xref: /openbmc/linux/fs/xfs/xfs_ioctl.c (revision f519f0be)
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  * Copyright (c) 2000-2005 Silicon Graphics, Inc.
4  * All Rights Reserved.
5  */
6 #include "xfs.h"
7 #include "xfs_fs.h"
8 #include "xfs_shared.h"
9 #include "xfs_format.h"
10 #include "xfs_log_format.h"
11 #include "xfs_trans_resv.h"
12 #include "xfs_mount.h"
13 #include "xfs_inode.h"
14 #include "xfs_ioctl.h"
15 #include "xfs_alloc.h"
16 #include "xfs_rtalloc.h"
17 #include "xfs_itable.h"
18 #include "xfs_error.h"
19 #include "xfs_attr.h"
20 #include "xfs_bmap.h"
21 #include "xfs_bmap_util.h"
22 #include "xfs_fsops.h"
23 #include "xfs_discard.h"
24 #include "xfs_quota.h"
25 #include "xfs_export.h"
26 #include "xfs_trace.h"
27 #include "xfs_icache.h"
28 #include "xfs_symlink.h"
29 #include "xfs_trans.h"
30 #include "xfs_acl.h"
31 #include "xfs_btree.h"
32 #include <linux/fsmap.h>
33 #include "xfs_fsmap.h"
34 #include "scrub/xfs_scrub.h"
35 #include "xfs_sb.h"
36 #include "xfs_ag.h"
37 #include "xfs_health.h"
38 
39 #include <linux/capability.h>
40 #include <linux/cred.h>
41 #include <linux/dcache.h>
42 #include <linux/mount.h>
43 #include <linux/namei.h>
44 #include <linux/pagemap.h>
45 #include <linux/slab.h>
46 #include <linux/exportfs.h>
47 
48 /*
49  * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
50  * a file or fs handle.
51  *
52  * XFS_IOC_PATH_TO_FSHANDLE
53  *    returns fs handle for a mount point or path within that mount point
54  * XFS_IOC_FD_TO_HANDLE
55  *    returns full handle for a FD opened in user space
56  * XFS_IOC_PATH_TO_HANDLE
57  *    returns full handle for a path
58  */
59 int
60 xfs_find_handle(
61 	unsigned int		cmd,
62 	xfs_fsop_handlereq_t	*hreq)
63 {
64 	int			hsize;
65 	xfs_handle_t		handle;
66 	struct inode		*inode;
67 	struct fd		f = {NULL};
68 	struct path		path;
69 	int			error;
70 	struct xfs_inode	*ip;
71 
72 	if (cmd == XFS_IOC_FD_TO_HANDLE) {
73 		f = fdget(hreq->fd);
74 		if (!f.file)
75 			return -EBADF;
76 		inode = file_inode(f.file);
77 	} else {
78 		error = user_lpath((const char __user *)hreq->path, &path);
79 		if (error)
80 			return error;
81 		inode = d_inode(path.dentry);
82 	}
83 	ip = XFS_I(inode);
84 
85 	/*
86 	 * We can only generate handles for inodes residing on a XFS filesystem,
87 	 * and only for regular files, directories or symbolic links.
88 	 */
89 	error = -EINVAL;
90 	if (inode->i_sb->s_magic != XFS_SB_MAGIC)
91 		goto out_put;
92 
93 	error = -EBADF;
94 	if (!S_ISREG(inode->i_mode) &&
95 	    !S_ISDIR(inode->i_mode) &&
96 	    !S_ISLNK(inode->i_mode))
97 		goto out_put;
98 
99 
100 	memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
101 
102 	if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
103 		/*
104 		 * This handle only contains an fsid, zero the rest.
105 		 */
106 		memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
107 		hsize = sizeof(xfs_fsid_t);
108 	} else {
109 		handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
110 					sizeof(handle.ha_fid.fid_len);
111 		handle.ha_fid.fid_pad = 0;
112 		handle.ha_fid.fid_gen = inode->i_generation;
113 		handle.ha_fid.fid_ino = ip->i_ino;
114 		hsize = sizeof(xfs_handle_t);
115 	}
116 
117 	error = -EFAULT;
118 	if (copy_to_user(hreq->ohandle, &handle, hsize) ||
119 	    copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
120 		goto out_put;
121 
122 	error = 0;
123 
124  out_put:
125 	if (cmd == XFS_IOC_FD_TO_HANDLE)
126 		fdput(f);
127 	else
128 		path_put(&path);
129 	return error;
130 }
131 
132 /*
133  * No need to do permission checks on the various pathname components
134  * as the handle operations are privileged.
135  */
136 STATIC int
137 xfs_handle_acceptable(
138 	void			*context,
139 	struct dentry		*dentry)
140 {
141 	return 1;
142 }
143 
144 /*
145  * Convert userspace handle data into a dentry.
146  */
147 struct dentry *
148 xfs_handle_to_dentry(
149 	struct file		*parfilp,
150 	void __user		*uhandle,
151 	u32			hlen)
152 {
153 	xfs_handle_t		handle;
154 	struct xfs_fid64	fid;
155 
156 	/*
157 	 * Only allow handle opens under a directory.
158 	 */
159 	if (!S_ISDIR(file_inode(parfilp)->i_mode))
160 		return ERR_PTR(-ENOTDIR);
161 
162 	if (hlen != sizeof(xfs_handle_t))
163 		return ERR_PTR(-EINVAL);
164 	if (copy_from_user(&handle, uhandle, hlen))
165 		return ERR_PTR(-EFAULT);
166 	if (handle.ha_fid.fid_len !=
167 	    sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
168 		return ERR_PTR(-EINVAL);
169 
170 	memset(&fid, 0, sizeof(struct fid));
171 	fid.ino = handle.ha_fid.fid_ino;
172 	fid.gen = handle.ha_fid.fid_gen;
173 
174 	return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
175 			FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
176 			xfs_handle_acceptable, NULL);
177 }
178 
179 STATIC struct dentry *
180 xfs_handlereq_to_dentry(
181 	struct file		*parfilp,
182 	xfs_fsop_handlereq_t	*hreq)
183 {
184 	return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
185 }
186 
187 int
188 xfs_open_by_handle(
189 	struct file		*parfilp,
190 	xfs_fsop_handlereq_t	*hreq)
191 {
192 	const struct cred	*cred = current_cred();
193 	int			error;
194 	int			fd;
195 	int			permflag;
196 	struct file		*filp;
197 	struct inode		*inode;
198 	struct dentry		*dentry;
199 	fmode_t			fmode;
200 	struct path		path;
201 
202 	if (!capable(CAP_SYS_ADMIN))
203 		return -EPERM;
204 
205 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
206 	if (IS_ERR(dentry))
207 		return PTR_ERR(dentry);
208 	inode = d_inode(dentry);
209 
210 	/* Restrict xfs_open_by_handle to directories & regular files. */
211 	if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
212 		error = -EPERM;
213 		goto out_dput;
214 	}
215 
216 #if BITS_PER_LONG != 32
217 	hreq->oflags |= O_LARGEFILE;
218 #endif
219 
220 	permflag = hreq->oflags;
221 	fmode = OPEN_FMODE(permflag);
222 	if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
223 	    (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
224 		error = -EPERM;
225 		goto out_dput;
226 	}
227 
228 	if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
229 		error = -EPERM;
230 		goto out_dput;
231 	}
232 
233 	/* Can't write directories. */
234 	if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
235 		error = -EISDIR;
236 		goto out_dput;
237 	}
238 
239 	fd = get_unused_fd_flags(0);
240 	if (fd < 0) {
241 		error = fd;
242 		goto out_dput;
243 	}
244 
245 	path.mnt = parfilp->f_path.mnt;
246 	path.dentry = dentry;
247 	filp = dentry_open(&path, hreq->oflags, cred);
248 	dput(dentry);
249 	if (IS_ERR(filp)) {
250 		put_unused_fd(fd);
251 		return PTR_ERR(filp);
252 	}
253 
254 	if (S_ISREG(inode->i_mode)) {
255 		filp->f_flags |= O_NOATIME;
256 		filp->f_mode |= FMODE_NOCMTIME;
257 	}
258 
259 	fd_install(fd, filp);
260 	return fd;
261 
262  out_dput:
263 	dput(dentry);
264 	return error;
265 }
266 
267 int
268 xfs_readlink_by_handle(
269 	struct file		*parfilp,
270 	xfs_fsop_handlereq_t	*hreq)
271 {
272 	struct dentry		*dentry;
273 	__u32			olen;
274 	int			error;
275 
276 	if (!capable(CAP_SYS_ADMIN))
277 		return -EPERM;
278 
279 	dentry = xfs_handlereq_to_dentry(parfilp, hreq);
280 	if (IS_ERR(dentry))
281 		return PTR_ERR(dentry);
282 
283 	/* Restrict this handle operation to symlinks only. */
284 	if (!d_is_symlink(dentry)) {
285 		error = -EINVAL;
286 		goto out_dput;
287 	}
288 
289 	if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
290 		error = -EFAULT;
291 		goto out_dput;
292 	}
293 
294 	error = vfs_readlink(dentry, hreq->ohandle, olen);
