struct file		*parfilp,
386 	void			__user *arg)
387 {
388 	int			error = -ENOMEM;
389 	attrlist_cursor_kern_t	*cursor;
390 	xfs_fsop_attrlist_handlereq_t al_hreq;
391 	struct dentry		*dentry;
392 	char			*kbuf;
393 
394 	if (!capable(CAP_SYS_ADMIN))
395 		return -EPERM;
396 	if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
397 		return -EFAULT;
398 	if (al_hreq.buflen < sizeof(struct attrlist) ||
399 	    al_hreq.buflen > XFS_XATTR_LIST_MAX)
400 		return -EINVAL;
401 
402 	/*
403 	 * Reject flags, only allow namespaces.
404 	 */
405 	if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
406 		return -EINVAL;
407 
408 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
409 	if (IS_ERR(dentry))
410 		return PTR_ERR(dentry);
411 
412 	kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
413 	if (!kbuf)
414 		goto out_dput;
415 
416 	cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
417 	error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
418 					al_hreq.flags, cursor);
419 	if (error)
420 		goto out_kfree;
421 
422 	if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
423 		error = -EFAULT;
424 
425 out_kfree:
426 	kmem_free(kbuf);
427 out_dput:
428 	dput(dentry);
429 	return error;
430 }
431 
432 int
433 xfs_attrmulti_attr_get(
434 	struct inode		*inode,
435 	unsigned char		*name,
436 	unsigned char		__user *ubuf,
437 	__uint32_t		*len,
438 	__uint32_t		flags)
439 {
440 	unsigned char		*kbuf;
441 	int			error = -EFAULT;
442 
443 	if (*len > XFS_XATTR_SIZE_MAX)
444 		return -EINVAL;
445 	kbuf = kmem_zalloc_large(*len, KM_SLEEP);
446 	if (!kbuf)
447 		return -ENOMEM;
448 
449 	error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
450 	if (error)
451 		goto out_kfree;
452 
453 	if (copy_to_user(ubuf, kbuf, *len))
454 		error = -EFAULT;
455 
456 out_kfree:
457 	kmem_free(kbuf);
458 	return error;
459 }
460 
461 int
462 xfs_attrmulti_attr_set(
463 	struct inode		*inode,
464 	unsigned char		*name,
465 	const unsigned char	__user *ubuf,
466 	__uint32_t		len,
467 	__uint32_t		flags)
468 {
469 	unsigned char		*kbuf;
470 	int			error;
471 
472 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
473 		return -EPERM;
474 	if (len > XFS_XATTR_SIZE_MAX)
475 		return -EINVAL;
476 
477 	kbuf = memdup_user(ubuf, len);
478 	if (IS_ERR(kbuf))
479 		return PTR_ERR(kbuf);
480 
481 	error = xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
482 	if (!error)
483 		xfs_forget_acl(inode, name, flags);
484 	kfree(kbuf);
485 	return error;
486 }
487 
488 int
489 xfs_attrmulti_attr_remove(
490 	struct inode		*inode,
491 	unsigned char		*name,
492 	__uint32_t		flags)
493 {
494 	int			error;
495 
496 	if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
497 		return -EPERM;
498 	error = xfs_attr_remove(XFS_I(inode), name, flags);
499 	if (!error)
500 		xfs_forget_acl(inode, name, flags);
501 	return error;
502 }
503 
504 STATIC int
505 xfs_attrmulti_by_handle(
506 	struct file		*parfilp,
507 	void			__user *arg)
508 {
509 	int			error;
510 	xfs_attr_multiop_t	*ops;
511 	xfs_fsop_attrmulti_handlereq_t am_hreq;
512 	struct dentry		*dentry;
513 	unsigned int		i, size;
514 	unsigned char		*attr_name;
515 
516 	if (!capable(CAP_SYS_ADMIN))
517 		return -EPERM;
518 	if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
519 		return -EFAULT;
520 
521 	/* overflow check */
522 	if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
523 		return -E2BIG;
524 
525 	dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
526 	if (IS_ERR(dentry))
527 		return PTR_ERR(dentry);
528 
529 	error = -E2BIG;
530 	size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
531 	if (!size || size > 16 * PAGE_SIZE)
532 		goto out_dput;
533 
534 	ops = memdup_user(am_hreq.ops, size);
535 	if (IS_ERR(ops)) {
536 		error = PTR_ERR(ops);
537 		goto out_dput;
538 	}
539 
540 	error = -ENOMEM;
541 	attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
542 	if (!attr_name)
543 		goto out_kfree_ops;
544 
545 	error = 0;
546 	for (i = 0; i < am_hreq.opcount; i++) {
547 		ops[i].am_error = strncpy_from_user((char *)attr_name,
548 				ops[i].am_attrname, MAXNAMELEN);
549 		if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
550 			error = -ERANGE;
551 		if (ops[i].am_error < 0)
552 			break;
553 
554 		switch (ops[i].am_opcode) {
555 		case ATTR_OP_GET:
556 			ops[i].am_error = xfs_attrmulti_attr_get(
557 					d_inode(dentry), attr_name,
558 					ops[i].am_attrvalue, &ops[i].am_length,
559 					ops[i].am_flags);
560 			break;
561 		case ATTR_OP_SET:
562 			ops[i].am_error = mnt_want_write_file(parfilp);
563 			if (ops[i].am_error)
564 				break;
565 			ops[i].am_error = xfs_attrmulti_attr_set(
566 					d_inode(dentry), attr_name,
567 					ops[i].am_attrvalue, ops[i].am_length,
568 					ops[i].am_flags);
569 			mnt_drop_write_file(parfilp);
570 			break;
571 		case ATTR_OP_REMOVE:
572 			ops[i].am_error = mnt_want_write_file(parfilp);
573 			if (ops[i].am_error)
574 				break;
575 			ops[i].am_error = xfs_attrmulti_attr_remove(
576 					d_inode(dentry), attr_name,
577 					ops[i].am_flags);
578 			mnt_drop_write_file(parfilp);
579 			break;
580 		default:
581 			ops[i].am_error = -EINVAL;
582 		}
583 	}
584 
585 	if (copy_to_user(am_hreq.ops, ops, size))
586 		error = -EFAULT;
587 
588 	kfree(attr_name);
589  out_kfree_ops:
590 	kfree(ops);
591  out_dput:
592 	dput(dentry);
593 	return error;
594 }
595 
596 int
597 xfs_ioc_space(
598 	struct xfs_inode	*ip,
599 	struct inode		*inode,
600 	struct file		*filp,
601 	int			ioflags,
602 	unsigned int		cmd,
603 	xfs_flock64_t		*bf)
604 {
605 	struct iattr		iattr;
606 	enum xfs_prealloc_flags	flags = 0;
607 	uint			iolock = XFS_IOLOCK_EXCL;
608 	int			error;
609 
610 	/*
611 	 * Only allow the sys admin to reserve space unless
612 	 * unwritten extents are enabled.
