xref: /linux/fs/xfs/xfs_ioctl.c (revision b615b9c36cae0468491547206406a909a9a37f26)
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_rtalloc.h"
15 #include "xfs_iwalk.h"
16 #include "xfs_itable.h"
17 #include "xfs_error.h"
18 #include "xfs_da_format.h"
19 #include "xfs_da_btree.h"
20 #include "xfs_attr.h"
21 #include "xfs_bmap.h"
22 #include "xfs_bmap_util.h"
23 #include "xfs_fsops.h"
24 #include "xfs_discard.h"
25 #include "xfs_quota.h"
26 #include "xfs_trace.h"
27 #include "xfs_icache.h"
28 #include "xfs_trans.h"
29 #include "xfs_btree.h"
30 #include <linux/fsmap.h>
31 #include "xfs_fsmap.h"
32 #include "scrub/xfs_scrub.h"
33 #include "xfs_sb.h"
34 #include "xfs_ag.h"
35 #include "xfs_health.h"
36 #include "xfs_reflink.h"
37 #include "xfs_ioctl.h"
38 #include "xfs_xattr.h"
39 #include "xfs_rtbitmap.h"
40 #include "xfs_file.h"
41 #include "xfs_exchrange.h"
42 #include "xfs_handle.h"
43 
44 #include <linux/mount.h>
45 #include <linux/fileattr.h>
46 
47 /* Return 0 on success or positive error */
48 int
49 xfs_fsbulkstat_one_fmt(
50 	struct xfs_ibulk		*breq,
51 	const struct xfs_bulkstat	*bstat)
52 {
53 	struct xfs_bstat		bs1;
54 
55 	xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
56 	if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
57 		return -EFAULT;
58 	return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
59 }
60 
61 int
62 xfs_fsinumbers_fmt(
63 	struct xfs_ibulk		*breq,
64 	const struct xfs_inumbers	*igrp)
65 {
66 	struct xfs_inogrp		ig1;
67 
68 	xfs_inumbers_to_inogrp(&ig1, igrp);
69 	if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
70 		return -EFAULT;
71 	return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
72 }
73 
74 STATIC int
75 xfs_ioc_fsbulkstat(
76 	struct file		*file,
77 	unsigned int		cmd,
78 	void			__user *arg)
79 {
80 	struct xfs_mount	*mp = XFS_I(file_inode(file))->i_mount;
81 	struct xfs_fsop_bulkreq	bulkreq;
82 	struct xfs_ibulk	breq = {
83 		.mp		= mp,
84 		.idmap		= file_mnt_idmap(file),
85 		.ocount		= 0,
86 	};
87 	xfs_ino_t		lastino;
88 	int			error;
89 
90 	/* done = 1 if there are more stats to get and if bulkstat */
91 	/* should be called again (unused here, but used in dmapi) */
92 
93 	if (!capable(CAP_SYS_ADMIN))
94 		return -EPERM;
95 
96 	if (xfs_is_shutdown(mp))
97 		return -EIO;
98 
99 	if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
100 		return -EFAULT;
101 
102 	if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
103 		return -EFAULT;
104 
105 	if (bulkreq.icount <= 0)
106 		return -EINVAL;
107 
108 	if (bulkreq.ubuffer == NULL)
109 		return -EINVAL;
110 
111 	breq.ubuffer = bulkreq.ubuffer;
112 	breq.icount = bulkreq.icount;
113 
114 	/*
115 	 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
116 	 * that we want to stat.  However, FSINUMBERS and FSBULKSTAT expect
117 	 * that *lastip contains either zero or the number of the last inode to
118 	 * be examined by the previous call and return results starting with
119 	 * the next inode after that.  The new bulk request back end functions
120 	 * take the inode to start with, so we have to compute the startino
121 	 * parameter from lastino to maintain correct function.  lastino == 0
122 	 * is a special case because it has traditionally meant "first inode
123 	 * in filesystem".
124 	 */
125 	if (cmd == XFS_IOC_FSINUMBERS) {
126 		breq.startino = lastino ? lastino + 1 : 0;
127 		error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
128 		lastino = breq.startino - 1;
129 	} else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
130 		breq.startino = lastino;
131 		breq.icount = 1;
132 		error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
133 	} else {	/* XFS_IOC_FSBULKSTAT */
134 		breq.startino = lastino ? lastino + 1 : 0;
135 		error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
136 		lastino = breq.startino - 1;
137 	}
138 
139 	if (error)
140 		return error;
141 
142 	if (bulkreq.lastip != NULL &&
143 	    copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
144 		return -EFAULT;
145 
146 	if (bulkreq.ocount != NULL &&
147 	    copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
148 		return -EFAULT;
149 
150 	return 0;
151 }
152 
153 /* Return 0 on success or positive error */
154 static int
155 xfs_bulkstat_fmt(
156 	struct xfs_ibulk		*breq,
157 	const struct xfs_bulkstat	*bstat)
158 {
159 	if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
160 		return -EFAULT;
161 	return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
162 }
163 
164 /*
165  * Check the incoming bulk request @hdr from userspace and initialize the
166  * internal @breq bulk request appropriately.  Returns 0 if the bulk request
167  * should proceed; -ECANCELED if there's nothing to do; or the usual
168  * negative error code.
169  */
170 static int
171 xfs_bulk_ireq_setup(
172 	struct xfs_mount	*mp,
173 	const struct xfs_bulk_ireq *hdr,
174 	struct xfs_ibulk	*breq,
175 	void __user		*ubuffer)
176 {
177 	if (hdr->icount == 0 ||
178 	    (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
179 	    memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
180 		return -EINVAL;
181 
182 	breq->startino = hdr->ino;
183 	breq->ubuffer = ubuffer;
184 	breq->icount = hdr->icount;
185 	breq->ocount = 0;
186 	breq->flags = 0;
187 
188 	/*
189 	 * The @ino parameter is a special value, so we must look it up here.
190 	 * We're not allowed to have IREQ_AGNO, and we only return one inode
191 	 * worth of data.
192 	 */
193 	if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
194 		if (hdr->flags & XFS_BULK_IREQ_AGNO)
195 			return -EINVAL;
196 
197 		switch (hdr->ino) {
198 		case XFS_BULK_IREQ_SPECIAL_ROOT:
199 			breq->startino = mp->m_sb.sb_rootino;
200 			break;
201 		default:
202 			return -EINVAL;
203 		}
204 		breq->icount = 1;
205 	}
206 
207 	/*
208 	 * The IREQ_AGNO flag means that we only want results from a given AG.