295 
296  out_dput:
297 	dput(dentry);
298 	return error;
299 }
300 
301 int
302 xfs_set_dmattrs(
303 	xfs_inode_t     *ip,
304 	uint		evmask,
305 	uint16_t	state)
306 {
307 	xfs_mount_t	*mp = ip->i_mount;
308 	xfs_trans_t	*tp;
309 	int		error;
310 
311 	if (!capable(CAP_SYS_ADMIN))
312 		return -EPERM;
313 
314 	if (XFS_FORCED_SHUTDOWN(mp))
315 		return -EIO;
316 
317 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
318 	if (error)
319 		return error;
320 
321 	xfs_ilock(ip, XFS_ILOCK_EXCL);
322 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
323 
324 	ip->i_d.di_dmevmask = evmask;
325 	ip->i_d.di_dmstate  = state;
326 
327 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
328 	error = xfs_trans_commit(tp);
329 
330 	return error;
331 }
332 
333 STATIC int
334 xfs_fssetdm_by_handle(
335 	struct file		*parfilp,
336 	void			__user *arg)
337 {
338 	int			error;
339 	struct fsdmidata	fsd;
340 	xfs_fsop_setdm_handlereq_t dmhreq;
341 	struct dentry		*dentry;
342 
343 	if (!capable(CAP_MKNOD))
344 		return -EPERM;
345 	if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
346 		return -EFAULT;
347 
348 	error = mnt_want_write_file(parfilp);
349 	if (error)
350 		return error;
351 
352 	dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
353 	if (IS_ERR(dentry)) {
354 		mnt_drop_write_file(parfilp);
355 		return PTR_ERR(dentry);
356 	}
357 
358 	if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
359 		error = -EPERM;
360 		goto out;
361 	}
362 
363 	if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
364 		error = -EFAULT;
365 		goto out;
366 	}
367 
368 	error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
369 				 fsd.fsd_dmstate);
370 
371  out:
372 	mnt_drop_write_file(parfilp);
373 	dput(dentry);
374 	return error;
375 }
376 
377 STATIC int
378 xfs_attrlist_by_handle(
379 	struct file		*parfilp,
380 	void			__user *arg)
381 {
382 	int			error = -ENOMEM;
383 	attrlist_cursor_kern_t	*cursor;
384 	struct xfs_fsop_attrlist_handlereq __user	*p = arg;
385 	xfs_fsop_attrlist_handlereq_t al_hreq;
386 	struct dentry		*dentry;
387 	char			*kbuf;
388 
389 	if (!capable(CAP_SYS_ADMIN))
390 		return -EPERM;
391 	if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
392 		return -EFAULT;
393 	if (al_hreq.buflen < sizeof(struct attrlist) ||
394 	    al_hreq.buflen > XFS_XATTR_LIST_MAX)
395 		return -EINVAL;
396 
397 	/*
398 	 * Reject flags, only allow namespaces.
399 	 */
400 	if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
401 		return -EINVAL;
402 
403 	dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
404 	if (IS_ERR(dentry))
405 		return PTR_ERR(dentry);
406 
407 	kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
408 	if (!kbuf)
409 		goto out_dput;
410 
411 	cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
412 	error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
413 					al_hreq.flags, cursor);
414 	if (error)
415 		goto out_kfree;
416 
417 	if (copy_to_user(&p->pos, cursor, sizeof(attrlist_cursor_kern_t))) {
418 		error = -EFAULT;
419 		goto out_kfree;
420 	}
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 file		*filp,
599 	unsigned int		cmd,
600 	xfs_flock64_t		*bf)
601 {
602 	struct inode		*inode = file_inode(filp);
603 	struct xfs_inode	*ip = XFS_I(inode);
604 	struct iattr		iattr;
605 	enum xfs_prealloc_flags	flags = 0;
606 	uint			iolock = XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL;
607 	int			error;
608 
609 	if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
610 		return -EPERM;
611 
612 	if (!(filp->f_mode & FMODE_WRITE))
613 		return -EBADF;
614 
615 	if (!S_ISREG(inode->i_mode))
616 		return -EINVAL;
617 
618 	if (filp->f_flags & O_DSYNC)
619 		flags |= XFS_PREALLOC_SYNC;
620 	if (filp->f_mode & FMODE_NOCMTIME)
621 		flags |= XFS_PREALLOC_INVISIBLE;
622 
623 	error = mnt_want_write_file(filp);
624 	if (error)
625 		return error;
626 
627 	xfs_ilock(ip, iolock);
628 	error = xfs_break_layouts(inode, &iolock, BREAK_UNMAP);
629 	if (error)
630 		goto out_unlock;
631 
632 	switch (bf->l_whence) {
633 	case 0: /*SEEK_SET*/
634 		break;
635 	case 1: /*SEEK_CUR*/
636 		bf->l_start += filp->f_pos;
637 		break;
638 	case 2: /*SEEK_END*/
639 		bf->l_start += XFS_ISIZE(ip);
640 		break;
641 	default:
642 		error = -EINVAL;
643 		goto out_unlock;
644 	}
645 
646 	/*
647 	 * length of <= 0 for resv/unresv/zero is invalid.  length for
648 	 * alloc/free is ignored completely and we have no idea what userspace
649 	 * might have set it to, so set it to zero to allow range
650 	 * checks to pass.
651 	 */
652 	switch (cmd) {
653 	case XFS_IOC_ZERO_RANGE:
654 	case XFS_IOC_RESVSP:
655 	case XFS_IOC_RESVSP64:
656 	case XFS_IOC_UNRESVSP:
657 	case XFS_IOC_UNRESVSP64:
658 		if (bf->l_len <= 0) {
659 			error = -EINVAL;
660 			goto out_unlock;
661 		}
662 		break;
663 	default:
664 		bf->l_len = 0;
665 		break;
666 	}
667 
668 	if (bf->l_start < 0 ||
669 	    bf->l_start > inode->i_sb->s_maxbytes ||
670 	    bf->l_start + bf->l_len < 0 ||
671 	    bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
672 		error = -EINVAL;
673 		goto out_unlock;
674 	}
675 
676 	switch (cmd) {
677 	case XFS_IOC_ZERO_RANGE:
678 		flags |= XFS_PREALLOC_SET;
679 		error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
680 		break;
681 	case XFS_IOC_RESVSP:
682 	case XFS_IOC_RESVSP64:
683 		flags |= XFS_PREALLOC_SET;
684 		error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
685 						XFS_BMAPI_PREALLOC);
686 		break;
687 	case XFS_IOC_UNRESVSP:
688 	case XFS_IOC_UNRESVSP64:
689 		error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
690 		break;
691 	case XFS_IOC_ALLOCSP:
692 	case XFS_IOC_ALLOCSP64:
693 	case XFS_IOC_FREESP:
694 	case XFS_IOC_FREESP64:
695 		flags |= XFS_PREALLOC_CLEAR;
696 		if (bf->l_start > XFS_ISIZE(ip)) {
697 			error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
698 					bf->l_start - XFS_ISIZE(ip), 0);
699 			if (error)
700 				goto out_unlock;
701 		}
702 
703 		iattr.ia_valid = ATTR_SIZE;
704 		iattr.ia_size = bf->l_start;
705 
706 		error = xfs_vn_setattr_size(file_dentry(filp), &iattr);
707 		break;
708 	default:
709 		ASSERT(0);
710 		error = -EINVAL;
711 	}
712 
713 	if (error)
714 		goto out_unlock;
715 
716 	error = xfs_update_prealloc_flags(ip, flags);
717 
718 out_unlock:
719 	xfs_iunlock(ip, iolock);
720 	mnt_drop_write_file(filp);
721 	return error;
722 }
723 
724 STATIC int
725 xfs_ioc_bulkstat(
726 	xfs_mount_t		*mp,
727 	unsigned int		cmd,
728 	void			__user *arg)
729 {
730 	xfs_fsop_bulkreq_t	bulkreq;
731 	int			count;	/* # of records returned */
732 	xfs_ino_t		inlast;	/* last inode number */
733 	int			done;
734 	int			error;
735 
736 	/* done = 1 if there are more stats to get and if bulkstat */
737 	/* should be called again (unused here, but used in dmapi) */
738 
739 	if (!capable(CAP_SYS_ADMIN))
740 		return -EPERM;
741 
742 	if (XFS_FORCED_SHUTDOWN(mp))
743 		return -EIO;
744 
745 	if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
746 		return -EFAULT;
747 
748 	if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
749 		return -EFAULT;
750 
751 	if ((count = bulkreq.icount) <= 0)
752 		return -EINVAL;
753 
754 	if (bulkreq.ubuffer == NULL)
755 		return -EINVAL;
756 
757 	if (cmd == XFS_IOC_FSINUMBERS)
758 		error = xfs_inumbers(mp, &inlast, &count,
759 					bulkreq.ubuffer, xfs_inumbers_fmt);
760 	else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
761 		error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
762 					sizeof(xfs_bstat_t), NULL, &done);