613 	 */
614 	if (!xfs_sb_version_hasextflgbit(&ip->i_mount->m_sb) &&
615 	    !capable(CAP_SYS_ADMIN))
616 		return -EPERM;
617 
618 	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
619 		return -EPERM;
620 
621 	if (!(filp->f_mode & FMODE_WRITE))
622 		return -EBADF;
623 
624 	if (!S_ISREG(inode->i_mode))
625 		return -EINVAL;
626 
627 	if (filp->f_flags & O_DSYNC)
628 		flags |= XFS_PREALLOC_SYNC;
629 	if (ioflags & XFS_IO_INVIS)
630 		flags |= XFS_PREALLOC_INVISIBLE;
631 
632 	error = mnt_want_write_file(filp);
633 	if (error)
634 		return error;
635 
636 	xfs_ilock(ip, iolock);
637 	error = xfs_break_layouts(inode, &iolock, false);
638 	if (error)
639 		goto out_unlock;
640 
641 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
642 	iolock |= XFS_MMAPLOCK_EXCL;
643 
644 	switch (bf->l_whence) {
645 	case 0: /*SEEK_SET*/
646 		break;
647 	case 1: /*SEEK_CUR*/
648 		bf->l_start += filp->f_pos;
649 		break;
650 	case 2: /*SEEK_END*/
651 		bf->l_start += XFS_ISIZE(ip);
652 		break;
653 	default:
654 		error = -EINVAL;
655 		goto out_unlock;
656 	}
657 
658 	/*
659 	 * length of <= 0 for resv/unresv/zero is invalid.  length for
660 	 * alloc/free is ignored completely and we have no idea what userspace
661 	 * might have set it to, so set it to zero to allow range
662 	 * checks to pass.
663 	 */
664 	switch (cmd) {
665 	case XFS_IOC_ZERO_RANGE:
666 	case XFS_IOC_RESVSP:
667 	case XFS_IOC_RESVSP64:
668 	case XFS_IOC_UNRESVSP:
669 	case XFS_IOC_UNRESVSP64:
670 		if (bf->l_len <= 0) {
671 			error = -EINVAL;
672 			goto out_unlock;
673 		}
674 		break;
675 	default:
676 		bf->l_len = 0;
677 		break;
678 	}
679 
680 	if (bf->l_start < 0 ||
681 	    bf->l_start > inode->i_sb->s_maxbytes ||
682 	    bf->l_start + bf->l_len < 0 ||
683 	    bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
684 		error = -EINVAL;
685 		goto out_unlock;
686 	}
687 
688 	switch (cmd) {
689 	case XFS_IOC_ZERO_RANGE:
690 		flags |= XFS_PREALLOC_SET;
691 		error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
692 		break;
693 	case XFS_IOC_RESVSP:
694 	case XFS_IOC_RESVSP64:
695 		flags |= XFS_PREALLOC_SET;
696 		error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
697 						XFS_BMAPI_PREALLOC);
698 		break;
699 	case XFS_IOC_UNRESVSP:
700 	case XFS_IOC_UNRESVSP64:
701 		error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
702 		break;
703 	case XFS_IOC_ALLOCSP:
704 	case XFS_IOC_ALLOCSP64:
705 	case XFS_IOC_FREESP:
706 	case XFS_IOC_FREESP64:
707 		flags |= XFS_PREALLOC_CLEAR;
708 		if (bf->l_start > XFS_ISIZE(ip)) {
709 			error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
710 					bf->l_start - XFS_ISIZE(ip), 0);
711 			if (error)
712 				goto out_unlock;
713 		}
714 
715 		iattr.ia_valid = ATTR_SIZE;
716 		iattr.ia_size = bf->l_start;
717 
718 		error = xfs_setattr_size(ip, &iattr);
719 		break;
720 	default:
721 		ASSERT(0);
722 		error = -EINVAL;
723 	}
724 
725 	if (error)
726 		goto out_unlock;
727 
728 	error = xfs_update_prealloc_flags(ip, flags);
729 
730 out_unlock:
731 	xfs_iunlock(ip, iolock);
732 	mnt_drop_write_file(filp);
733 	return error;
734 }
735 
736 STATIC int
737 xfs_ioc_bulkstat(
738 	xfs_mount_t		*mp,
739 	unsigned int		cmd,
740 	void			__user *arg)
741 {
742 	xfs_fsop_bulkreq_t	bulkreq;
743 	int			count;	/* # of records returned */
744 	xfs_ino_t		inlast;	/* last inode number */
745 	int			done;
746 	int			error;
747 
748 	/* done = 1 if there are more stats to get and if bulkstat */
749 	/* should be called again (unused here, but used in dmapi) */
750 
751 	if (!capable(CAP_SYS_ADMIN))
752 		return -EPERM;
753 
754 	if (XFS_FORCED_SHUTDOWN(mp))
755 		return -EIO;
756 
757 	if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
758 		return -EFAULT;
759 
760 	if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
761 		return -EFAULT;
762 
763 	if ((count = bulkreq.icount) <= 0)
764 		return -EINVAL;
765 
766 	if (bulkreq.ubuffer == NULL)
767 		return -EINVAL;
768 
769 	if (cmd == XFS_IOC_FSINUMBERS)
770 		error = xfs_inumbers(mp, &inlast, &count,
771 					bulkreq.ubuffer, xfs_inumbers_fmt);
772 	else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
773 		error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
774 					sizeof(xfs_bstat_t), NULL, &done);
775 	else	/* XFS_IOC_FSBULKSTAT */
776 		error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
777 				     sizeof(xfs_bstat_t), bulkreq.ubuffer,
778 				     &done);
779 
780 	if (error)
781 		return error;
782 
783 	if (bulkreq.ocount != NULL) {
784 		if (copy_to_user(bulkreq.lastip, &inlast,
785 						sizeof(xfs_ino_t)))
786 			return -EFAULT;
787 
788 		if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
789 			return -EFAULT;
790 	}
791 
792 	return 0;
793 }
794 
795 STATIC int
796 xfs_ioc_fsgeometry_v1(
797 	xfs_mount_t		*mp,
798 	void			__user *arg)
799 {
800 	xfs_fsop_geom_t         fsgeo;
801 	int			error;
802 
803 	error = xfs_fs_geometry(mp, &fsgeo, 3);
804 	if (error)
805 		return error;
806 
807 	/*
808 	 * Caller should have passed an argument of type
809 	 * xfs_fsop_geom_v1_t.  This is a proper subset of the
810 	 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
811 	 */
812 	if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
813 		return -EFAULT;
814 	return 0;
815 }
816 
817 STATIC int
818 xfs_ioc_fsgeometry(
819 	xfs_mount_t		*mp,
820 	void			__user *arg)
821 {
822 	xfs_fsop_geom_t		fsgeo;
823 	int			error;
824 
825 	error = xfs_fs_geometry(mp, &fsgeo, 4);
826 	if (error)
827 		return error;
828 
829 	if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
830 		return -EFAULT;
831 	return 0;
832 }
833 
834 /*
835  * Linux extended inode flags interface.