209 	 * If @hdr->ino is zero, we start iterating in that AG.  If @hdr->ino is
210 	 * beyond the specified AG then we return no results.
211 	 */
212 	if (hdr->flags & XFS_BULK_IREQ_AGNO) {
213 		if (hdr->agno >= mp->m_sb.sb_agcount)
214 			return -EINVAL;
215 
216 		if (breq->startino == 0)
217 			breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
218 		else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
219 			return -EINVAL;
220 
221 		breq->flags |= XFS_IBULK_SAME_AG;
222 
223 		/* Asking for an inode past the end of the AG?  We're done! */
224 		if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
225 			return -ECANCELED;
226 	} else if (hdr->agno)
227 		return -EINVAL;
228 
229 	/* Asking for an inode past the end of the FS?  We're done! */
230 	if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
231 		return -ECANCELED;
232 
233 	if (hdr->flags & XFS_BULK_IREQ_NREXT64)
234 		breq->flags |= XFS_IBULK_NREXT64;
235 
236 	return 0;
237 }
238 
239 /*
240  * Update the userspace bulk request @hdr to reflect the end state of the
241  * internal bulk request @breq.
242  */
243 static void
244 xfs_bulk_ireq_teardown(
245 	struct xfs_bulk_ireq	*hdr,
246 	struct xfs_ibulk	*breq)
247 {
248 	hdr->ino = breq->startino;
249 	hdr->ocount = breq->ocount;
250 }
251 
252 /* Handle the v5 bulkstat ioctl. */
253 STATIC int
254 xfs_ioc_bulkstat(
255 	struct file			*file,
256 	unsigned int			cmd,
257 	struct xfs_bulkstat_req __user	*arg)
258 {
259 	struct xfs_mount		*mp = XFS_I(file_inode(file))->i_mount;
260 	struct xfs_bulk_ireq		hdr;
261 	struct xfs_ibulk		breq = {
262 		.mp			= mp,
263 		.idmap			= file_mnt_idmap(file),
264 	};
265 	int				error;
266 
267 	if (!capable(CAP_SYS_ADMIN))
268 		return -EPERM;
269 
270 	if (xfs_is_shutdown(mp))
271 		return -EIO;
272 
273 	if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
274 		return -EFAULT;
275 
276 	error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
277 	if (error == -ECANCELED)
278 		goto out_teardown;
279 	if (error < 0)
280 		return error;
281 
282 	error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
283 	if (error)
284 		return error;
285 
286 out_teardown:
287 	xfs_bulk_ireq_teardown(&hdr, &breq);
288 	if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
289 		return -EFAULT;
290 
291 	return 0;
292 }
293 
294 STATIC int
295 xfs_inumbers_fmt(
296 	struct xfs_ibulk		*breq,
297 	const struct xfs_inumbers	*igrp)
298 {
299 	if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
300 		return -EFAULT;
301 	return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
302 }
303 
304 /* Handle the v5 inumbers ioctl. */
305 STATIC int
306 xfs_ioc_inumbers(
307 	struct xfs_mount		*mp,
308 	unsigned int			cmd,
309 	struct xfs_inumbers_req __user	*arg)
310 {
311 	struct xfs_bulk_ireq		hdr;
312 	struct xfs_ibulk		breq = {
313 		.mp			= mp,
314 	};
315 	int				error;
316 
317 	if (!capable(CAP_SYS_ADMIN))
318 		return -EPERM;
319 
320 	if (xfs_is_shutdown(mp))
321 		return -EIO;
322 
323 	if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
324 		return -EFAULT;
325 
326 	error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
327 	if (error == -ECANCELED)
328 		goto out_teardown;
329 	if (error < 0)
330 		return error;
331 
332 	error = xfs_inumbers(&breq, xfs_inumbers_fmt);
333 	if (error)
334 		return error;
335 
336 out_teardown:
337 	xfs_bulk_ireq_teardown(&hdr, &breq);
338 	if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
339 		return -EFAULT;
340 
341 	return 0;
342 }
343 
344 STATIC int
345 xfs_ioc_fsgeometry(
346 	struct xfs_mount	*mp,
347 	void			__user *arg,
348 	int			struct_version)
349 {
350 	struct xfs_fsop_geom	fsgeo;
351 	size_t			len;
352 
353 	xfs_fs_geometry(mp, &fsgeo, struct_version);
354 
355 	if (struct_version <= 3)
356 		len = sizeof(struct xfs_fsop_geom_v1);
357 	else if (struct_version == 4)
358 		len = sizeof(struct xfs_fsop_geom_v4);
359 	else {
360 		xfs_fsop_geom_health(mp, &fsgeo);
361 		len = sizeof(fsgeo);
362 	}
363 
364 	if (copy_to_user(arg, &fsgeo, len))
365 		return -EFAULT;
366 	return 0;
367 }
368 
369 STATIC int
370 xfs_ioc_ag_geometry(
371 	struct xfs_mount	*mp,
372 	void			__user *arg)
373 {
374 	struct xfs_perag	*pag;
375 	struct xfs_ag_geometry	ageo;
376 	int			error;
377 
378 	if (copy_from_user(&ageo, arg, sizeof(ageo)))
379 		return -EFAULT;
380 	if (ageo.ag_flags)
381 		return -EINVAL;
382 	if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
383 		return -EINVAL;
384 
385 	pag = xfs_perag_get(mp, ageo.ag_number);
386 	if (!pag)
387 		return -EINVAL;
388 
389 	error = xfs_ag_get_geometry(pag, &ageo);
390 	xfs_perag_put(pag);
391 	if (error)
392 		return error;
393 
394 	if (copy_to_user(arg, &ageo, sizeof(ageo)))
395 		return -EFAULT;
396 	return 0;
397 }
398 
399 /*
400  * Linux extended inode flags interface.
401  */
402 
403 static void
404 xfs_fill_fsxattr(
405 	struct xfs_inode	*ip,
406 	int			whichfork,
407 	struct fileattr		*fa)
408 {
409 	struct xfs_mount	*mp = ip->i_mount;
410 	struct xfs_ifork	*ifp = xfs_ifork_ptr(ip, whichfork);
411 
412 	fileattr_fill_xflags(fa, xfs_ip2xflags(ip));
413 
414 	if (ip->i_diflags & XFS_DIFLAG_EXTSIZE) {
415 		fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
416 	} else if (ip->i_diflags & XFS_DIFLAG_EXTSZINHERIT) {
417 		/*
418 		 * Don't let a misaligned extent size hint on a directory
419 		 * escape to userspace if it won't pass the setattr checks
420 		 * later.