763 	else	/* XFS_IOC_FSBULKSTAT */
764 		error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
765 				     sizeof(xfs_bstat_t), bulkreq.ubuffer,
766 				     &done);
767 
768 	if (error)
769 		return error;
770 
771 	if (bulkreq.ocount != NULL) {
772 		if (copy_to_user(bulkreq.lastip, &inlast,
773 						sizeof(xfs_ino_t)))
774 			return -EFAULT;
775 
776 		if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
777 			return -EFAULT;
778 	}
779 
780 	return 0;
781 }
782 
783 STATIC int
784 xfs_ioc_fsgeometry(
785 	struct xfs_mount	*mp,
786 	void			__user *arg,
787 	int			struct_version)
788 {
789 	struct xfs_fsop_geom	fsgeo;
790 	size_t			len;
791 
792 	xfs_fs_geometry(&mp->m_sb, &fsgeo, struct_version);
793 
794 	if (struct_version <= 3)
795 		len = sizeof(struct xfs_fsop_geom_v1);
796 	else if (struct_version == 4)
797 		len = sizeof(struct xfs_fsop_geom_v4);
798 	else {
799 		xfs_fsop_geom_health(mp, &fsgeo);
800 		len = sizeof(fsgeo);
801 	}
802 
803 	if (copy_to_user(arg, &fsgeo, len))
804 		return -EFAULT;
805 	return 0;
806 }
807 
808 STATIC int
809 xfs_ioc_ag_geometry(
810 	struct xfs_mount	*mp,
811 	void			__user *arg)
812 {
813 	struct xfs_ag_geometry	ageo;
814 	int			error;
815 
816 	if (copy_from_user(&ageo, arg, sizeof(ageo)))
817 		return -EFAULT;
818 
819 	error = xfs_ag_get_geometry(mp, ageo.ag_number, &ageo);
820 	if (error)
821 		return error;
822 
823 	if (copy_to_user(arg, &ageo, sizeof(ageo)))
824 		return -EFAULT;
825 	return 0;
826 }
827 
828 /*
829  * Linux extended inode flags interface.
830  */
831 
832 STATIC unsigned int
833 xfs_merge_ioc_xflags(
834 	unsigned int	flags,
835 	unsigned int	start)
836 {
837 	unsigned int	xflags = start;
838 
839 	if (flags & FS_IMMUTABLE_FL)
840 		xflags |= FS_XFLAG_IMMUTABLE;
841 	else
842 		xflags &= ~FS_XFLAG_IMMUTABLE;
843 	if (flags & FS_APPEND_FL)
844 		xflags |= FS_XFLAG_APPEND;
845 	else
846 		xflags &= ~FS_XFLAG_APPEND;
847 	if (flags & FS_SYNC_FL)
848 		xflags |= FS_XFLAG_SYNC;
849 	else
850 		xflags &= ~FS_XFLAG_SYNC;
851 	if (flags & FS_NOATIME_FL)
852 		xflags |= FS_XFLAG_NOATIME;
853 	else
854 		xflags &= ~FS_XFLAG_NOATIME;
855 	if (flags & FS_NODUMP_FL)
856 		xflags |= FS_XFLAG_NODUMP;
857 	else
858 		xflags &= ~FS_XFLAG_NODUMP;
859 
860 	return xflags;
861 }
862 
863 STATIC unsigned int
864 xfs_di2lxflags(
865 	uint16_t	di_flags)
866 {
867 	unsigned int	flags = 0;
868 
869 	if (di_flags & XFS_DIFLAG_IMMUTABLE)
870 		flags |= FS_IMMUTABLE_FL;
871 	if (di_flags & XFS_DIFLAG_APPEND)
872 		flags |= FS_APPEND_FL;
873 	if (di_flags & XFS_DIFLAG_SYNC)
874 		flags |= FS_SYNC_FL;
875 	if (di_flags & XFS_DIFLAG_NOATIME)
876 		flags |= FS_NOATIME_FL;
877 	if (di_flags & XFS_DIFLAG_NODUMP)
878 		flags |= FS_NODUMP_FL;
879 	return flags;
880 }
881 
882 STATIC int
883 xfs_ioc_fsgetxattr(
884 	xfs_inode_t		*ip,
885 	int			attr,
886 	void			__user *arg)
887 {
888 	struct fsxattr		fa;
889 
890 	memset(&fa, 0, sizeof(struct fsxattr));
891 
892 	xfs_ilock(ip, XFS_ILOCK_SHARED);
893 	fa.fsx_xflags = xfs_ip2xflags(ip);
894 	fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
895 	fa.fsx_cowextsize = ip->i_d.di_cowextsize <<
896 			ip->i_mount->m_sb.sb_blocklog;
897 	fa.fsx_projid = xfs_get_projid(ip);
898 
899 	if (attr) {
900 		if (ip->i_afp) {
901 			if (ip->i_afp->if_flags & XFS_IFEXTENTS)
902 				fa.fsx_nextents = xfs_iext_count(ip->i_afp);
903 			else
904 				fa.fsx_nextents = ip->i_d.di_anextents;
905 		} else
906 			fa.fsx_nextents = 0;
907 	} else {
908 		if (ip->i_df.if_flags & XFS_IFEXTENTS)
909 			fa.fsx_nextents = xfs_iext_count(&ip->i_df);
910 		else
911 			fa.fsx_nextents = ip->i_d.di_nextents;
912 	}
913 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
914 
915 	if (copy_to_user(arg, &fa, sizeof(fa)))
916 		return -EFAULT;
917 	return 0;
918 }
919 
920 STATIC uint16_t
921 xfs_flags2diflags(
922 	struct xfs_inode	*ip,
923 	unsigned int		xflags)
924 {
925 	/* can't set PREALLOC this way, just preserve it */
926 	uint16_t		di_flags =
927 		(ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
928 
929 	if (xflags & FS_XFLAG_IMMUTABLE)
930 		di_flags |= XFS_DIFLAG_IMMUTABLE;
931 	if (xflags & FS_XFLAG_APPEND)
932 		di_flags |= XFS_DIFLAG_APPEND;
933 	if (xflags & FS_XFLAG_SYNC)
934 		di_flags |= XFS_DIFLAG_SYNC;
935 	if (xflags & FS_XFLAG_NOATIME)
936 		di_flags |= XFS_DIFLAG_NOATIME;
937 	if (xflags & FS_XFLAG_NODUMP)
938 		di_flags |= XFS_DIFLAG_NODUMP;
939 	if (xflags & FS_XFLAG_NODEFRAG)
940 		di_flags |= XFS_DIFLAG_NODEFRAG;
941 	if (xflags & FS_XFLAG_FILESTREAM)
942 		di_flags |= XFS_DIFLAG_FILESTREAM;
943 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
944 		if (xflags & FS_XFLAG_RTINHERIT)
945 			di_flags |= XFS_DIFLAG_RTINHERIT;
946 		if (xflags & FS_XFLAG_NOSYMLINKS)
947 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
948 		if (xflags & FS_XFLAG_EXTSZINHERIT)
949 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
950 		if (xflags & FS_XFLAG_PROJINHERIT)
951 			di_flags |= XFS_DIFLAG_PROJINHERIT;
952 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
953 		if (xflags & FS_XFLAG_REALTIME)
954 			di_flags |= XFS_DIFLAG_REALTIME;
955 		if (xflags & FS_XFLAG_EXTSIZE)
956 			di_flags |= XFS_DIFLAG_EXTSIZE;
957 	}
958 
959 	return di_flags;
960 }
961 
962 STATIC uint64_t
963 xfs_flags2diflags2(
964 	struct xfs_inode	*ip,
965 	unsigned int		xflags)
966 {
967 	uint64_t		di_flags2 =
968 		(ip->i_d.di_flags2 & XFS_DIFLAG2_REFLINK);
969 
970 	if (xflags & FS_XFLAG_DAX)
971 		di_flags2 |= XFS_DIFLAG2_DAX;
972 	if (xflags & FS_XFLAG_COWEXTSIZE)
973 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
974 
975 	return di_flags2;
976 }
977 
978 STATIC void
979 xfs_diflags_to_linux(
980 	struct xfs_inode	*ip)
981 {
982 	struct inode		*inode = VFS_I(ip);
983 	unsigned int		xflags = xfs_ip2xflags(ip);
984 
985 	if (xflags & FS_XFLAG_IMMUTABLE)
986 		inode->i_flags |= S_IMMUTABLE;
987 	else
988 		inode->i_flags &= ~S_IMMUTABLE;
989 	if (xflags & FS_XFLAG_APPEND)
990 		inode->i_flags |= S_APPEND;
991 	else
992 		inode->i_flags &= ~S_APPEND;
993 	if (xflags & FS_XFLAG_SYNC)
994 		inode->i_flags |= S_SYNC;
995 	else
996 		inode->i_flags &= ~S_SYNC;
997 	if (xflags & FS_XFLAG_NOATIME)
998 		inode->i_flags |= S_NOATIME;
999 	else
1000 		inode->i_flags &= ~S_NOATIME;
1001 #if 0	/* disabled until the flag switching races are sorted out */
1002 	if (xflags & FS_XFLAG_DAX)
1003 		inode->i_flags |= S_DAX;
1004 	else
1005 		inode->i_flags &= ~S_DAX;
1006 #endif
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 	uint64_t		di_flags2;
1017 
1018 	/* Can't change realtime flag if any extents are allocated. */
1019 	if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1020 	    XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1021 		return -EINVAL;
1022 
1023 	/* If realtime flag is set then must have realtime device */
1024 	if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1025 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1026 		    (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1027 			return -EINVAL;
1028 	}
1029 
1030 	/* Clear reflink if we are actually able to set the rt flag. */
1031 	if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1032 		ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
1033 
1034 	/* Don't allow us to set DAX mode for a reflinked file for now. */
1035 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1036 		return -EINVAL;
1037 
1038 	/*
1039 	 * Can't modify an immutable/append-only file unless
1040 	 * we have appropriate permission.