836  */
837 
838 STATIC unsigned int
839 xfs_merge_ioc_xflags(
840 	unsigned int	flags,
841 	unsigned int	start)
842 {
843 	unsigned int	xflags = start;
844 
845 	if (flags & FS_IMMUTABLE_FL)
846 		xflags |= FS_XFLAG_IMMUTABLE;
847 	else
848 		xflags &= ~FS_XFLAG_IMMUTABLE;
849 	if (flags & FS_APPEND_FL)
850 		xflags |= FS_XFLAG_APPEND;
851 	else
852 		xflags &= ~FS_XFLAG_APPEND;
853 	if (flags & FS_SYNC_FL)
854 		xflags |= FS_XFLAG_SYNC;
855 	else
856 		xflags &= ~FS_XFLAG_SYNC;
857 	if (flags & FS_NOATIME_FL)
858 		xflags |= FS_XFLAG_NOATIME;
859 	else
860 		xflags &= ~FS_XFLAG_NOATIME;
861 	if (flags & FS_NODUMP_FL)
862 		xflags |= FS_XFLAG_NODUMP;
863 	else
864 		xflags &= ~FS_XFLAG_NODUMP;
865 
866 	return xflags;
867 }
868 
869 STATIC unsigned int
870 xfs_di2lxflags(
871 	__uint16_t	di_flags)
872 {
873 	unsigned int	flags = 0;
874 
875 	if (di_flags & XFS_DIFLAG_IMMUTABLE)
876 		flags |= FS_IMMUTABLE_FL;
877 	if (di_flags & XFS_DIFLAG_APPEND)
878 		flags |= FS_APPEND_FL;
879 	if (di_flags & XFS_DIFLAG_SYNC)
880 		flags |= FS_SYNC_FL;
881 	if (di_flags & XFS_DIFLAG_NOATIME)
882 		flags |= FS_NOATIME_FL;
883 	if (di_flags & XFS_DIFLAG_NODUMP)
884 		flags |= FS_NODUMP_FL;
885 	return flags;
886 }
887 
888 STATIC int
889 xfs_ioc_fsgetxattr(
890 	xfs_inode_t		*ip,
891 	int			attr,
892 	void			__user *arg)
893 {
894 	struct fsxattr		fa;
895 
896 	memset(&fa, 0, sizeof(struct fsxattr));
897 
898 	xfs_ilock(ip, XFS_ILOCK_SHARED);
899 	fa.fsx_xflags = xfs_ip2xflags(ip);
900 	fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
901 	fa.fsx_projid = xfs_get_projid(ip);
902 
903 	if (attr) {
904 		if (ip->i_afp) {
905 			if (ip->i_afp->if_flags & XFS_IFEXTENTS)
906 				fa.fsx_nextents = ip->i_afp->if_bytes /
907 							sizeof(xfs_bmbt_rec_t);
908 			else
909 				fa.fsx_nextents = ip->i_d.di_anextents;
910 		} else
911 			fa.fsx_nextents = 0;
912 	} else {
913 		if (ip->i_df.if_flags & XFS_IFEXTENTS)
914 			fa.fsx_nextents = ip->i_df.if_bytes /
915 						sizeof(xfs_bmbt_rec_t);
916 		else
917 			fa.fsx_nextents = ip->i_d.di_nextents;
918 	}
919 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
920 
921 	if (copy_to_user(arg, &fa, sizeof(fa)))
922 		return -EFAULT;
923 	return 0;
924 }
925 
926 STATIC void
927 xfs_set_diflags(
928 	struct xfs_inode	*ip,
929 	unsigned int		xflags)
930 {
931 	unsigned int		di_flags;
932 	uint64_t		di_flags2;
933 
934 	/* can't set PREALLOC this way, just preserve it */
935 	di_flags = (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
936 	if (xflags & FS_XFLAG_IMMUTABLE)
937 		di_flags |= XFS_DIFLAG_IMMUTABLE;
938 	if (xflags & FS_XFLAG_APPEND)
939 		di_flags |= XFS_DIFLAG_APPEND;
940 	if (xflags & FS_XFLAG_SYNC)
941 		di_flags |= XFS_DIFLAG_SYNC;
942 	if (xflags & FS_XFLAG_NOATIME)
943 		di_flags |= XFS_DIFLAG_NOATIME;
944 	if (xflags & FS_XFLAG_NODUMP)
945 		di_flags |= XFS_DIFLAG_NODUMP;
946 	if (xflags & FS_XFLAG_NODEFRAG)
947 		di_flags |= XFS_DIFLAG_NODEFRAG;
948 	if (xflags & FS_XFLAG_FILESTREAM)
949 		di_flags |= XFS_DIFLAG_FILESTREAM;
950 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
951 		if (xflags & FS_XFLAG_RTINHERIT)
952 			di_flags |= XFS_DIFLAG_RTINHERIT;
953 		if (xflags & FS_XFLAG_NOSYMLINKS)
954 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
955 		if (xflags & FS_XFLAG_EXTSZINHERIT)
956 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
957 		if (xflags & FS_XFLAG_PROJINHERIT)
958 			di_flags |= XFS_DIFLAG_PROJINHERIT;
959 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
960 		if (xflags & FS_XFLAG_REALTIME)
961 			di_flags |= XFS_DIFLAG_REALTIME;
962 		if (xflags & FS_XFLAG_EXTSIZE)
963 			di_flags |= XFS_DIFLAG_EXTSIZE;
964 	}
965 	ip->i_d.di_flags = di_flags;
966 
967 	/* diflags2 only valid for v3 inodes. */
968 	if (ip->i_d.di_version < 3)
969 		return;
970 
971 	di_flags2 = 0;
972 	if (xflags & FS_XFLAG_DAX)
973 		di_flags2 |= XFS_DIFLAG2_DAX;
974 
975 	ip->i_d.di_flags2 = di_flags2;
976 
977 }
978 
979 STATIC void
980 xfs_diflags_to_linux(
981 	struct xfs_inode	*ip)
982 {
983 	struct inode		*inode = VFS_I(ip);
984 	unsigned int		xflags = xfs_ip2xflags(ip);
985 
986 	if (xflags & FS_XFLAG_IMMUTABLE)
987 		inode->i_flags |= S_IMMUTABLE;
988 	else
989 		inode->i_flags &= ~S_IMMUTABLE;
990 	if (xflags & FS_XFLAG_APPEND)
991 		inode->i_flags |= S_APPEND;
992 	else
993 		inode->i_flags &= ~S_APPEND;
994 	if (xflags & FS_XFLAG_SYNC)
995 		inode->i_flags |= S_SYNC;
996 	else
997 		inode->i_flags &= ~S_SYNC;
998 	if (xflags & FS_XFLAG_NOATIME)
999 		inode->i_flags |= S_NOATIME;
1000 	else
1001 		inode->i_flags &= ~S_NOATIME;
1002 	if (xflags & FS_XFLAG_DAX)
1003 		inode->i_flags |= S_DAX;
1004 	else
1005 		inode->i_flags &= ~S_DAX;
1006 
1007 }
1008 
1009 static int