421 		 */
422 		if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
423 		    xfs_extlen_to_rtxmod(mp, ip->i_extsize) > 0) {
424 			fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE |
425 					    FS_XFLAG_EXTSZINHERIT);
426 			fa->fsx_extsize = 0;
427 		} else {
428 			fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
429 		}
430 	}
431 
432 	if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
433 		fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
434 	fa->fsx_projid = ip->i_projid;
435 	if (ifp && !xfs_need_iread_extents(ifp))
436 		fa->fsx_nextents = xfs_iext_count(ifp);
437 	else
438 		fa->fsx_nextents = xfs_ifork_nextents(ifp);
439 }
440 
441 STATIC int
442 xfs_ioc_fsgetxattra(
443 	xfs_inode_t		*ip,
444 	void			__user *arg)
445 {
446 	struct fileattr		fa;
447 
448 	xfs_ilock(ip, XFS_ILOCK_SHARED);
449 	xfs_fill_fsxattr(ip, XFS_ATTR_FORK, &fa);
450 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
451 
452 	return copy_fsxattr_to_user(&fa, arg);
453 }
454 
455 int
456 xfs_fileattr_get(
457 	struct dentry		*dentry,
458 	struct fileattr		*fa)
459 {
460 	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
461 
462 	if (d_is_special(dentry))
463 		return -ENOTTY;
464 
465 	xfs_ilock(ip, XFS_ILOCK_SHARED);
466 	xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
467 	xfs_iunlock(ip, XFS_ILOCK_SHARED);
468 
469 	return 0;
470 }
471 
472 static int
473 xfs_ioctl_setattr_xflags(
474 	struct xfs_trans	*tp,
475 	struct xfs_inode	*ip,
476 	struct fileattr		*fa)
477 {
478 	struct xfs_mount	*mp = ip->i_mount;
479 	bool			rtflag = (fa->fsx_xflags & FS_XFLAG_REALTIME);
480 	uint64_t		i_flags2;
481 
482 	if (rtflag != XFS_IS_REALTIME_INODE(ip)) {
483 		/* Can't change realtime flag if any extents are allocated. */
484 		if (ip->i_df.if_nextents || ip->i_delayed_blks)
485 			return -EINVAL;
486 
487 		/*
488 		 * If S_DAX is enabled on this file, we can only switch the
489 		 * device if both support fsdax.  We can't update S_DAX because
490 		 * there might be other threads walking down the access paths.
491 		 */
492 		if (IS_DAX(VFS_I(ip)) &&
493 		    (mp->m_ddev_targp->bt_daxdev == NULL ||
494 		     (mp->m_rtdev_targp &&
495 		      mp->m_rtdev_targp->bt_daxdev == NULL)))
496 			return -EINVAL;
497 	}
498 
499 	if (rtflag) {
500 		/* If realtime flag is set then must have realtime device */
501 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
502 		    xfs_extlen_to_rtxmod(mp, ip->i_extsize))
503 			return -EINVAL;
504 
505 		/* Clear reflink if we are actually able to set the rt flag. */
506 		if (xfs_is_reflink_inode(ip))
507 			ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
508 	}
509 
510 	/* diflags2 only valid for v3 inodes. */
511 	i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
512 	if (i_flags2 && !xfs_has_v3inodes(mp))
513 		return -EINVAL;
514 
515 	ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
516 	ip->i_diflags2 = i_flags2;
517 
518 	xfs_diflags_to_iflags(ip, false);
519 
520 	/*
521 	 * Make the stable writes flag match that of the device the inode
522 	 * resides on when flipping the RT flag.
523 	 */
524 	if (rtflag != XFS_IS_REALTIME_INODE(ip) && S_ISREG(VFS_I(ip)->i_mode))
525 		xfs_update_stable_writes(ip);
526 
527 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
528 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
529 	XFS_STATS_INC(mp, xs_ig_attrchg);
530 	return 0;
531 }
532 
533 static void
534 xfs_ioctl_setattr_prepare_dax(
535 	struct xfs_inode	*ip,
536 	struct fileattr		*fa)
537 {
538 	struct xfs_mount	*mp = ip->i_mount;
539 	struct inode            *inode = VFS_I(ip);
540 
541 	if (S_ISDIR(inode->i_mode))
542 		return;
543 
544 	if (xfs_has_dax_always(mp) || xfs_has_dax_never(mp))
545 		return;
546 
547 	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
548 	    !(ip->i_diflags2 & XFS_DIFLAG2_DAX)) ||
549 	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
550 	     (ip->i_diflags2 & XFS_DIFLAG2_DAX)))
551 		d_mark_dontcache(inode);
552 }
553 
554 /*
555  * Set up the transaction structure for the setattr operation, checking that we
556  * have permission to do so. On success, return a clean transaction and the
557  * inode locked exclusively ready for further operation specific checks. On
558  * failure, return an error without modifying or locking the inode.
559  */
560 static struct xfs_trans *
561 xfs_ioctl_setattr_get_trans(
562 	struct xfs_inode	*ip,
563 	struct xfs_dquot	*pdqp)
564 {
565 	struct xfs_mount	*mp = ip->i_mount;
566 	struct xfs_trans	*tp;
567 	int			error = -EROFS;
568 
569 	if (xfs_is_readonly(mp))
570 		goto out_error;
571 	error = -EIO;
572 	if (xfs_is_shutdown(mp))
573 		goto out_error;
574 
575 	error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
576 			has_capability_noaudit(current, CAP_FOWNER), &tp);
577 	if (error)
578 		goto out_error;
579 
580 	if (xfs_has_wsync(mp))
581 		xfs_trans_set_sync(tp);
582 
583 	return tp;
584 
585 out_error:
586 	return ERR_PTR(error);
587 }
588 
589 /*
590  * Validate a proposed extent size hint.  For regular files, the hint can only
591  * be changed if no extents are allocated.
592  */
593 static int
594 xfs_ioctl_setattr_check_extsize(
595 	struct xfs_inode	*ip,
596 	struct fileattr		*fa)
597 {
598 	struct xfs_mount	*mp = ip->i_mount;
599 	xfs_failaddr_t		failaddr;
600 	uint16_t		new_diflags;
601 
602 	if (!fa->fsx_valid)
603 		return 0;
604 
605 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
606 	    XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
607 		return -EINVAL;
608 
609 	if (fa->fsx_extsize & mp->m_blockmask)
610 		return -EINVAL;
611 
612 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
613 
614 	/*
615 	 * Inode verifiers do not check that the extent size hint is an integer
616 	 * multiple of the rt extent size on a directory with both rtinherit
617 	 * and extszinherit flags set.  Don't let sysadmins misconfigure
618 	 * directories.