1041 	 */
1042 	if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1043 	     (fa->fsx_xflags & (FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))) &&
1044 	    !capable(CAP_LINUX_IMMUTABLE))
1045 		return -EPERM;
1046 
1047 	/* diflags2 only valid for v3 inodes. */
1048 	di_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1049 	if (di_flags2 && ip->i_d.di_version < 3)
1050 		return -EINVAL;
1051 
1052 	ip->i_d.di_flags = xfs_flags2diflags(ip, fa->fsx_xflags);
1053 	ip->i_d.di_flags2 = di_flags2;
1054 
1055 	xfs_diflags_to_linux(ip);
1056 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1057 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1058 	XFS_STATS_INC(mp, xs_ig_attrchg);
1059 	return 0;
1060 }
1061 
1062 /*
1063  * If we are changing DAX flags, we have to ensure the file is clean and any
1064  * cached objects in the address space are invalidated and removed. This
1065  * requires us to lock out other IO and page faults similar to a truncate
1066  * operation. The locks need to be held until the transaction has been committed
1067  * so that the cache invalidation is atomic with respect to the DAX flag
1068  * manipulation.
1069  */
1070 static int
1071 xfs_ioctl_setattr_dax_invalidate(
1072 	struct xfs_inode	*ip,
1073 	struct fsxattr		*fa,
1074 	int			*join_flags)
1075 {
1076 	struct inode		*inode = VFS_I(ip);
1077 	struct super_block	*sb = inode->i_sb;
1078 	int			error;
1079 
1080 	*join_flags = 0;
1081 
1082 	/*
1083 	 * It is only valid to set the DAX flag on regular files and
1084 	 * directories on filesystems where the block size is equal to the page
1085 	 * size. On directories it serves as an inherited hint so we don't
1086 	 * have to check the device for dax support or flush pagecache.
1087 	 */
1088 	if (fa->fsx_xflags & FS_XFLAG_DAX) {
1089 		if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1090 			return -EINVAL;
1091 		if (S_ISREG(inode->i_mode) &&
1092 		    !bdev_dax_supported(xfs_find_bdev_for_inode(VFS_I(ip)),
1093 				sb->s_blocksize))
1094 			return -EINVAL;
1095 	}
1096 
1097 	/* If the DAX state is not changing, we have nothing to do here. */
1098 	if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
1099 		return 0;
1100 	if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
1101 		return 0;
1102 
1103 	if (S_ISDIR(inode->i_mode))
1104 		return 0;
1105 
1106 	/* lock, flush and invalidate mapping in preparation for flag change */
1107 	xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1108 	error = filemap_write_and_wait(inode->i_mapping);
1109 	if (error)
1110 		goto out_unlock;
1111 	error = invalidate_inode_pages2(inode->i_mapping);
1112 	if (error)
1113 		goto out_unlock;
1114 
1115 	*join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
1116 	return 0;
1117 
1118 out_unlock:
1119 	xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1120 	return error;
1121 
1122 }
1123 
1124 /*
1125  * Set up the transaction structure for the setattr operation, checking that we
1126  * have permission to do so. On success, return a clean transaction and the
1127  * inode locked exclusively ready for further operation specific checks. On
1128  * failure, return an error without modifying or locking the inode.
1129  *
1130  * The inode might already be IO locked on call. If this is the case, it is
1131  * indicated in @join_flags and we take full responsibility for ensuring they
1132  * are unlocked from now on. Hence if we have an error here, we still have to
1133  * unlock them. Otherwise, once they are joined to the transaction, they will
1134  * be unlocked on commit/cancel.
1135  */
1136 static struct xfs_trans *
1137 xfs_ioctl_setattr_get_trans(
1138 	struct xfs_inode	*ip,
1139 	int			join_flags)
1140 {
1141 	struct xfs_mount	*mp = ip->i_mount;
1142 	struct xfs_trans	*tp;
1143 	int			error = -EROFS;
1144 
1145 	if (mp->m_flags & XFS_MOUNT_RDONLY)
1146 		goto out_unlock;
1147 	error = -EIO;
1148 	if (XFS_FORCED_SHUTDOWN(mp))
1149 		goto out_unlock;
1150 
1151 	error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
1152 	if (error)
1153 		goto out_unlock;
1154 
1155 	xfs_ilock(ip, XFS_ILOCK_EXCL);
1156 	xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
1157 	join_flags = 0;
1158 
1159 	/*
1160 	 * CAP_FOWNER overrides the following restrictions:
1161 	 *
1162 	 * The user ID of the calling process must be equal to the file owner
1163 	 * ID, except in cases where the CAP_FSETID capability is applicable.
1164 	 */
1165 	if (!inode_owner_or_capable(VFS_I(ip))) {
1166 		error = -EPERM;
1167 		goto out_cancel;
1168 	}
1169 
1170 	if (mp->m_flags & XFS_MOUNT_WSYNC)
1171 		xfs_trans_set_sync(tp);
1172 
1173 	return tp;
1174 
1175 out_cancel:
1176 	xfs_trans_cancel(tp);
1177 out_unlock:
1178 	if (join_flags)
1179 		xfs_iunlock(ip, join_flags);
1180 	return ERR_PTR(error);
1181 }
1182 
1183 /*
1184  * extent size hint validation is somewhat cumbersome. Rules are:
1185  *
1186  * 1. extent size hint is only valid for directories and regular files
1187  * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1188  * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1189  * 4. can only be changed on regular files if no extents are allocated
1190  * 5. can be changed on directories at any time
1191  * 6. extsize hint of 0 turns off hints, clears inode flags.
1192  * 7. Extent size must be a multiple of the appropriate block size.
1193  * 8. for non-realtime files, the extent size hint must be limited
1194  *    to half the AG size to avoid alignment extending the extent beyond the
1195  *    limits of the AG.
1196  *
1197  * Please keep this function in sync with xfs_scrub_inode_extsize.