1010 xfs_ioctl_setattr_xflags(
1011 	struct xfs_trans	*tp,
1012 	struct xfs_inode	*ip,
1013 	struct fsxattr		*fa)
1014 {
1015 	struct xfs_mount	*mp = ip->i_mount;
1016 
1017 	/* Can't change realtime flag if any extents are allocated. */
1018 	if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1019 	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1020 		return -EINVAL;
1021 
1022 	/* If realtime flag is set then must have realtime device */
1023 	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1024 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1025 		    (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1026 			return -EINVAL;
1027 	}
1028 
1029 	/*
1030 	 * Can't modify an immutable/append-only file unless
1031 	 * we have appropriate permission.
1032 	 */
1033 	if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1034 	     (fa->fsx_xflags & (FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))) &&
1035 	    !capable(CAP_LINUX_IMMUTABLE))
1036 		return -EPERM;
1037 
1038 	xfs_set_diflags(ip, fa->fsx_xflags);
1039 	xfs_diflags_to_linux(ip);
1040 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1041 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1042 	XFS_STATS_INC(mp, xs_ig_attrchg);
1043 	return 0;
1044 }
1045 
1046 /*
1047  * If we are changing DAX flags, we have to ensure the file is clean and any
1048  * cached objects in the address space are invalidated and removed. This
1049  * requires us to lock out other IO and page faults similar to a truncate
1050  * operation. The locks need to be held until the transaction has been committed
1051  * so that the cache invalidation is atomic with respect to the DAX flag
1052  * manipulation.
1053  */
1054 static int
1055 xfs_ioctl_setattr_dax_invalidate(
1056 	struct xfs_inode	*ip,
1057 	struct fsxattr		*fa,
1058 	int			*join_flags)
1059 {
1060 	struct inode		*inode = VFS_I(ip);
1061 	int			error;
1062 
1063 	*join_flags = 0;
1064 
1065 	/*
1066 	 * It is only valid to set the DAX flag on regular files and
1067 	 * directories on filesystems where the block size is equal to the page
1068 	 * size. On directories it serves as an inherit hint.
1069 	 */
1070 	if (fa->fsx_xflags & FS_XFLAG_DAX) {
1071 		if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1072 			return -EINVAL;
1073 		if (ip->i_mount->m_sb.sb_blocksize != PAGE_SIZE)
1074 			return -EINVAL;
1075 	}
1076 
1077 	/* If the DAX state is not changing, we have nothing to do here. */
1078 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
1079 		return 0;
1080 	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
1081 		return 0;
1082 
1083 	/* lock, flush and invalidate mapping in preparation for flag change */
1084 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1085 	error = filemap_write_and_wait(inode->i_mapping);
1086 	if (error)
1087 		goto out_unlock;
1088 	error = invalidate_inode_pages2(inode->i_mapping);
1089 	if (error)
1090 		goto out_unlock;
1091 
1092 	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
1093 	return 0;
1094 
1095 out_unlock:
1096 	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1097 	return error;
1098 
1099 }
1100 
1101 /*
1102  * Set up the transaction structure for the setattr operation, checking that we
1103  * have permission to do so. On success, return a clean transaction and the
1104  * inode locked exclusively ready for further operation specific checks. On
1105  * failure, return an error without modifying or locking the inode.
1106  *
1107  * The inode might already be IO locked on call. If this is the case, it is
1108  * indicated in @join_flags and we take full responsibility for ensuring they
1109  * are unlocked from now on. Hence if we have an error here, we still have to
1110  * unlock them. Otherwise, once they are joined to the transaction, they will
1111  * be unlocked on commit/cancel.
1112  */
1113 static struct xfs_trans *
1114 xfs_ioctl_setattr_get_trans(
1115 	struct xfs_inode	*ip,
1116 	int			join_flags)
1117 {
1118 	struct xfs_mount	*mp = ip->i_mount;
1119 	struct xfs_trans	*tp;
1120 	int			error = -EROFS;
1121 
1122 	if (mp->m_flags & XFS_MOUNT_RDONLY)
1123 		goto out_unlock;
1124 	error = -EIO;
1125 	if (XFS_FORCED_SHUTDOWN(mp))
1126 		goto out_unlock;
1127 
1128 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
1129 	if (error)
1130 		return ERR_PTR(error);
1131 
1132 	xfs_ilock(ip, XFS_ILOCK_EXCL);
1133 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
1134 	join_flags = 0;
1135 
1136 	/*
1137 	 * CAP_FOWNER overrides the following restrictions:
1138 	 *
1139 	 * The user ID of the calling process must be equal to the file owner
1140 	 * ID, except in cases where the CAP_FSETID capability is applicable.