619 	 */
620 	if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
621 	    (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
622 		unsigned int	rtextsize_bytes;
623 
624 		rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
625 		if (fa->fsx_extsize % rtextsize_bytes)
626 			return -EINVAL;
627 	}
628 
629 	failaddr = xfs_inode_validate_extsize(ip->i_mount,
630 			XFS_B_TO_FSB(mp, fa->fsx_extsize),
631 			VFS_I(ip)->i_mode, new_diflags);
632 	return failaddr != NULL ? -EINVAL : 0;
633 }
634 
635 static int
636 xfs_ioctl_setattr_check_cowextsize(
637 	struct xfs_inode	*ip,
638 	struct fileattr		*fa)
639 {
640 	struct xfs_mount	*mp = ip->i_mount;
641 	xfs_failaddr_t		failaddr;
642 	uint64_t		new_diflags2;
643 	uint16_t		new_diflags;
644 
645 	if (!fa->fsx_valid)
646 		return 0;
647 
648 	if (fa->fsx_cowextsize & mp->m_blockmask)
649 		return -EINVAL;
650 
651 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
652 	new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
653 
654 	failaddr = xfs_inode_validate_cowextsize(ip->i_mount,
655 			XFS_B_TO_FSB(mp, fa->fsx_cowextsize),
656 			VFS_I(ip)->i_mode, new_diflags, new_diflags2);
657 	return failaddr != NULL ? -EINVAL : 0;
658 }
659 
660 static int
661 xfs_ioctl_setattr_check_projid(
662 	struct xfs_inode	*ip,
663 	struct fileattr		*fa)
664 {
665 	if (!fa->fsx_valid)
666 		return 0;
667 
668 	/* Disallow 32bit project ids if 32bit IDs are not enabled. */
669 	if (fa->fsx_projid > (uint16_t)-1 &&
670 	    !xfs_has_projid32(ip->i_mount))
671 		return -EINVAL;
672 	return 0;
673 }
674 
675 int
676 xfs_fileattr_set(
677 	struct mnt_idmap	*idmap,
678 	struct dentry		*dentry,
679 	struct fileattr		*fa)
680 {
681 	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
682 	struct xfs_mount	*mp = ip->i_mount;
683 	struct xfs_trans	*tp;
684 	struct xfs_dquot	*pdqp = NULL;
685 	struct xfs_dquot	*olddquot = NULL;
686 	int			error;
687 
688 	trace_xfs_ioctl_setattr(ip);
689 
690 	if (d_is_special(dentry))
691 		return -ENOTTY;
692 
693 	if (!fa->fsx_valid) {
694 		if (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL |
695 				  FS_NOATIME_FL | FS_NODUMP_FL |
696 				  FS_SYNC_FL | FS_DAX_FL | FS_PROJINHERIT_FL))
697 			return -EOPNOTSUPP;
698 	}
699 
700 	error = xfs_ioctl_setattr_check_projid(ip, fa);
701 	if (error)
702 		return error;
703 
704 	/*
705 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
706 	 * before we start any other transactions. Trying to do this later
707 	 * is messy. We don't care to take a readlock to look at the ids
708 	 * in inode here, because we can't hold it across the trans_reserve.
709 	 * If the IDs do change before we take the ilock, we're covered
710 	 * because the i_*dquot fields will get updated anyway.
711 	 */
712 	if (fa->fsx_valid && XFS_IS_QUOTA_ON(mp)) {
713 		error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
714 				VFS_I(ip)->i_gid, fa->fsx_projid,
715 				XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
716 		if (error)
717 			return error;
718 	}
719 
720 	xfs_ioctl_setattr_prepare_dax(ip, fa);
721 
722 	tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
723 	if (IS_ERR(tp)) {
724 		error = PTR_ERR(tp);
725 		goto error_free_dquots;
726 	}
727 
728 	error = xfs_ioctl_setattr_check_extsize(ip, fa);
729 	if (error)
730 		goto error_trans_cancel;
731 
732 	error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
733 	if (error)
734 		goto error_trans_cancel;
735 
736 	error = xfs_ioctl_setattr_xflags(tp, ip, fa);
737 	if (error)
738 		goto error_trans_cancel;
739 
740 	if (!fa->fsx_valid)
741 		goto skip_xattr;
742 	/*
743 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
744 	 * overrides the following restrictions:
745 	 *
746 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
747 	 * successful return from chown()
748 	 */
749 
750 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
751 	    !capable_wrt_inode_uidgid(idmap, VFS_I(ip), CAP_FSETID))
752 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
753 
754 	/* Change the ownerships and register project quota modifications */
755 	if (ip->i_projid != fa->fsx_projid) {
756 		if (XFS_IS_PQUOTA_ON(mp)) {
757 			olddquot = xfs_qm_vop_chown(tp, ip,
758 						&ip->i_pdquot, pdqp);
759 		}
760 		ip->i_projid = fa->fsx_projid;
761 	}
762 
763 	/*
764 	 * Only set the extent size hint if we've already determined that the
765 	 * extent size hint should be set on the inode. If no extent size flags
766 	 * are set on the inode then unconditionally clear the extent size hint.
767 	 */
768 	if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
769 		ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
770 	else
771 		ip->i_extsize = 0;
772 
773 	if (xfs_has_v3inodes(mp)) {
774 		if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
775 			ip->i_cowextsize = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
776 		else
777 			ip->i_cowextsize = 0;
778 	}
779 
780 skip_xattr:
781 	error = xfs_trans_commit(tp);
782 
783 	/*
784 	 * Release any dquot(s) the inode had kept before chown.