1198  */
1199 static int
1200 xfs_ioctl_setattr_check_extsize(
1201 	struct xfs_inode	*ip,
1202 	struct fsxattr		*fa)
1203 {
1204 	struct xfs_mount	*mp = ip->i_mount;
1205 
1206 	if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(VFS_I(ip)->i_mode))
1207 		return -EINVAL;
1208 
1209 	if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
1210 	    !S_ISDIR(VFS_I(ip)->i_mode))
1211 		return -EINVAL;
1212 
1213 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_d.di_nextents &&
1214 	    ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1215 		return -EINVAL;
1216 
1217 	if (fa->fsx_extsize != 0) {
1218 		xfs_extlen_t    size;
1219 		xfs_fsblock_t   extsize_fsb;
1220 
1221 		extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1222 		if (extsize_fsb > MAXEXTLEN)
1223 			return -EINVAL;
1224 
1225 		if (XFS_IS_REALTIME_INODE(ip) ||
1226 		    (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1227 			size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1228 		} else {
1229 			size = mp->m_sb.sb_blocksize;
1230 			if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1231 				return -EINVAL;
1232 		}
1233 
1234 		if (fa->fsx_extsize % size)
1235 			return -EINVAL;
1236 	} else
1237 		fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
1238 
1239 	return 0;
1240 }
1241 
1242 /*
1243  * CoW extent size hint validation rules are:
1244  *
1245  * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
1246  *    The inode does not have to have any shared blocks, but it must be a v3.
1247  * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
1248  *    for a directory, the hint is propagated to new files.
1249  * 3. Can be changed on files & directories at any time.
1250  * 4. CoW extsize hint of 0 turns off hints, clears inode flags.
1251  * 5. Extent size must be a multiple of the appropriate block size.
1252  * 6. The extent size hint must be limited to half the AG size to avoid
1253  *    alignment extending the extent beyond the limits of the AG.
1254  *
1255  * Please keep this function in sync with xfs_scrub_inode_cowextsize.
1256  */
1257 static int
1258 xfs_ioctl_setattr_check_cowextsize(
1259 	struct xfs_inode	*ip,
1260 	struct fsxattr		*fa)
1261 {
1262 	struct xfs_mount	*mp = ip->i_mount;
1263 
1264 	if (!(fa->fsx_xflags & FS_XFLAG_COWEXTSIZE))
1265 		return 0;
1266 
1267 	if (!xfs_sb_version_hasreflink(&ip->i_mount->m_sb) ||
1268 	    ip->i_d.di_version != 3)
1269 		return -EINVAL;
1270 
1271 	if (!S_ISREG(VFS_I(ip)->i_mode) && !S_ISDIR(VFS_I(ip)->i_mode))
1272 		return -EINVAL;
1273 
1274 	if (fa->fsx_cowextsize != 0) {
1275 		xfs_extlen_t    size;
1276 		xfs_fsblock_t   cowextsize_fsb;
1277 
1278 		cowextsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1279 		if (cowextsize_fsb > MAXEXTLEN)
1280 			return -EINVAL;
1281 
1282 		size = mp->m_sb.sb_blocksize;
1283 		if (cowextsize_fsb > mp->m_sb.sb_agblocks / 2)
1284 			return -EINVAL;
1285 
1286 		if (fa->fsx_cowextsize % size)
1287 			return -EINVAL;
1288 	} else
1289 		fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
1290 
1291 	return 0;
1292 }
1293 
1294 static int
1295 xfs_ioctl_setattr_check_projid(
1296 	struct xfs_inode	*ip,
1297 	struct fsxattr		*fa)
1298 {
1299 	/* Disallow 32bit project ids if projid32bit feature is not enabled. */
1300 	if (fa->fsx_projid > (uint16_t)-1 &&
1301 	    !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1302 		return -EINVAL;
1303 
1304 	/*
1305 	 * Project Quota ID state is only allowed to change from within the init
1306 	 * namespace. Enforce that restriction only if we are trying to change
1307 	 * the quota ID state. Everything else is allowed in user namespaces.
1308 	 */
1309 	if (current_user_ns() == &init_user_ns)
1310 		return 0;
1311 
1312 	if (xfs_get_projid(ip) != fa->fsx_projid)
1313 		return -EINVAL;
1314 	if ((fa->fsx_xflags & FS_XFLAG_PROJINHERIT) !=
1315 	    (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1316 		return -EINVAL;
1317 
1318 	return 0;
1319 }
1320 
1321 STATIC int
1322 xfs_ioctl_setattr(
1323 	xfs_inode_t		*ip,
1324 	struct fsxattr		*fa)
1325 {
1326 	struct xfs_mount	*mp = ip->i_mount;
1327 	struct xfs_trans	*tp;
1328 	struct xfs_dquot	*udqp = NULL;
1329 	struct xfs_dquot	*pdqp = NULL;
1330 	struct xfs_dquot	*olddquot = NULL;
1331 	int			code;
1332 	int			join_flags = 0;
1333 
1334 	trace_xfs_ioctl_setattr(ip);
1335 
1336 	code = xfs_ioctl_setattr_check_projid(ip, fa);
1337 	if (code)
1338 		return code;
1339 
1340 	/*
1341 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
1342 	 * before we start any other transactions. Trying to do this later
1343 	 * is messy. We don't care to take a readlock to look at the ids
1344 	 * in inode here, because we can't hold it across the trans_reserve.
1345 	 * If the IDs do change before we take the ilock, we're covered
1346 	 * because the i_*dquot fields will get updated anyway.
1347 	 */
1348 	if (XFS_IS_QUOTA_ON(mp)) {
1349 		code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1350 					 ip->i_d.di_gid, fa->fsx_projid,
1351 					 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1352 		if (code)
1353 			return code;
1354 	}
1355 
1356 	/*
1357 	 * Changing DAX config may require inode locking for mapping
1358 	 * invalidation. These need to be held all the way to transaction commit
1359 	 * or cancel time, so need to be passed through to
1360 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1361 	 * appropriately.
1362 	 */
1363 	code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
1364 	if (code)
1365 		goto error_free_dquots;
1366 
1367 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1368 	if (IS_ERR(tp)) {
1369 		code = PTR_ERR(tp);
1370 		goto error_free_dquots;
1371 	}
1372 
1373 
1374 	if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1375 	    xfs_get_projid(ip) != fa->fsx_projid) {
1376 		code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1377 				capable(CAP_FOWNER) ?  XFS_QMOPT_FORCE_RES : 0);
1378 		if (code)	/* out of quota */
1379 			goto error_trans_cancel;
1380 	}
1381 
1382 	code = xfs_ioctl_setattr_check_extsize(ip, fa);
1383 	if (code)
1384 		goto error_trans_cancel;
1385 
1386 	code = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1387 	if (code)
1388 		goto error_trans_cancel;
1389 
1390 	code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1391 	if (code)
1392 		goto error_trans_cancel;
1393 
1394 	/*
1395 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
1396 	 * overrides the following restrictions:
1397 	 *
1398 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1399 	 * successful return from chown()
1400 	 */
1401 
1402 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1403 	    !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1404 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1405 
1406 	/* Change the ownerships and register project quota modifications */
1407 	if (xfs_get_projid(ip) != fa->fsx_projid) {
1408 		if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1409 			olddquot = xfs_qm_vop_chown(tp, ip,
1410 						&ip->i_pdquot, pdqp);
1411 		}
1412 		ASSERT(ip->i_d.di_version > 1);
1413 		xfs_set_projid(ip, fa->fsx_projid);
1414 	}
1415 
1416 	/*
1417 	 * Only set the extent size hint if we've already determined that the
1418 	 * extent size hint should be set on the inode. If no extent size flags
1419 	 * are set on the inode then unconditionally clear the extent size hint.
1420 	 */
1421 	if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1422 		ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1423 	else
1424 		ip->i_d.di_extsize = 0;
1425 	if (ip->i_d.di_version == 3 &&
1426 	    (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
1427 		ip->i_d.di_cowextsize = fa->fsx_cowextsize >>
1428 				mp->m_sb.sb_blocklog;
1429 	else
1430 		ip->i_d.di_cowextsize = 0;
1431 
1432 	code = xfs_trans_commit(tp);
1433 
1434 	/*
1435 	 * Release any dquot(s) the inode had kept before chown.