1141 	 */
1142 	if (!inode_owner_or_capable(VFS_I(ip))) {
1143 		error = -EPERM;
1144 		goto out_cancel;
1145 	}
1146 
1147 	if (mp->m_flags & XFS_MOUNT_WSYNC)
1148 		xfs_trans_set_sync(tp);
1149 
1150 	return tp;
1151 
1152 out_cancel:
1153 	xfs_trans_cancel(tp);
1154 out_unlock:
1155 	if (join_flags)
1156 		xfs_iunlock(ip, join_flags);
1157 	return ERR_PTR(error);
1158 }
1159 
1160 /*
1161  * extent size hint validation is somewhat cumbersome. Rules are:
1162  *
1163  * 1. extent size hint is only valid for directories and regular files
1164  * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1165  * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1166  * 4. can only be changed on regular files if no extents are allocated
1167  * 5. can be changed on directories at any time
1168  * 6. extsize hint of 0 turns off hints, clears inode flags.
1169  * 7. Extent size must be a multiple of the appropriate block size.
1170  * 8. for non-realtime files, the extent size hint must be limited
1171  *    to half the AG size to avoid alignment extending the extent beyond the
1172  *    limits of the AG.
1173  */
1174 static int
1175 xfs_ioctl_setattr_check_extsize(
1176 	struct xfs_inode	*ip,
1177 	struct fsxattr		*fa)
1178 {
1179 	struct xfs_mount	*mp = ip->i_mount;
1180 
1181 	if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(VFS_I(ip)->i_mode))
1182 		return -EINVAL;
1183 
1184 	if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
1185 	    !S_ISDIR(VFS_I(ip)->i_mode))
1186 		return -EINVAL;
1187 
1188 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_d.di_nextents &&
1189 	    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1190 		return -EINVAL;
1191 
1192 	if (fa->fsx_extsize != 0) {
1193 		xfs_extlen_t    size;
1194 		xfs_fsblock_t   extsize_fsb;
1195 
1196 		extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1197 		if (extsize_fsb > MAXEXTLEN)
1198 			return -EINVAL;
1199 
1200 		if (XFS_IS_REALTIME_INODE(ip) ||
1201 		    (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1202 			size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1203 		} else {
1204 			size = mp->m_sb.sb_blocksize;
1205 			if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1206 				return -EINVAL;
1207 		}
1208 
1209 		if (fa->fsx_extsize % size)
1210 			return -EINVAL;
1211 	} else
1212 		fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
1213 
1214 	return 0;
1215 }
1216 
1217 static int
1218 xfs_ioctl_setattr_check_projid(
1219 	struct xfs_inode	*ip,
1220 	struct fsxattr		*fa)
1221 {
1222 	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1223 	if (fa->fsx_projid > (__uint16_t)-1 &&
1224 	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1225 		return -EINVAL;
1226 
1227 	/*
1228 	 * Project Quota ID state is only allowed to change from within the init
1229 	 * namespace. Enforce that restriction only if we are trying to change
1230 	 * the quota ID state. Everything else is allowed in user namespaces.
1231 	 */
1232 	if (current_user_ns() == &init_user_ns)
1233 		return 0;
1234 
1235 	if (xfs_get_projid(ip) != fa->fsx_projid)
1236 		return -EINVAL;
1237 	if ((fa->fsx_xflags & FS_XFLAG_PROJINHERIT) !=
1238 	    (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1239 		return -EINVAL;
1240 
1241 	return 0;
1242 }
1243 
1244 STATIC int
1245 xfs_ioctl_setattr(
1246 	xfs_inode_t		*ip,
1247 	struct fsxattr		*fa)
1248 {
1249 	struct xfs_mount	*mp = ip->i_mount;
1250 	struct xfs_trans	*tp;
1251 	struct xfs_dquot	*udqp = NULL;
1252 	struct xfs_dquot	*pdqp = NULL;
1253 	struct xfs_dquot	*olddquot = NULL;
1254 	int			code;
1255 	int			join_flags = 0;
1256 
1257 	trace_xfs_ioctl_setattr(ip);
1258 
1259 	code = xfs_ioctl_setattr_check_projid(ip, fa);
1260 	if (code)
1261 		return code;
1262 
1263 	/*
1264 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1265 	 * before we start any other transactions. Trying to do this later
1266 	 * is messy. We don't care to take a readlock to look at the ids
1267 	 * in inode here, because we can't hold it across the trans_reserve.
1268 	 * If the IDs do change before we take the ilock, we're covered
1269 	 * because the i_*dquot fields will get updated anyway.
1270 	 */
1271 	if (XFS_IS_QUOTA_ON(mp)) {
1272 		code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1273 					 ip->i_d.di_gid, fa->fsx_projid,
1274 					 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1275 		if (code)
1276 			return code;
1277 	}
1278 
1279 	/*
1280 	 * Changing DAX config may require inode locking for mapping
1281 	 * invalidation. These need to be held all the way to transaction commit
1282 	 * or cancel time, so need to be passed through to
1283 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1284 	 * appropriately.
1285 	 */
1286 	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
1287 	if (code)
1288 		goto error_free_dquots;
1289 
1290 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1291 	if (IS_ERR(tp)) {
1292 		code = PTR_ERR(tp);
1293 		goto error_free_dquots;
1294 	}
1295 
1296 
1297 	if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1298 	    xfs_get_projid(ip) != fa->fsx_projid) {
1299 		code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1300 				capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
1301 		if (code)	/* out of quota */
1302 			goto error_trans_cancel;
1303 	}
1304 
1305 	code = xfs_ioctl_setattr_check_extsize(ip, fa);
1306 	if (code)
1307 		goto error_trans_cancel;
1308 
1309 	code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1310 	if (code)
1311 		goto error_trans_cancel;
1312 
1313 	/*
1314 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1315 	 * overrides the following restrictions:
1316 	 *
1317 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1318 	 * successful return from chown()
1319 	 */
1320 
1321 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1322 	    !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1323 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1324 
1325 	/* Change the ownerships and register project quota modifications */
1326 	if (xfs_get_projid(ip) != fa->fsx_projid) {
1327 		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1328 			olddquot = xfs_qm_vop_chown(tp, ip,
1329 						&ip->i_pdquot, pdqp);
1330 		}
1331 		ASSERT(ip->i_d.di_version > 1);
1332 		xfs_set_projid(ip, fa->fsx_projid);
1333 	}
1334 
1335 	/*
1336 	 * Only set the extent size hint if we've already determined that the
1337 	 * extent size hint should be set on the inode. If no extent size flags
1338 	 * are set on the inode then unconditionally clear the extent size hint.
1339 	 */
1340 	if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1341 		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1342 	else
1343 		ip->i_d.di_extsize = 0;
1344 
1345 	code = xfs_trans_commit(tp);
1346 
1347 	/*
1348 	 * Release any dquot(s) the inode had kept before chown.