785 	 */
786 	xfs_qm_dqrele(olddquot);
787 	xfs_qm_dqrele(pdqp);
788 
789 	return error;
790 
791 error_trans_cancel:
792 	xfs_trans_cancel(tp);
793 error_free_dquots:
794 	xfs_qm_dqrele(pdqp);
795 	return error;
796 }
797 
798 static bool
799 xfs_getbmap_format(
800 	struct kgetbmap		*p,
801 	struct getbmapx __user	*u,
802 	size_t			recsize)
803 {
804 	if (put_user(p->bmv_offset, &u->bmv_offset) ||
805 	    put_user(p->bmv_block, &u->bmv_block) ||
806 	    put_user(p->bmv_length, &u->bmv_length) ||
807 	    put_user(0, &u->bmv_count) ||
808 	    put_user(0, &u->bmv_entries))
809 		return false;
810 	if (recsize < sizeof(struct getbmapx))
811 		return true;
812 	if (put_user(0, &u->bmv_iflags) ||
813 	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
814 	    put_user(0, &u->bmv_unused1) ||
815 	    put_user(0, &u->bmv_unused2))
816 		return false;
817 	return true;
818 }
819 
820 STATIC int
821 xfs_ioc_getbmap(
822 	struct file		*file,
823 	unsigned int		cmd,
824 	void			__user *arg)
825 {
826 	struct getbmapx		bmx = { 0 };
827 	struct kgetbmap		*buf;
828 	size_t			recsize;
829 	int			error, i;
830 
831 	switch (cmd) {
832 	case XFS_IOC_GETBMAPA:
833 		bmx.bmv_iflags = BMV_IF_ATTRFORK;
834 		fallthrough;
835 	case XFS_IOC_GETBMAP:
836 		/* struct getbmap is a strict subset of struct getbmapx. */
837 		recsize = sizeof(struct getbmap);
838 		break;
839 	case XFS_IOC_GETBMAPX:
840 		recsize = sizeof(struct getbmapx);
841 		break;
842 	default:
843 		return -EINVAL;
844 	}
845 
846 	if (copy_from_user(&bmx, arg, recsize))
847 		return -EFAULT;
848 
849 	if (bmx.bmv_count < 2)
850 		return -EINVAL;
851 	if (bmx.bmv_count >= INT_MAX / recsize)
852 		return -ENOMEM;
853 
854 	buf = kvcalloc(bmx.bmv_count, sizeof(*buf), GFP_KERNEL);
855 	if (!buf)
856 		return -ENOMEM;
857 
858 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
859 	if (error)
860 		goto out_free_buf;
861 
862 	error = -EFAULT;
863 	if (copy_to_user(arg, &bmx, recsize))
864 		goto out_free_buf;
865 	arg += recsize;
866 
867 	for (i = 0; i < bmx.bmv_entries; i++) {
868 		if (!xfs_getbmap_format(buf + i, arg, recsize))
869 			goto out_free_buf;
870 		arg += recsize;
871 	}
872 
873 	error = 0;
874 out_free_buf:
875 	kvfree(buf);
876 	return error;
877 }
878 
879 STATIC int
880 xfs_ioc_getfsmap(
881 	struct xfs_inode	*ip,
882 	struct fsmap_head	__user *arg)
883 {
884 	struct xfs_fsmap_head	xhead = {0};
885 	struct fsmap_head	head;
886 	struct fsmap		*recs;
887 	unsigned int		count;
888 	__u32			last_flags = 0;
889 	bool			done = false;
890 	int			error;
891 
892 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
893 		return -EFAULT;
894 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
895 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
896 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
897 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
898 		       sizeof(head.fmh_keys[1].fmr_reserved)))
899 		return -EINVAL;
900 
901 	/*
902 	 * Use an internal memory buffer so that we don't have to copy fsmap
903 	 * data to userspace while holding locks.  Start by trying to allocate
904 	 * up to 128k for the buffer, but fall back to a single page if needed.
905 	 */
906 	count = min_t(unsigned int, head.fmh_count,
907 			131072 / sizeof(struct fsmap));
908 	recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
909 	if (!recs) {
910 		count = min_t(unsigned int, head.fmh_count,
911 				PAGE_SIZE / sizeof(struct fsmap));
912 		recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
913 		if (!recs)
914 			return -ENOMEM;
915 	}
916 
917 	xhead.fmh_iflags = head.fmh_iflags;
918 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
919 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
920 
921 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
922 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
923 
924 	head.fmh_entries = 0;
925 	do {
926 		struct fsmap __user	*user_recs;
927 		struct fsmap		*last_rec;
928 
929 		user_recs = &arg->fmh_recs[head.fmh_entries];
930 		xhead.fmh_entries = 0;
931 		xhead.fmh_count = min_t(unsigned int, count,
932 					head.fmh_count - head.fmh_entries);
933 
934 		/* Run query, record how many entries we got. */
935 		error = xfs_getfsmap(ip->i_mount, &xhead, recs);
936 		switch (error) {
937 		case 0:
938 			/*
939 			 * There are no more records in the result set.  Copy
940 			 * whatever we got to userspace and break out.
941 			 */
942 			done = true;
943 			break;
944 		case -ECANCELED:
945 			/*
946 			 * The internal memory buffer is full.  Copy whatever
947 			 * records we got to userspace and go again if we have
948 			 * not yet filled the userspace buffer.
949 			 */
950 			error = 0;
951 			break;
952 		default:
953 			goto out_free;
954 		}
955 		head.fmh_entries += xhead.fmh_entries;
956 		head.fmh_oflags = xhead.fmh_oflags;
957 
958 		/*
959 		 * If the caller wanted a record count or there aren't any
960 		 * new records to return, we're done.
961 		 */
962 		if (head.fmh_count == 0 || xhead.fmh_entries == 0)
963 			break;
964 
965 		/* Copy all the records we got out to userspace. */
966 		if (copy_to_user(user_recs, recs,
967 				 xhead.fmh_entries * sizeof(struct fsmap))) {
968 			error = -EFAULT;
969 			goto out_free;
970 		}
971 
972 		/* Remember the last record flags we copied to userspace. */
973 		last_rec = &recs[xhead.fmh_entries - 1];
974 		last_flags = last_rec->fmr_flags;
975 
976 		/* Set up the low key for the next iteration. */
977 		xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
978 		trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
979 	} while (!done && head.fmh_entries < head.fmh_count);
980 
981 	/*
982 	 * If there are no more records in the query result set and we're not
983 	 * in counting mode, mark the last record returned with the LAST flag.