1436 	 */
1437 	xfs_qm_dqrele(olddquot);
1438 	xfs_qm_dqrele(udqp);
1439 	xfs_qm_dqrele(pdqp);
1440 
1441 	return code;
1442 
1443 error_trans_cancel:
1444 	xfs_trans_cancel(tp);
1445 error_free_dquots:
1446 	xfs_qm_dqrele(udqp);
1447 	xfs_qm_dqrele(pdqp);
1448 	return code;
1449 }
1450 
1451 STATIC int
1452 xfs_ioc_fssetxattr(
1453 	xfs_inode_t		*ip,
1454 	struct file		*filp,
1455 	void			__user *arg)
1456 {
1457 	struct fsxattr		fa;
1458 	int error;
1459 
1460 	if (copy_from_user(&fa, arg, sizeof(fa)))
1461 		return -EFAULT;
1462 
1463 	error = mnt_want_write_file(filp);
1464 	if (error)
1465 		return error;
1466 	error = xfs_ioctl_setattr(ip, &fa);
1467 	mnt_drop_write_file(filp);
1468 	return error;
1469 }
1470 
1471 STATIC int
1472 xfs_ioc_getxflags(
1473 	xfs_inode_t		*ip,
1474 	void			__user *arg)
1475 {
1476 	unsigned int		flags;
1477 
1478 	flags = xfs_di2lxflags(ip->i_d.di_flags);
1479 	if (copy_to_user(arg, &flags, sizeof(flags)))
1480 		return -EFAULT;
1481 	return 0;
1482 }
1483 
1484 STATIC int
1485 xfs_ioc_setxflags(
1486 	struct xfs_inode	*ip,
1487 	struct file		*filp,
1488 	void			__user *arg)
1489 {
1490 	struct xfs_trans	*tp;
1491 	struct fsxattr		fa;
1492 	unsigned int		flags;
1493 	int			join_flags = 0;
1494 	int			error;
1495 
1496 	if (copy_from_user(&flags, arg, sizeof(flags)))
1497 		return -EFAULT;
1498 
1499 	if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1500 		      FS_NOATIME_FL | FS_NODUMP_FL | \
1501 		      FS_SYNC_FL))
1502 		return -EOPNOTSUPP;
1503 
1504 	fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1505 
1506 	error = mnt_want_write_file(filp);
1507 	if (error)
1508 		return error;
1509 
1510 	/*
1511 	 * Changing DAX config may require inode locking for mapping
1512 	 * invalidation. These need to be held all the way to transaction commit
1513 	 * or cancel time, so need to be passed through to
1514 	 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1515 	 * appropriately.
1516 	 */
1517 	error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
1518 	if (error)
1519 		goto out_drop_write;
1520 
1521 	tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1522 	if (IS_ERR(tp)) {
1523 		error = PTR_ERR(tp);
1524 		goto out_drop_write;
1525 	}
1526 
1527 	error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1528 	if (error) {
1529 		xfs_trans_cancel(tp);
1530 		goto out_drop_write;
1531 	}
1532 
1533 	error = xfs_trans_commit(tp);
1534 out_drop_write:
1535 	mnt_drop_write_file(filp);
1536 	return error;
1537 }
1538 
1539 static bool
1540 xfs_getbmap_format(
1541 	struct kgetbmap		*p,
1542 	struct getbmapx __user	*u,
1543 	size_t			recsize)
1544 {
1545 	if (put_user(p->bmv_offset, &u->bmv_offset) ||
1546 	    put_user(p->bmv_block, &u->bmv_block) ||
1547 	    put_user(p->bmv_length, &u->bmv_length) ||
1548 	    put_user(0, &u->bmv_count) ||
1549 	    put_user(0, &u->bmv_entries))
1550 		return false;
1551 	if (recsize < sizeof(struct getbmapx))
1552 		return true;
1553 	if (put_user(0, &u->bmv_iflags) ||
1554 	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
1555 	    put_user(0, &u->bmv_unused1) ||
1556 	    put_user(0, &u->bmv_unused2))
1557 		return false;
1558 	return true;
1559 }
1560 
1561 STATIC int
1562 xfs_ioc_getbmap(
1563 	struct file		*file,
1564 	unsigned int		cmd,
1565 	void			__user *arg)
1566 {
1567 	struct getbmapx		bmx = { 0 };
1568 	struct kgetbmap		*buf;
1569 	size_t			recsize;
1570 	int			error, i;
1571 
1572 	switch (cmd) {
1573 	case XFS_IOC_GETBMAPA:
1574 		bmx.bmv_iflags = BMV_IF_ATTRFORK;
1575 		/*FALLTHRU*/
1576 	case XFS_IOC_GETBMAP:
1577 		if (file->f_mode & FMODE_NOCMTIME)
1578 			bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1579 		/* struct getbmap is a strict subset of struct getbmapx. */
1580 		recsize = sizeof(struct getbmap);
1581 		break;
1582 	case XFS_IOC_GETBMAPX:
1583 		recsize = sizeof(struct getbmapx);
1584 		break;
1585 	default:
1586 		return -EINVAL;
1587 	}
1588 
1589 	if (copy_from_user(&bmx, arg, recsize))
1590 		return -EFAULT;
1591 
1592 	if (bmx.bmv_count < 2)
1593 		return -EINVAL;
1594 	if (bmx.bmv_count > ULONG_MAX / recsize)
1595 		return -ENOMEM;
1596 
1597 	buf = kmem_zalloc_large(bmx.bmv_count * sizeof(*buf), 0);
1598 	if (!buf)
1599 		return -ENOMEM;
1600 
1601 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1602 	if (error)
1603 		goto out_free_buf;
1604 
1605 	error = -EFAULT;
1606 	if (copy_to_user(arg, &bmx, recsize))
1607 		goto out_free_buf;
1608 	arg += recsize;
1609 
1610 	for (i = 0; i < bmx.bmv_entries; i++) {
1611 		if (!xfs_getbmap_format(buf + i, arg, recsize))
1612 			goto out_free_buf;
1613 		arg += recsize;
1614 	}
1615 
1616 	error = 0;
1617 out_free_buf:
1618 	kmem_free(buf);
1619 	return error;
1620 }
1621 
1622 struct getfsmap_info {
1623 	struct xfs_mount	*mp;
1624 	struct fsmap_head __user *data;
1625 	unsigned int		idx;
1626 	__u32			last_flags;
1627 };
1628 
1629 STATIC int
1630 xfs_getfsmap_format(struct xfs_fsmap *xfm, void *priv)
1631 {
1632 	struct getfsmap_info	*info = priv;
1633 	struct fsmap		fm;
1634 
1635 	trace_xfs_getfsmap_mapping(info->mp, xfm);
1636 
1637 	info->last_flags = xfm->fmr_flags;
1638 	xfs_fsmap_from_internal(&fm, xfm);
1639 	if (copy_to_user(&info->data->fmh_recs[info->idx++], &fm,
1640 			sizeof(struct fsmap)))
1641 		return -EFAULT;
1642 
1643 	return 0;
1644 }
1645 
1646 STATIC int
1647 xfs_ioc_getfsmap(
1648 	struct xfs_inode	*ip,
1649 	struct fsmap_head	__user *arg)
1650 {
1651 	struct getfsmap_info	info = { NULL };
1652 	struct xfs_fsmap_head	xhead = {0};
1653 	struct fsmap_head	head;
1654 	bool			aborted = false;
1655 	int			error;
1656 
1657 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1658 		return -EFAULT;
1659 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1660 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1661 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
1662 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1663 		       sizeof(head.fmh_keys[1].fmr_reserved)))
1664 		return -EINVAL;
1665 
1666 	xhead.fmh_iflags = head.fmh_iflags;
1667 	xhead.fmh_count = head.fmh_count;
1668 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1669 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1670 
1671 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1672 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1673 
1674 	info.mp = ip->i_mount;
1675 	info.data = arg;
1676 	error = xfs_getfsmap(ip->i_mount, &xhead, xfs_getfsmap_format, &info);
1677 	if (error == XFS_BTREE_QUERY_RANGE_ABORT) {
1678 		error = 0;
1679 		aborted = true;
1680 	} else if (error)
1681 		return error;
1682 
1683 	/* If we didn't abort, set the "last" flag in the last fmx */
1684 	if (!aborted && info.idx) {
1685 		info.last_flags |= FMR_OF_LAST;
1686 		if (copy_to_user(&info.data->fmh_recs[info.idx - 1].fmr_flags,
1687 				&info.last_flags, sizeof(info.last_flags)))
1688 			return -EFAULT;
1689 	}
1690 
1691 	/* copy back header */
1692 	head.fmh_entries = xhead.fmh_entries;
1693 	head.fmh_oflags = xhead.fmh_oflags;
1694 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head)))
1695 		return -EFAULT;
1696 
1697 	return 0;
1698 }
1699 
1700 STATIC int
1701 xfs_ioc_scrub_metadata(
1702 	struct xfs_inode		*ip,
1703 	void				__user *arg)
1704 {
1705 	struct xfs_scrub_metadata	scrub;
1706 	int				error;
1707 
1708 	if (!capable(CAP_SYS_ADMIN))
1709 		return -EPERM;
1710 
1711 	if (copy_from_user(&scrub, arg, sizeof(scrub)))
1712 		return -EFAULT;
1713 
1714 	error = xfs_scrub_metadata(ip, &scrub);
1715 	if (error)
1716 		return error;
1717 
1718 	if (copy_to_user(arg, &scrub, sizeof(scrub)))
1719 		return -EFAULT;
1720 
1721 	return 0;
1722 }
1723 
1724 int
1725 xfs_ioc_swapext(
1726 	xfs_swapext_t	*sxp)
1727 {
1728 	xfs_inode_t     *ip, *tip;
1729 	struct fd	f, tmp;
1730 	int		error = 0;
1731 
1732 	/* Pull information for the target fd */
1733 	f = fdget((int)sxp->sx_fdtarget);
1734 	if (!f.file) {
1735 		error = -EINVAL;
1736 		goto out;
1737 	}
1738 
1739 	if (!(f.file->f_mode & FMODE_WRITE) ||
1740 	    !(f.file->f_mode & FMODE_READ) ||
1741 	    (f.file->f_flags & O_APPEND)) {
1742 		error = -EBADF;
1743 		goto out_put_file;
1744 	}
1745 
1746 	tmp = fdget((int)sxp->sx_fdtmp);
1747 	if (!tmp.file) {
1748 		error = -EINVAL;
1749 		goto out_put_file;
1750 	}
1751 
1752 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1753 	    !(tmp.file->f_mode & FMODE_READ) ||
1754 	    (tmp.file->f_flags & O_APPEND)) {
1755 		error = -EBADF;
1756 		goto out_put_tmp_file;
1757 	}
1758 
1759 	if (IS_SWAPFILE(file_inode(f.file)) ||
1760 	    IS_SWAPFILE(file_inode(tmp.file))) {
1761 		error = -EINVAL;
1762 		goto out_put_tmp_file;
1763 	}
1764 
1765 	/*
1766 	 * We need to ensure that the fds passed in point to XFS inodes
1767 	 * before we cast and access them as XFS structures as we have no
1768 	 * control over what the user passes us here.