1349 	 */
1350 	xfs_qm_dqrele(olddquot);
1351 	xfs_qm_dqrele(udqp);
1352 	xfs_qm_dqrele(pdqp);
1353 
1354 	return code;
1355 
1356 error_trans_cancel:
1357 	xfs_trans_cancel(tp);
1358 error_free_dquots:
1359 	xfs_qm_dqrele(udqp);
1360 	xfs_qm_dqrele(pdqp);
1361 	return code;
1362 }
1363 
1364 STATIC int
1365 xfs_ioc_fssetxattr(
1366 	xfs_inode_t		*ip,
1367 	struct file		*filp,
1368 	void			__user *arg)
1369 {
1370 	struct fsxattr		fa;
1371 	int error;
1372 
1373 	if (copy_from_user(&fa, arg, sizeof(fa)))
1374 		return -EFAULT;
1375 
1376 	error = mnt_want_write_file(filp);
1377 	if (error)
1378 		return error;
1379 	error = xfs_ioctl_setattr(ip, &fa);
1380 	mnt_drop_write_file(filp);
1381 	return error;
1382 }
1383 
1384 STATIC int
1385 xfs_ioc_getxflags(
1386 	xfs_inode_t		*ip,
1387 	void			__user *arg)
1388 {
1389 	unsigned int		flags;
1390 
1391 	flags = xfs_di2lxflags(ip->i_d.di_flags);
1392 	if (copy_to_user(arg, &flags, sizeof(flags)))
1393 		return -EFAULT;
1394 	return 0;
1395 }
1396 
1397 STATIC int
1398 xfs_ioc_setxflags(
1399 	struct xfs_inode	*ip,
1400 	struct file		*filp,
1401 	void			__user *arg)
1402 {
1403 	struct xfs_trans	*tp;
1404 	struct fsxattr		fa;
1405 	unsigned int		flags;
1406 	int			join_flags = 0;
1407 	int			error;
1408 
1409 	if (copy_from_user(&flags, arg, sizeof(flags)))
1410 		return -EFAULT;
1411 
1412 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1413 		      FS_NOATIME_FL | FS_NODUMP_FL | \
1414 		      FS_SYNC_FL))
1415 		return -EOPNOTSUPP;
1416 
1417 	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1418 
1419 	error = mnt_want_write_file(filp);
1420 	if (error)
1421 		return error;
1422 
1423 	/*
1424 	 * Changing DAX config may require inode locking for mapping
1425 	 * invalidation. These need to be held all the way to transaction commit
1426 	 * or cancel time, so need to be passed through to
1427 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1428 	 * appropriately.
1429 	 */
1430 	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
1431 	if (error)
1432 		goto out_drop_write;
1433 
1434 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1435 	if (IS_ERR(tp)) {
1436 		error = PTR_ERR(tp);
1437 		goto out_drop_write;
1438 	}
1439 
1440 	error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1441 	if (error) {
1442 		xfs_trans_cancel(tp);
1443 		goto out_drop_write;
1444 	}
1445 
1446 	error = xfs_trans_commit(tp);
1447 out_drop_write:
1448 	mnt_drop_write_file(filp);
1449 	return error;
1450 }
1451 
1452 STATIC int
1453 xfs_getbmap_format(void **ap, struct getbmapx *bmv, int *full)
1454 {
1455 	struct getbmap __user	*base = (struct getbmap __user *)*ap;
1456 
1457 	/* copy only getbmap portion (not getbmapx) */
1458 	if (copy_to_user(base, bmv, sizeof(struct getbmap)))
1459 		return -EFAULT;
1460 
1461 	*ap += sizeof(struct getbmap);
1462 	return 0;
1463 }
1464 
1465 STATIC int
1466 xfs_ioc_getbmap(
1467 	struct xfs_inode	*ip,
1468 	int			ioflags,
1469 	unsigned int		cmd,
1470 	void			__user *arg)
1471 {
1472 	struct getbmapx		bmx;
1473 	int			error;
1474 
1475 	if (copy_from_user(&bmx, arg, sizeof(struct getbmapx)))
1476 		return -EFAULT;
1477 
1478 	if (bmx.bmv_count < 2)
1479 		return -EINVAL;
1480 
1481 	bmx.bmv_iflags = (cmd == XFS_IOC_GETBMAPA ? BMV_IF_ATTRFORK : 0);
1482 	if (ioflags & XFS_IO_INVIS)
1483 		bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1484 
1485 	error = xfs_getbmap(ip, &bmx, xfs_getbmap_format,
1486 			    (__force struct getbmap *)arg+1);
1487 	if (error)
1488 		return error;
1489 
1490 	/* copy back header - only size of getbmap */
1491 	if (copy_to_user(arg, &bmx, sizeof(struct getbmap)))
1492 		return -EFAULT;
1493 	return 0;
1494 }
1495 
1496 STATIC int
1497 xfs_getbmapx_format(void **ap, struct getbmapx *bmv, int *full)
1498 {
1499 	struct getbmapx __user	*base = (struct getbmapx __user *)*ap;
1500 
1501 	if (copy_to_user(base, bmv, sizeof(struct getbmapx)))
1502 		return -EFAULT;
1503 
1504 	*ap += sizeof(struct getbmapx);
1505 	return 0;
1506 }
1507 
1508 STATIC int
1509 xfs_ioc_getbmapx(
1510 	struct xfs_inode	*ip,
1511 	void			__user *arg)
1512 {
1513 	struct getbmapx		bmx;
1514 	int			error;
1515 
1516 	if (copy_from_user(&bmx, arg, sizeof(bmx)))
1517 		return -EFAULT;
1518 
1519 	if (bmx.bmv_count < 2)
1520 		return -EINVAL;
1521 
1522 	if (bmx.bmv_iflags & (~BMV_IF_VALID))
1523 		return -EINVAL;
1524 
1525 	error = xfs_getbmap(ip, &bmx, xfs_getbmapx_format,
1526 			    (__force struct getbmapx *)arg+1);
1527 	if (error)
1528 		return error;
1529 
1530 	/* copy back header */
1531 	if (copy_to_user(arg, &bmx, sizeof(struct getbmapx)))
1532 		return -EFAULT;
1533 
1534 	return 0;
1535 }
1536 
1537 int
1538 xfs_ioc_swapext(
1539 	xfs_swapext_t	*sxp)
1540 {
1541 	xfs_inode_t     *ip, *tip;