984 	 */
985 	if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
986 		struct fsmap __user	*user_rec;
987 
988 		last_flags |= FMR_OF_LAST;
989 		user_rec = &arg->fmh_recs[head.fmh_entries - 1];
990 
991 		if (copy_to_user(&user_rec->fmr_flags, &last_flags,
992 					sizeof(last_flags))) {
993 			error = -EFAULT;
994 			goto out_free;
995 		}
996 	}
997 
998 	/* copy back header */
999 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1000 		error = -EFAULT;
1001 		goto out_free;
1002 	}
1003 
1004 out_free:
1005 	kvfree(recs);
1006 	return error;
1007 }
1008 
1009 int
1010 xfs_ioc_swapext(
1011 	xfs_swapext_t	*sxp)
1012 {
1013 	xfs_inode_t     *ip, *tip;
1014 	struct fd	f, tmp;
1015 	int		error = 0;
1016 
1017 	/* Pull information for the target fd */
1018 	f = fdget((int)sxp->sx_fdtarget);
1019 	if (!f.file) {
1020 		error = -EINVAL;
1021 		goto out;
1022 	}
1023 
1024 	if (!(f.file->f_mode & FMODE_WRITE) ||
1025 	    !(f.file->f_mode & FMODE_READ) ||
1026 	    (f.file->f_flags & O_APPEND)) {
1027 		error = -EBADF;
1028 		goto out_put_file;
1029 	}
1030 
1031 	tmp = fdget((int)sxp->sx_fdtmp);
1032 	if (!tmp.file) {
1033 		error = -EINVAL;
1034 		goto out_put_file;
1035 	}
1036 
1037 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1038 	    !(tmp.file->f_mode & FMODE_READ) ||
1039 	    (tmp.file->f_flags & O_APPEND)) {
1040 		error = -EBADF;
1041 		goto out_put_tmp_file;
1042 	}
1043 
1044 	if (IS_SWAPFILE(file_inode(f.file)) ||
1045 	    IS_SWAPFILE(file_inode(tmp.file))) {
1046 		error = -EINVAL;
1047 		goto out_put_tmp_file;
1048 	}
1049 
1050 	/*
1051 	 * We need to ensure that the fds passed in point to XFS inodes
1052 	 * before we cast and access them as XFS structures as we have no
1053 	 * control over what the user passes us here.
1054 	 */
1055 	if (f.file->f_op != &xfs_file_operations ||
1056 	    tmp.file->f_op != &xfs_file_operations) {
1057 		error = -EINVAL;
1058 		goto out_put_tmp_file;
1059 	}
1060 
1061 	ip = XFS_I(file_inode(f.file));
1062 	tip = XFS_I(file_inode(tmp.file));
1063 
1064 	if (ip->i_mount != tip->i_mount) {
1065 		error = -EINVAL;
1066 		goto out_put_tmp_file;
1067 	}
1068 
1069 	if (ip->i_ino == tip->i_ino) {
1070 		error = -EINVAL;
1071 		goto out_put_tmp_file;
1072 	}
1073 
1074 	if (xfs_is_shutdown(ip->i_mount)) {
1075 		error = -EIO;
1076 		goto out_put_tmp_file;
1077 	}
1078 
1079 	error = xfs_swap_extents(ip, tip, sxp);
1080 
1081  out_put_tmp_file:
1082 	fdput(tmp);
1083  out_put_file:
1084 	fdput(f);
1085  out:
1086 	return error;
1087 }
1088 
1089 static int
1090 xfs_ioc_getlabel(
1091 	struct xfs_mount	*mp,
1092 	char			__user *user_label)
1093 {
1094 	struct xfs_sb		*sbp = &mp->m_sb;
1095 	char			label[XFSLABEL_MAX + 1];
1096 
1097 	/* Paranoia */
1098 	BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1099 
1100 	/* 1 larger than sb_fname, so this ensures a trailing NUL char */
1101 	memset(label, 0, sizeof(label));
1102 	spin_lock(&mp->m_sb_lock);
1103 	strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1104 	spin_unlock(&mp->m_sb_lock);
1105 
1106 	if (copy_to_user(user_label, label, sizeof(label)))
1107 		return -EFAULT;
1108 	return 0;
1109 }
1110 
1111 static int
1112 xfs_ioc_setlabel(
1113 	struct file		*filp,
1114 	struct xfs_mount	*mp,
1115 	char			__user *newlabel)
1116 {
1117 	struct xfs_sb		*sbp = &mp->m_sb;
1118 	char			label[XFSLABEL_MAX + 1];
1119 	size_t			len;
1120 	int			error;
1121 
1122 	if (!capable(CAP_SYS_ADMIN))
1123 		return -EPERM;
1124 	/*
1125 	 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1126 	 * smaller, at 12 bytes.  We copy one more to be sure we find the
1127 	 * (required) NULL character to test the incoming label length.
1128 	 * NB: The on disk label doesn't need to be null terminated.
1129 	 */
1130 	if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1131 		return -EFAULT;
1132 	len = strnlen(label, XFSLABEL_MAX + 1);
1133 	if (len > sizeof(sbp->sb_fname))
1134 		return -EINVAL;
1135 
1136 	error = mnt_want_write_file(filp);
1137 	if (error)
1138 		return error;
1139 
1140 	spin_lock(&mp->m_sb_lock);
1141 	memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1142 	memcpy(sbp->sb_fname, label, len);
1143 	spin_unlock(&mp->m_sb_lock);
1144 
1145 	/*
1146 	 * Now we do several things to satisfy userspace.
1147 	 * In addition to normal logging of the primary superblock, we also
1148 	 * immediately write these changes to sector zero for the primary, then
1149 	 * update all backup supers (as xfs_db does for a label change), then
1150 	 * invalidate the block device page cache.  This is so that any prior
1151 	 * buffered reads from userspace (i.e. from blkid) are invalidated,
1152 	 * and userspace will see the newly-written label.
1153 	 */
1154 	error = xfs_sync_sb_buf(mp);
1155 	if (error)
1156 		goto out;
1157 	/*
1158 	 * growfs also updates backup supers so lock against that.