1769 	 */
1770 	if (f.file->f_op != &xfs_file_operations ||
1771 	    tmp.file->f_op != &xfs_file_operations) {
1772 		error = -EINVAL;
1773 		goto out_put_tmp_file;
1774 	}
1775 
1776 	ip = XFS_I(file_inode(f.file));
1777 	tip = XFS_I(file_inode(tmp.file));
1778 
1779 	if (ip->i_mount != tip->i_mount) {
1780 		error = -EINVAL;
1781 		goto out_put_tmp_file;
1782 	}
1783 
1784 	if (ip->i_ino == tip->i_ino) {
1785 		error = -EINVAL;
1786 		goto out_put_tmp_file;
1787 	}
1788 
1789 	if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1790 		error = -EIO;
1791 		goto out_put_tmp_file;
1792 	}
1793 
1794 	error = xfs_swap_extents(ip, tip, sxp);
1795 
1796  out_put_tmp_file:
1797 	fdput(tmp);
1798  out_put_file:
1799 	fdput(f);
1800  out:
1801 	return error;
1802 }
1803 
1804 static int
1805 xfs_ioc_getlabel(
1806 	struct xfs_mount	*mp,
1807 	char			__user *user_label)
1808 {
1809 	struct xfs_sb		*sbp = &mp->m_sb;
1810 	char			label[XFSLABEL_MAX + 1];
1811 
1812 	/* Paranoia */
1813 	BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1814 
1815 	/* 1 larger than sb_fname, so this ensures a trailing NUL char */
1816 	memset(label, 0, sizeof(label));
1817 	spin_lock(&mp->m_sb_lock);
1818 	strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1819 	spin_unlock(&mp->m_sb_lock);
1820 
1821 	if (copy_to_user(user_label, label, sizeof(label)))
1822 		return -EFAULT;
1823 	return 0;
1824 }
1825 
1826 static int
1827 xfs_ioc_setlabel(
1828 	struct file		*filp,
1829 	struct xfs_mount	*mp,
1830 	char			__user *newlabel)
1831 {
1832 	struct xfs_sb		*sbp = &mp->m_sb;
1833 	char			label[XFSLABEL_MAX + 1];
1834 	size_t			len;
1835 	int			error;
1836 
1837 	if (!capable(CAP_SYS_ADMIN))
1838 		return -EPERM;
1839 	/*
1840 	 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1841 	 * smaller, at 12 bytes.  We copy one more to be sure we find the
1842 	 * (required) NULL character to test the incoming label length.
1843 	 * NB: The on disk label doesn't need to be null terminated.
1844 	 */
1845 	if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1846 		return -EFAULT;
1847 	len = strnlen(label, XFSLABEL_MAX + 1);
1848 	if (len > sizeof(sbp->sb_fname))
1849 		return -EINVAL;
1850 
1851 	error = mnt_want_write_file(filp);
1852 	if (error)
1853 		return error;
1854 
1855 	spin_lock(&mp->m_sb_lock);
1856 	memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1857 	memcpy(sbp->sb_fname, label, len);
1858 	spin_unlock(&mp->m_sb_lock);
1859 
1860 	/*
1861 	 * Now we do several things to satisfy userspace.
1862 	 * In addition to normal logging of the primary superblock, we also
1863 	 * immediately write these changes to sector zero for the primary, then
1864 	 * update all backup supers (as xfs_db does for a label change), then
1865 	 * invalidate the block device page cache.  This is so that any prior
1866 	 * buffered reads from userspace (i.e. from blkid) are invalidated,
1867 	 * and userspace will see the newly-written label.
1868 	 */
1869 	error = xfs_sync_sb_buf(mp);
1870 	if (error)
1871 		goto out;
1872 	/*
1873 	 * growfs also updates backup supers so lock against that.
1874 	 */
1875 	mutex_lock(&mp->m_growlock);
1876 	error = xfs_update_secondary_sbs(mp);
1877 	mutex_unlock(&mp->m_growlock);
1878 
1879 	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1880 
1881 out:
1882 	mnt_drop_write_file(filp);
1883 	return error;
1884 }
1885 
1886 /*
1887  * Note: some of the ioctl's return positive numbers as a
1888  * byte count indicating success, such as readlink_by_handle.
1889  * So we don't "sign flip" like most other routines.  This means
1890  * true errors need to be returned as a negative value.
1891  */
1892 long
1893 xfs_file_ioctl(
1894 	struct file		*filp,
1895 	unsigned int		cmd,
1896 	unsigned long		p)
1897 {
1898 	struct inode		*inode = file_inode(filp);
1899 	struct xfs_inode	*ip = XFS_I(inode);
1900 	struct xfs_mount	*mp = ip->i_mount;
1901 	void			__user *arg = (void __user *)p;
1902 	int			error;
1903 
1904 	trace_xfs_file_ioctl(ip);
1905 
1906 	switch (cmd) {
1907 	case FITRIM:
1908 		return xfs_ioc_trim(mp, arg);
1909 	case FS_IOC_GETFSLABEL:
1910 		return xfs_ioc_getlabel(mp, arg);
1911 	case FS_IOC_SETFSLABEL:
1912 		return xfs_ioc_setlabel(filp, mp, arg);
1913 	case XFS_IOC_ALLOCSP:
1914 	case XFS_IOC_FREESP:
1915 	case XFS_IOC_RESVSP:
1916 	case XFS_IOC_UNRESVSP:
1917 	case XFS_IOC_ALLOCSP64:
1918 	case XFS_IOC_FREESP64:
1919 	case XFS_IOC_RESVSP64:
1920 	case XFS_IOC_UNRESVSP64:
1921 	case XFS_IOC_ZERO_RANGE: {
1922 		xfs_flock64_t		bf;
1923 
1924 		if (copy_from_user(&bf, arg, sizeof(bf)))
1925 			return -EFAULT;
1926 		return xfs_ioc_space(filp, cmd, &bf);
1927 	}
1928 	case XFS_IOC_DIOINFO: {
1929 		struct dioattr	da;