1542 	struct fd	f, tmp;
1543 	int		error = 0;
1544 
1545 	/* Pull information for the target fd */
1546 	f = fdget((int)sxp->sx_fdtarget);
1547 	if (!f.file) {
1548 		error = -EINVAL;
1549 		goto out;
1550 	}
1551 
1552 	if (!(f.file->f_mode & FMODE_WRITE) ||
1553 	    !(f.file->f_mode & FMODE_READ) ||
1554 	    (f.file->f_flags & O_APPEND)) {
1555 		error = -EBADF;
1556 		goto out_put_file;
1557 	}
1558 
1559 	tmp = fdget((int)sxp->sx_fdtmp);
1560 	if (!tmp.file) {
1561 		error = -EINVAL;
1562 		goto out_put_file;
1563 	}
1564 
1565 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1566 	    !(tmp.file->f_mode & FMODE_READ) ||
1567 	    (tmp.file->f_flags & O_APPEND)) {
1568 		error = -EBADF;
1569 		goto out_put_tmp_file;
1570 	}
1571 
1572 	if (IS_SWAPFILE(file_inode(f.file)) ||
1573 	    IS_SWAPFILE(file_inode(tmp.file))) {
1574 		error = -EINVAL;
1575 		goto out_put_tmp_file;
1576 	}
1577 
1578 	ip = XFS_I(file_inode(f.file));
1579 	tip = XFS_I(file_inode(tmp.file));
1580 
1581 	if (ip->i_mount != tip->i_mount) {
1582 		error = -EINVAL;
1583 		goto out_put_tmp_file;
1584 	}
1585 
1586 	if (ip->i_ino == tip->i_ino) {
1587 		error = -EINVAL;
1588 		goto out_put_tmp_file;
1589 	}
1590 
1591 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1592 		error = -EIO;
1593 		goto out_put_tmp_file;
1594 	}
1595 
1596 	error = xfs_swap_extents(ip, tip, sxp);
1597 
1598  out_put_tmp_file:
1599 	fdput(tmp);
1600  out_put_file:
1601 	fdput(f);
1602  out:
1603 	return error;
1604 }
1605 
1606 /*
1607  * Note: some of the ioctl's return positive numbers as a
1608  * byte count indicating success, such as readlink_by_handle.
1609  * So we don't "sign flip" like most other routines.  This means
1610  * true errors need to be returned as a negative value.
1611  */
1612 long
1613 xfs_file_ioctl(
1614 	struct file		*filp,
1615 	unsigned int		cmd,
1616 	unsigned long		p)
1617 {
1618 	struct inode		*inode = file_inode(filp);
1619 	struct xfs_inode	*ip = XFS_I(inode);
1620 	struct xfs_mount	*mp = ip->i_mount;
1621 	void			__user *arg = (void __user *)p;
1622 	int			ioflags = 0;
1623 	int			error;
1624 
1625 	if (filp->f_mode & FMODE_NOCMTIME)
1626 		ioflags |= XFS_IO_INVIS;
1627 
1628 	trace_xfs_file_ioctl(ip);
1629 
1630 	switch (cmd) {
1631 	case FITRIM:
1632 		return xfs_ioc_trim(mp, arg);
1633 	case XFS_IOC_ALLOCSP:
1634 	case XFS_IOC_FREESP:
1635 	case XFS_IOC_RESVSP:
1636 	case XFS_IOC_UNRESVSP:
1637 	case XFS_IOC_ALLOCSP64:
1638 	case XFS_IOC_FREESP64:
1639 	case XFS_IOC_RESVSP64:
1640 	case XFS_IOC_UNRESVSP64:
1641 	case XFS_IOC_ZERO_RANGE: {
1642 		xfs_flock64_t		bf;
1643 
1644 		if (copy_from_user(&bf, arg, sizeof(bf)))
1645 			return -EFAULT;
1646 		return xfs_ioc_space(ip, inode, filp, ioflags, cmd, &bf);
1647 	}
1648 	case XFS_IOC_DIOINFO: {
1649 		struct dioattr	da;
1650 		xfs_buftarg_t	*target =
1651 			XFS_IS_REALTIME_INODE(ip) ?
1652 			mp->m_rtdev_targp : mp->m_ddev_targp;
1653 
1654 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1655 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1656 
1657 		if (copy_to_user(arg, &da, sizeof(da)))
1658 			return -EFAULT;
1659 		return 0;
1660 	}
1661 
1662 	case XFS_IOC_FSBULKSTAT_SINGLE:
1663 	case XFS_IOC_FSBULKSTAT:
1664 	case XFS_IOC_FSINUMBERS:
1665 		return xfs_ioc_bulkstat(mp, cmd, arg);
1666 
1667 	case XFS_IOC_FSGEOMETRY_V1:
1668 		return xfs_ioc_fsgeometry_v1(mp, arg);
1669 
1670 	case XFS_IOC_FSGEOMETRY:
1671 		return xfs_ioc_fsgeometry(mp, arg);
1672 
1673 	case XFS_IOC_GETVERSION:
1674 		return put_user(inode->i_generation, (int __user *)arg);
1675 
1676 	case XFS_IOC_FSGETXATTR:
1677 		return xfs_ioc_fsgetxattr(ip, 0, arg);
1678 	case XFS_IOC_FSGETXATTRA:
1679 		return xfs_ioc_fsgetxattr(ip, 1, arg);
1680 	case XFS_IOC_FSSETXATTR:
1681 		return xfs_ioc_fssetxattr(ip, filp, arg);
1682 	case XFS_IOC_GETXFLAGS:
1683 		return xfs_ioc_getxflags(ip, arg);
1684 	case XFS_IOC_SETXFLAGS:
1685 		return xfs_ioc_setxflags(ip, filp, arg);
1686 
1687 	case XFS_IOC_FSSETDM: {
1688 		struct fsdmidata	dmi;
1689 
1690 		if (copy_from_user(&dmi, arg, sizeof(dmi)))
1691 			return -EFAULT;
1692 
1693 		error = mnt_want_write_file(filp);
1694 		if (error)
1695 			return error;
1696 
1697 		error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1698 				dmi.fsd_dmstate);
1699 		mnt_drop_write_file(filp);
1700 		return error;
1701 	}
1702 
1703 	case XFS_IOC_GETBMAP:
1704 	case XFS_IOC_GETBMAPA:
1705 		return xfs_ioc_getbmap(ip, ioflags, cmd, arg);
1706 
1707 	case XFS_IOC_GETBMAPX:
1708 		return xfs_ioc_getbmapx(ip, arg);
1709 
1710 	case XFS_IOC_FD_TO_HANDLE:
1711 	case XFS_IOC_PATH_TO_HANDLE:
1712 	case XFS_IOC_PATH_TO_FSHANDLE: {
1713 		xfs_fsop_handlereq_t	hreq;
1714 
1715 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1716 			return -EFAULT;