1159 	 */
1160 	mutex_lock(&mp->m_growlock);
1161 	error = xfs_update_secondary_sbs(mp);
1162 	mutex_unlock(&mp->m_growlock);
1163 
1164 	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1165 
1166 out:
1167 	mnt_drop_write_file(filp);
1168 	return error;
1169 }
1170 
1171 static inline int
1172 xfs_fs_eofblocks_from_user(
1173 	struct xfs_fs_eofblocks		*src,
1174 	struct xfs_icwalk		*dst)
1175 {
1176 	if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1177 		return -EINVAL;
1178 
1179 	if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1180 		return -EINVAL;
1181 
1182 	if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1183 	    memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1184 		return -EINVAL;
1185 
1186 	dst->icw_flags = 0;
1187 	if (src->eof_flags & XFS_EOF_FLAGS_SYNC)
1188 		dst->icw_flags |= XFS_ICWALK_FLAG_SYNC;
1189 	if (src->eof_flags & XFS_EOF_FLAGS_UID)
1190 		dst->icw_flags |= XFS_ICWALK_FLAG_UID;
1191 	if (src->eof_flags & XFS_EOF_FLAGS_GID)
1192 		dst->icw_flags |= XFS_ICWALK_FLAG_GID;
1193 	if (src->eof_flags & XFS_EOF_FLAGS_PRID)
1194 		dst->icw_flags |= XFS_ICWALK_FLAG_PRID;
1195 	if (src->eof_flags & XFS_EOF_FLAGS_MINFILESIZE)
1196 		dst->icw_flags |= XFS_ICWALK_FLAG_MINFILESIZE;
1197 
1198 	dst->icw_prid = src->eof_prid;
1199 	dst->icw_min_file_size = src->eof_min_file_size;
1200 
1201 	dst->icw_uid = INVALID_UID;
1202 	if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1203 		dst->icw_uid = make_kuid(current_user_ns(), src->eof_uid);
1204 		if (!uid_valid(dst->icw_uid))
1205 			return -EINVAL;
1206 	}
1207 
1208 	dst->icw_gid = INVALID_GID;
1209 	if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1210 		dst->icw_gid = make_kgid(current_user_ns(), src->eof_gid);
1211 		if (!gid_valid(dst->icw_gid))
1212 			return -EINVAL;
1213 	}
1214 	return 0;
1215 }
1216 
1217 static int
1218 xfs_ioctl_getset_resblocks(
1219 	struct file		*filp,
1220 	unsigned int		cmd,
1221 	void __user		*arg)
1222 {
1223 	struct xfs_mount	*mp = XFS_I(file_inode(filp))->i_mount;
1224 	struct xfs_fsop_resblks	fsop = { };
1225 	int			error;
1226 
1227 	if (!capable(CAP_SYS_ADMIN))
1228 		return -EPERM;
1229 
1230 	if (cmd == XFS_IOC_SET_RESBLKS) {
1231 		if (xfs_is_readonly(mp))
1232 			return -EROFS;
1233 
1234 		if (copy_from_user(&fsop, arg, sizeof(fsop)))
1235 			return -EFAULT;
1236 
1237 		error = mnt_want_write_file(filp);
1238 		if (error)
1239 			return error;
1240 		error = xfs_reserve_blocks(mp, fsop.resblks);
1241 		mnt_drop_write_file(filp);
1242 		if (error)
1243 			return error;
1244 	}
1245 
1246 	spin_lock(&mp->m_sb_lock);
1247 	fsop.resblks = mp->m_resblks;
1248 	fsop.resblks_avail = mp->m_resblks_avail;
1249 	spin_unlock(&mp->m_sb_lock);
1250 
1251 	if (copy_to_user(arg, &fsop, sizeof(fsop)))
1252 		return -EFAULT;
1253 	return 0;
1254 }
1255 
1256 static int
1257 xfs_ioctl_fs_counts(
1258 	struct xfs_mount	*mp,
1259 	struct xfs_fsop_counts __user	*uarg)
1260 {
1261 	struct xfs_fsop_counts	out = {
1262 		.allocino = percpu_counter_read_positive(&mp->m_icount),
1263 		.freeino  = percpu_counter_read_positive(&mp->m_ifree),
1264 		.freedata = percpu_counter_read_positive(&mp->m_fdblocks) -
1265 				xfs_fdblocks_unavailable(mp),
1266 		.freertx  = percpu_counter_read_positive(&mp->m_frextents),
1267 	};
1268 
1269 	if (copy_to_user(uarg, &out, sizeof(out)))
1270 		return -EFAULT;
1271 	return 0;
1272 }
1273 
1274 /*
1275  * These long-unused ioctls were removed from the official ioctl API in 5.17,
1276  * but retain these definitions so that we can log warnings about them.
1277  */
1278 #define XFS_IOC_ALLOCSP		_IOW ('X', 10, struct xfs_flock64)
1279 #define XFS_IOC_FREESP		_IOW ('X', 11, struct xfs_flock64)
1280 #define XFS_IOC_ALLOCSP64	_IOW ('X', 36, struct xfs_flock64)
1281 #define XFS_IOC_FREESP64	_IOW ('X', 37, struct xfs_flock64)
1282 
1283 /*
1284  * Note: some of the ioctl's return positive numbers as a
1285  * byte count indicating success, such as readlink_by_handle.
1286  * So we don't "sign flip" like most other routines.  This means
1287  * true errors need to be returned as a negative value.
1288  */
1289 long
1290 xfs_file_ioctl(
1291 	struct file		*filp,
1292 	unsigned int		cmd,
1293 	unsigned long		p)
1294 {
1295 	struct inode		*inode = file_inode(filp);
1296 	struct xfs_inode	*ip = XFS_I(inode);
1297 	struct xfs_mount	*mp = ip->i_mount;
1298 	void			__user *arg = (void __user *)p;
1299 	int			error;
1300 
1301 	trace_xfs_file_ioctl(ip);
1302 
1303 	switch (cmd) {
1304 	case FITRIM:
1305 		return xfs_ioc_trim(mp, arg);
1306 	case FS_IOC_GETFSLABEL:
1307 		return xfs_ioc_getlabel(mp, arg);
1308 	case FS_IOC_SETFSLABEL:
1309 		return xfs_ioc_setlabel(filp, mp, arg);
1310 	case XFS_IOC_ALLOCSP:
1311 	case XFS_IOC_FREESP:
1312 	case XFS_IOC_ALLOCSP64:
1313 	case XFS_IOC_FREESP64:
1314 		xfs_warn_once(mp,
1315 	"%s should use fallocate; XFS_IOC_{ALLOC,FREE}SP ioctl unsupported",
1316 				current->comm);
1317 		return -ENOTTY;
1318 	case XFS_IOC_DIOINFO: {
1319 		struct xfs_buftarg	*target = xfs_inode_buftarg(ip);
1320 		struct dioattr		da;
1321 
1322 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1323 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1324 
1325 		if (copy_to_user(arg, &da, sizeof(da)))
1326 			return -EFAULT;
1327 		return 0;
1328 	}
1329 