1930 		xfs_buftarg_t	*target =
1931 			XFS_IS_REALTIME_INODE(ip) ?
1932 			mp->m_rtdev_targp : mp->m_ddev_targp;
1933 
1934 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1935 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1936 
1937 		if (copy_to_user(arg, &da, sizeof(da)))
1938 			return -EFAULT;
1939 		return 0;
1940 	}
1941 
1942 	case XFS_IOC_FSBULKSTAT_SINGLE:
1943 	case XFS_IOC_FSBULKSTAT:
1944 	case XFS_IOC_FSINUMBERS:
1945 		return xfs_ioc_bulkstat(mp, cmd, arg);
1946 
1947 	case XFS_IOC_FSGEOMETRY_V1:
1948 		return xfs_ioc_fsgeometry(mp, arg, 3);
1949 	case XFS_IOC_FSGEOMETRY_V4:
1950 		return xfs_ioc_fsgeometry(mp, arg, 4);
1951 	case XFS_IOC_FSGEOMETRY:
1952 		return xfs_ioc_fsgeometry(mp, arg, 5);
1953 
1954 	case XFS_IOC_AG_GEOMETRY:
1955 		return xfs_ioc_ag_geometry(mp, arg);
1956 
1957 	case XFS_IOC_GETVERSION:
1958 		return put_user(inode->i_generation, (int __user *)arg);
1959 
1960 	case XFS_IOC_FSGETXATTR:
1961 		return xfs_ioc_fsgetxattr(ip, 0, arg);
1962 	case XFS_IOC_FSGETXATTRA:
1963 		return xfs_ioc_fsgetxattr(ip, 1, arg);
1964 	case XFS_IOC_FSSETXATTR:
1965 		return xfs_ioc_fssetxattr(ip, filp, arg);
1966 	case XFS_IOC_GETXFLAGS:
1967 		return xfs_ioc_getxflags(ip, arg);
1968 	case XFS_IOC_SETXFLAGS:
1969 		return xfs_ioc_setxflags(ip, filp, arg);
1970 
1971 	case XFS_IOC_FSSETDM: {
1972 		struct fsdmidata	dmi;
1973 
1974 		if (copy_from_user(&dmi, arg, sizeof(dmi)))
1975 			return -EFAULT;
1976 
1977 		error = mnt_want_write_file(filp);
1978 		if (error)
1979 			return error;
1980 
1981 		error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1982 				dmi.fsd_dmstate);
1983 		mnt_drop_write_file(filp);
1984 		return error;
1985 	}
1986 
1987 	case XFS_IOC_GETBMAP:
1988 	case XFS_IOC_GETBMAPA:
1989 	case XFS_IOC_GETBMAPX:
1990 		return xfs_ioc_getbmap(filp, cmd, arg);
1991 
1992 	case FS_IOC_GETFSMAP:
1993 		return xfs_ioc_getfsmap(ip, arg);
1994 
1995 	case XFS_IOC_SCRUB_METADATA:
1996 		return xfs_ioc_scrub_metadata(ip, arg);
1997 
1998 	case XFS_IOC_FD_TO_HANDLE:
1999 	case XFS_IOC_PATH_TO_HANDLE:
2000 	case XFS_IOC_PATH_TO_FSHANDLE: {
2001 		xfs_fsop_handlereq_t	hreq;
2002 
2003 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
2004 			return -EFAULT;
2005 		return xfs_find_handle(cmd, &hreq);
2006 	}
2007 	case XFS_IOC_OPEN_BY_HANDLE: {
2008 		xfs_fsop_handlereq_t	hreq;
2009 
2010 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2011 			return -EFAULT;
2012 		return xfs_open_by_handle(filp, &hreq);
2013 	}
2014 	case XFS_IOC_FSSETDM_BY_HANDLE:
2015 		return xfs_fssetdm_by_handle(filp, arg);
2016 
2017 	case XFS_IOC_READLINK_BY_HANDLE: {
2018 		xfs_fsop_handlereq_t	hreq;
2019 
2020 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2021 			return -EFAULT;
2022 		return xfs_readlink_by_handle(filp, &hreq);
2023 	}
2024 	case XFS_IOC_ATTRLIST_BY_HANDLE:
2025 		return xfs_attrlist_by_handle(filp, arg);
2026 
2027 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
2028 		return xfs_attrmulti_by_handle(filp, arg);
2029 
2030 	case XFS_IOC_SWAPEXT: {
2031 		struct xfs_swapext	sxp;
2032 
2033 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
2034 			return -EFAULT;
2035 		error = mnt_want_write_file(filp);
2036 		if (error)
2037 			return error;
2038 		error = xfs_ioc_swapext(&sxp);
2039 		mnt_drop_write_file(filp);
2040 		return error;
2041 	}
2042 
2043 	case XFS_IOC_FSCOUNTS: {
2044 		xfs_fsop_counts_t out;
2045 
2046 		xfs_fs_counts(mp, &out);
2047 
2048 		if (copy_to_user(arg, &out, sizeof(out)))
2049 			return -EFAULT;
2050 		return 0;
2051 	}
2052 
2053 	case XFS_IOC_SET_RESBLKS: {
2054 		xfs_fsop_resblks_t inout;
2055 		uint64_t	   in;
2056 
2057 		if (!capable(CAP_SYS_ADMIN))
2058 			return -EPERM;
2059 
2060 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2061 			return -EROFS;
2062 
2063 		if (copy_from_user(&inout, arg, sizeof(inout)))
2064 			return -EFAULT;
2065 
2066 		error = mnt_want_write_file(filp);
2067 		if (error)
2068 			return error;
2069 
2070 		/* input parameter is passed in resblks field of structure */
2071 		in = inout.resblks;
2072 		error = xfs_reserve_blocks(mp, &in, &inout);
2073 		mnt_drop_write_file(filp);
2074 		if (error)
2075 			return error;
2076 
2077 		if (copy_to_user(arg, &inout, sizeof(inout)))
2078 			return -EFAULT;
2079 		return 0;
2080 	}
2081 
2082 	case XFS_IOC_GET_RESBLKS: {
2083 		xfs_fsop_resblks_t out;
2084 
2085 		if (!capable(CAP_SYS_ADMIN))
2086 			return -EPERM;
2087 
2088 		error = xfs_reserve_blocks(mp, NULL, &out);
2089 		if (error)
2090 			return error;
2091 
2092 		if (copy_to_user(arg, &out, sizeof(out)))
2093 			return -EFAULT;
2094 
2095 		return 0;
2096 	}
2097 
2098 	case XFS_IOC_FSGROWFSDATA: {
2099 		xfs_growfs_data_t in;
2100 
2101 		if (copy_from_user(&in, arg, sizeof(in)))
2102 			return -EFAULT;
2103 
2104 		error = mnt_want_write_file(filp);
2105 		if (error)
2106 			return error;
2107 		error = xfs_growfs_data(mp, &in);
2108 		mnt_drop_write_file(filp);
2109 		return error;
2110 	}
2111 
2112 	case XFS_IOC_FSGROWFSLOG: {
2113 		xfs_growfs_log_t in;
2114 
2115 		if (copy_from_user(&in, arg, sizeof(in)))
2116 			return -EFAULT;
2117 
2118 		error = mnt_want_write_file(filp);
2119 		if (error)
2120 			return error;
2121 		error = xfs_growfs_log(mp, &in);
2122 		mnt_drop_write_file(filp);
2123 		return error;
2124 	}
2125 
2126 	case XFS_IOC_FSGROWFSRT: {
2127 		xfs_growfs_rt_t in;
2128 
2129 		if (copy_from_user(&in, arg, sizeof(in)))
2130 			return -EFAULT;
2131 
2132 		error = mnt_want_write_file(filp);
2133 		if (error)
2134 			return error;
2135 		error = xfs_growfs_rt(mp, &in);
2136 		mnt_drop_write_file(filp);
2137 		return error;
2138 	}
2139 
2140 	case XFS_IOC_GOINGDOWN: {
2141 		uint32_t in;
2142 
2143 		if (!capable(CAP_SYS_ADMIN))
2144 			return -EPERM;
2145 
2146 		if (get_user(in, (uint32_t __user *)arg))
2147 			return -EFAULT;
2148 
2149 		return xfs_fs_goingdown(mp, in);
2150 	}
2151 
2152 	case XFS_IOC_ERROR_INJECTION: {
2153 		xfs_error_injection_t in;
2154 
2155 		if (!capable(CAP_SYS_ADMIN))
2156 			return -EPERM;
2157 
2158 		if (copy_from_user(&in, arg, sizeof(in)))
2159 			return -EFAULT;
2160 
2161 		return xfs_errortag_add(mp, in.errtag);
2162 	}
2163 
2164 	case XFS_IOC_ERROR_CLEARALL:
2165 		if (!capable(CAP_SYS_ADMIN))
2166 			return -EPERM;
2167 
2168 		return xfs_errortag_clearall(mp);
2169 
2170 	case XFS_IOC_FREE_EOFBLOCKS: {
2171 		struct xfs_fs_eofblocks eofb;
2172 		struct xfs_eofblocks keofb;
2173 
2174 		if (!capable(CAP_SYS_ADMIN))
2175 			return -EPERM;
2176 
2177 		if (mp->m_flags & XFS_MOUNT_RDONLY)
2178 			return -EROFS;
2179 
2180 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
2181 			return -EFAULT;
2182 
2183 		error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2184 		if (error)
2185 			return error;
2186 
2187 		return xfs_icache_free_eofblocks(mp, &keofb);
2188 	}
2189 
2190 	default:
2191 		return -ENOTTY;
2192 	}
2193 }
2194