1717 		return xfs_find_handle(cmd, &hreq);
1718 	}
1719 	case XFS_IOC_OPEN_BY_HANDLE: {
1720 		xfs_fsop_handlereq_t	hreq;
1721 
1722 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1723 			return -EFAULT;
1724 		return xfs_open_by_handle(filp, &hreq);
1725 	}
1726 	case XFS_IOC_FSSETDM_BY_HANDLE:
1727 		return xfs_fssetdm_by_handle(filp, arg);
1728 
1729 	case XFS_IOC_READLINK_BY_HANDLE: {
1730 		xfs_fsop_handlereq_t	hreq;
1731 
1732 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1733 			return -EFAULT;
1734 		return xfs_readlink_by_handle(filp, &hreq);
1735 	}
1736 	case XFS_IOC_ATTRLIST_BY_HANDLE:
1737 		return xfs_attrlist_by_handle(filp, arg);
1738 
1739 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
1740 		return xfs_attrmulti_by_handle(filp, arg);
1741 
1742 	case XFS_IOC_SWAPEXT: {
1743 		struct xfs_swapext	sxp;
1744 
1745 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1746 			return -EFAULT;
1747 		error = mnt_want_write_file(filp);
1748 		if (error)
1749 			return error;
1750 		error = xfs_ioc_swapext(&sxp);
1751 		mnt_drop_write_file(filp);
1752 		return error;
1753 	}
1754 
1755 	case XFS_IOC_FSCOUNTS: {
1756 		xfs_fsop_counts_t out;
1757 
1758 		error = xfs_fs_counts(mp, &out);
1759 		if (error)
1760 			return error;
1761 
1762 		if (copy_to_user(arg, &out, sizeof(out)))
1763 			return -EFAULT;
1764 		return 0;
1765 	}
1766 
1767 	case XFS_IOC_SET_RESBLKS: {
1768 		xfs_fsop_resblks_t inout;
1769 		__uint64_t	   in;
1770 
1771 		if (!capable(CAP_SYS_ADMIN))
1772 			return -EPERM;
1773 
1774 		if (mp->m_flags & XFS_MOUNT_RDONLY)
1775 			return -EROFS;
1776 
1777 		if (copy_from_user(&inout, arg, sizeof(inout)))
1778 			return -EFAULT;
1779 
1780 		error = mnt_want_write_file(filp);
1781 		if (error)
1782 			return error;
1783 
1784 		/* input parameter is passed in resblks field of structure */
1785 		in = inout.resblks;
1786 		error = xfs_reserve_blocks(mp, &in, &inout);
1787 		mnt_drop_write_file(filp);
1788 		if (error)
1789 			return error;
1790 
1791 		if (copy_to_user(arg, &inout, sizeof(inout)))
1792 			return -EFAULT;
1793 		return 0;
1794 	}
1795 
1796 	case XFS_IOC_GET_RESBLKS: {
1797 		xfs_fsop_resblks_t out;
1798 
1799 		if (!capable(CAP_SYS_ADMIN))
1800 			return -EPERM;
1801 
1802 		error = xfs_reserve_blocks(mp, NULL, &out);
1803 		if (error)
1804 			return error;
1805 
1806 		if (copy_to_user(arg, &out, sizeof(out)))
1807 			return -EFAULT;
1808 
1809 		return 0;
1810 	}
1811 
1812 	case XFS_IOC_FSGROWFSDATA: {
1813 		xfs_growfs_data_t in;
1814 
1815 		if (copy_from_user(&in, arg, sizeof(in)))
1816 			return -EFAULT;
1817 
1818 		error = mnt_want_write_file(filp);
1819 		if (error)
1820 			return error;
1821 		error = xfs_growfs_data(mp, &in);
1822 		mnt_drop_write_file(filp);
1823 		return error;
1824 	}
1825 
1826 	case XFS_IOC_FSGROWFSLOG: {
1827 		xfs_growfs_log_t in;
1828 
1829 		if (copy_from_user(&in, arg, sizeof(in)))
1830 			return -EFAULT;
1831 
1832 		error = mnt_want_write_file(filp);
1833 		if (error)
1834 			return error;
1835 		error = xfs_growfs_log(mp, &in);
1836 		mnt_drop_write_file(filp);
1837 		return error;
1838 	}
1839 
1840 	case XFS_IOC_FSGROWFSRT: {
1841 		xfs_growfs_rt_t in;
1842 
1843 		if (copy_from_user(&in, arg, sizeof(in)))
1844 			return -EFAULT;
1845 
1846 		error = mnt_want_write_file(filp);
1847 		if (error)
1848 			return error;
1849 		error = xfs_growfs_rt(mp, &in);
1850 		mnt_drop_write_file(filp);
1851 		return error;
1852 	}
1853 
1854 	case XFS_IOC_GOINGDOWN: {
1855 		__uint32_t in;
1856 
1857 		if (!capable(CAP_SYS_ADMIN))
1858 			return -EPERM;
1859 
1860 		if (get_user(in, (__uint32_t __user *)arg))
1861 			return -EFAULT;
1862 
1863 		return xfs_fs_goingdown(mp, in);
1864 	}
1865 
1866 	case XFS_IOC_ERROR_INJECTION: {
1867 		xfs_error_injection_t in;
1868 
1869 		if (!capable(CAP_SYS_ADMIN))
1870 			return -EPERM;
1871 
1872 		if (copy_from_user(&in, arg, sizeof(in)))
1873 			return -EFAULT;
1874 
1875 		return xfs_errortag_add(in.errtag, mp);
1876 	}
1877 
1878 	case XFS_IOC_ERROR_CLEARALL:
1879 		if (!capable(CAP_SYS_ADMIN))
1880 			return -EPERM;
1881 
1882 		return xfs_errortag_clearall(mp, 1);
1883 
1884 	case XFS_IOC_FREE_EOFBLOCKS: {
1885 		struct xfs_fs_eofblocks eofb;
1886 		struct xfs_eofblocks keofb;
1887 
1888 		if (!capable(CAP_SYS_ADMIN))
1889 			return -EPERM;
1890 
1891 		if (mp->m_flags & XFS_MOUNT_RDONLY)
1892 			return -EROFS;
1893 
1894 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
1895 			return -EFAULT;
1896 
1897 		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
1898 		if (error)
1899 			return error;
1900 
1901 		return xfs_icache_free_eofblocks(mp, &keofb);
1902 	}
1903 
1904 	default:
1905 		return -ENOTTY;
1906 	}
1907 }
1908