1330 	case XFS_IOC_FSBULKSTAT_SINGLE:
1331 	case XFS_IOC_FSBULKSTAT:
1332 	case XFS_IOC_FSINUMBERS:
1333 		return xfs_ioc_fsbulkstat(filp, cmd, arg);
1334 
1335 	case XFS_IOC_BULKSTAT:
1336 		return xfs_ioc_bulkstat(filp, cmd, arg);
1337 	case XFS_IOC_INUMBERS:
1338 		return xfs_ioc_inumbers(mp, cmd, arg);
1339 
1340 	case XFS_IOC_FSGEOMETRY_V1:
1341 		return xfs_ioc_fsgeometry(mp, arg, 3);
1342 	case XFS_IOC_FSGEOMETRY_V4:
1343 		return xfs_ioc_fsgeometry(mp, arg, 4);
1344 	case XFS_IOC_FSGEOMETRY:
1345 		return xfs_ioc_fsgeometry(mp, arg, 5);
1346 
1347 	case XFS_IOC_AG_GEOMETRY:
1348 		return xfs_ioc_ag_geometry(mp, arg);
1349 
1350 	case XFS_IOC_GETVERSION:
1351 		return put_user(inode->i_generation, (int __user *)arg);
1352 
1353 	case XFS_IOC_FSGETXATTRA:
1354 		return xfs_ioc_fsgetxattra(ip, arg);
1355 	case XFS_IOC_GETPARENTS:
1356 		return xfs_ioc_getparents(filp, arg);
1357 	case XFS_IOC_GETPARENTS_BY_HANDLE:
1358 		return xfs_ioc_getparents_by_handle(filp, arg);
1359 	case XFS_IOC_GETBMAP:
1360 	case XFS_IOC_GETBMAPA:
1361 	case XFS_IOC_GETBMAPX:
1362 		return xfs_ioc_getbmap(filp, cmd, arg);
1363 
1364 	case FS_IOC_GETFSMAP:
1365 		return xfs_ioc_getfsmap(ip, arg);
1366 
1367 	case XFS_IOC_SCRUBV_METADATA:
1368 		return xfs_ioc_scrubv_metadata(filp, arg);
1369 	case XFS_IOC_SCRUB_METADATA:
1370 		return xfs_ioc_scrub_metadata(filp, arg);
1371 
1372 	case XFS_IOC_FD_TO_HANDLE:
1373 	case XFS_IOC_PATH_TO_HANDLE:
1374 	case XFS_IOC_PATH_TO_FSHANDLE: {
1375 		xfs_fsop_handlereq_t	hreq;
1376 
1377 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1378 			return -EFAULT;
1379 		return xfs_find_handle(cmd, &hreq);
1380 	}
1381 	case XFS_IOC_OPEN_BY_HANDLE: {
1382 		xfs_fsop_handlereq_t	hreq;
1383 
1384 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1385 			return -EFAULT;
1386 		return xfs_open_by_handle(filp, &hreq);
1387 	}
1388 
1389 	case XFS_IOC_READLINK_BY_HANDLE: {
1390 		xfs_fsop_handlereq_t	hreq;
1391 
1392 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1393 			return -EFAULT;
1394 		return xfs_readlink_by_handle(filp, &hreq);
1395 	}
1396 	case XFS_IOC_ATTRLIST_BY_HANDLE:
1397 		return xfs_attrlist_by_handle(filp, arg);
1398 
1399 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
1400 		return xfs_attrmulti_by_handle(filp, arg);
1401 
1402 	case XFS_IOC_SWAPEXT: {
1403 		struct xfs_swapext	sxp;
1404 
1405 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1406 			return -EFAULT;
1407 		error = mnt_want_write_file(filp);
1408 		if (error)
1409 			return error;
1410 		error = xfs_ioc_swapext(&sxp);
1411 		mnt_drop_write_file(filp);
1412 		return error;
1413 	}
1414 
1415 	case XFS_IOC_FSCOUNTS:
1416 		return xfs_ioctl_fs_counts(mp, arg);
1417 
1418 	case XFS_IOC_SET_RESBLKS:
1419 	case XFS_IOC_GET_RESBLKS:
1420 		return xfs_ioctl_getset_resblocks(filp, cmd, arg);
1421 
1422 	case XFS_IOC_FSGROWFSDATA: {
1423 		struct xfs_growfs_data in;
1424 
1425 		if (copy_from_user(&in, arg, sizeof(in)))
1426 			return -EFAULT;
1427 
1428 		error = mnt_want_write_file(filp);
1429 		if (error)
1430 			return error;
1431 		error = xfs_growfs_data(mp, &in);
1432 		mnt_drop_write_file(filp);
1433 		return error;
1434 	}
1435 
1436 	case XFS_IOC_FSGROWFSLOG: {
1437 		struct xfs_growfs_log in;
1438 
1439 		if (copy_from_user(&in, arg, sizeof(in)))
1440 			return -EFAULT;
1441 
1442 		error = mnt_want_write_file(filp);
1443 		if (error)
1444 			return error;
1445 		error = xfs_growfs_log(mp, &in);
1446 		mnt_drop_write_file(filp);
1447 		return error;
1448 	}
1449 
1450 	case XFS_IOC_FSGROWFSRT: {
1451 		xfs_growfs_rt_t in;
1452 
1453 		if (copy_from_user(&in, arg, sizeof(in)))
1454 			return -EFAULT;
1455 
1456 		error = mnt_want_write_file(filp);
1457 		if (error)
1458 			return error;
1459 		error = xfs_growfs_rt(mp, &in);
1460 		mnt_drop_write_file(filp);
1461 		return error;
1462 	}
1463 
1464 	case XFS_IOC_GOINGDOWN: {
1465 		uint32_t in;
1466 
1467 		if (!capable(CAP_SYS_ADMIN))
1468 			return -EPERM;
1469 
1470 		if (get_user(in, (uint32_t __user *)arg))
1471 			return -EFAULT;
1472 
1473 		return xfs_fs_goingdown(mp, in);
1474 	}
1475 
1476 	case XFS_IOC_ERROR_INJECTION: {
1477 		xfs_error_injection_t in;
1478 
1479 		if (!capable(CAP_SYS_ADMIN))
1480 			return -EPERM;
1481 
1482 		if (copy_from_user(&in, arg, sizeof(in)))
1483 			return -EFAULT;
1484 
1485 		return xfs_errortag_add(mp, in.errtag);
1486 	}
1487 
1488 	case XFS_IOC_ERROR_CLEARALL:
1489 		if (!capable(CAP_SYS_ADMIN))
1490 			return -EPERM;
1491 
1492 		return xfs_errortag_clearall(mp);
1493 
1494 	case XFS_IOC_FREE_EOFBLOCKS: {
1495 		struct xfs_fs_eofblocks	eofb;
1496 		struct xfs_icwalk	icw;
1497 
1498 		if (!capable(CAP_SYS_ADMIN))
1499 			return -EPERM;
1500 
1501 		if (xfs_is_readonly(mp))
1502 			return -EROFS;
1503 
1504 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
1505 			return -EFAULT;
1506 
1507 		error = xfs_fs_eofblocks_from_user(&eofb, &icw);
1508 		if (error)
1509 			return error;
1510 
1511 		trace_xfs_ioc_free_eofblocks(mp, &icw, _RET_IP_);
1512 
1513 		sb_start_write(mp->m_super);
1514 		error = xfs_blockgc_free_space(mp, &icw);
1515 		sb_end_write(mp->m_super);
1516 		return error;
1517 	}
1518 
1519 	case XFS_IOC_EXCHANGE_RANGE:
1520 		return xfs_ioc_exchange_range(filp, arg);
1521 
1522 	default:
1523 		return -ENOTTY;
1524 	}
1525 }
1526