xref: /linux/fs/xfs/xfs_ioctl.c (revision c5288cda69ee2d8607f5026bd599a5cebf0ee783)
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 uint16_t
473 xfs_flags2diflags(
474 	struct xfs_inode	*ip,
475 	unsigned int		xflags)
476 {
477 	/* can't set PREALLOC this way, just preserve it */
478 	uint16_t		di_flags =
479 		(ip->i_diflags & XFS_DIFLAG_PREALLOC);
480 
481 	if (xflags & FS_XFLAG_IMMUTABLE)
482 		di_flags |= XFS_DIFLAG_IMMUTABLE;
483 	if (xflags & FS_XFLAG_APPEND)
484 		di_flags |= XFS_DIFLAG_APPEND;
485 	if (xflags & FS_XFLAG_SYNC)
486 		di_flags |= XFS_DIFLAG_SYNC;
487 	if (xflags & FS_XFLAG_NOATIME)
488 		di_flags |= XFS_DIFLAG_NOATIME;
489 	if (xflags & FS_XFLAG_NODUMP)
490 		di_flags |= XFS_DIFLAG_NODUMP;
491 	if (xflags & FS_XFLAG_NODEFRAG)
492 		di_flags |= XFS_DIFLAG_NODEFRAG;
493 	if (xflags & FS_XFLAG_FILESTREAM)
494 		di_flags |= XFS_DIFLAG_FILESTREAM;
495 	if (S_ISDIR(VFS_I(ip)->i_mode)) {
496 		if (xflags & FS_XFLAG_RTINHERIT)
497 			di_flags |= XFS_DIFLAG_RTINHERIT;
498 		if (xflags & FS_XFLAG_NOSYMLINKS)
499 			di_flags |= XFS_DIFLAG_NOSYMLINKS;
500 		if (xflags & FS_XFLAG_EXTSZINHERIT)
501 			di_flags |= XFS_DIFLAG_EXTSZINHERIT;
502 		if (xflags & FS_XFLAG_PROJINHERIT)
503 			di_flags |= XFS_DIFLAG_PROJINHERIT;
504 	} else if (S_ISREG(VFS_I(ip)->i_mode)) {
505 		if (xflags & FS_XFLAG_REALTIME)
506 			di_flags |= XFS_DIFLAG_REALTIME;
507 		if (xflags & FS_XFLAG_EXTSIZE)
508 			di_flags |= XFS_DIFLAG_EXTSIZE;
509 	}
510 
511 	return di_flags;
512 }
513 
514 STATIC uint64_t
515 xfs_flags2diflags2(
516 	struct xfs_inode	*ip,
517 	unsigned int		xflags)
518 {
519 	uint64_t		di_flags2 =
520 		(ip->i_diflags2 & (XFS_DIFLAG2_REFLINK |
521 				   XFS_DIFLAG2_BIGTIME |
522 				   XFS_DIFLAG2_NREXT64));
523 
524 	if (xflags & FS_XFLAG_DAX)
525 		di_flags2 |= XFS_DIFLAG2_DAX;
526 	if (xflags & FS_XFLAG_COWEXTSIZE)
527 		di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
528 
529 	return di_flags2;
530 }
531 
532 static int
533 xfs_ioctl_setattr_xflags(
534 	struct xfs_trans	*tp,
535 	struct xfs_inode	*ip,
536 	struct fileattr		*fa)
537 {
538 	struct xfs_mount	*mp = ip->i_mount;
539 	bool			rtflag = (fa->fsx_xflags & FS_XFLAG_REALTIME);
540 	uint64_t		i_flags2;
541 
542 	if (rtflag != XFS_IS_REALTIME_INODE(ip)) {
543 		/* Can't change realtime flag if any extents are allocated. */
544 		if (ip->i_df.if_nextents || ip->i_delayed_blks)
545 			return -EINVAL;
546 	}
547 
548 	if (rtflag) {
549 		/* If realtime flag is set then must have realtime device */
550 		if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
551 		    xfs_extlen_to_rtxmod(mp, ip->i_extsize))
552 			return -EINVAL;
553 
554 		/* Clear reflink if we are actually able to set the rt flag. */
555 		if (xfs_is_reflink_inode(ip))
556 			ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
557 	}
558 
559 	/* diflags2 only valid for v3 inodes. */
560 	i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
561 	if (i_flags2 && !xfs_has_v3inodes(mp))
562 		return -EINVAL;
563 
564 	ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
565 	ip->i_diflags2 = i_flags2;
566 
567 	xfs_diflags_to_iflags(ip, false);
568 
569 	/*
570 	 * Make the stable writes flag match that of the device the inode
571 	 * resides on when flipping the RT flag.
572 	 */
573 	if (rtflag != XFS_IS_REALTIME_INODE(ip) && S_ISREG(VFS_I(ip)->i_mode))
574 		xfs_update_stable_writes(ip);
575 
576 	xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
577 	xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
578 	XFS_STATS_INC(mp, xs_ig_attrchg);
579 	return 0;
580 }
581 
582 static void
583 xfs_ioctl_setattr_prepare_dax(
584 	struct xfs_inode	*ip,
585 	struct fileattr		*fa)
586 {
587 	struct xfs_mount	*mp = ip->i_mount;
588 	struct inode            *inode = VFS_I(ip);
589 
590 	if (S_ISDIR(inode->i_mode))
591 		return;
592 
593 	if (xfs_has_dax_always(mp) || xfs_has_dax_never(mp))
594 		return;
595 
596 	if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
597 	    !(ip->i_diflags2 & XFS_DIFLAG2_DAX)) ||
598 	    (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
599 	     (ip->i_diflags2 & XFS_DIFLAG2_DAX)))
600 		d_mark_dontcache(inode);
601 }
602 
603 /*
604  * Set up the transaction structure for the setattr operation, checking that we
605  * have permission to do so. On success, return a clean transaction and the
606  * inode locked exclusively ready for further operation specific checks. On
607  * failure, return an error without modifying or locking the inode.
608  */
609 static struct xfs_trans *
610 xfs_ioctl_setattr_get_trans(
611 	struct xfs_inode	*ip,
612 	struct xfs_dquot	*pdqp)
613 {
614 	struct xfs_mount	*mp = ip->i_mount;
615 	struct xfs_trans	*tp;
616 	int			error = -EROFS;
617 
618 	if (xfs_is_readonly(mp))
619 		goto out_error;
620 	error = -EIO;
621 	if (xfs_is_shutdown(mp))
622 		goto out_error;
623 
624 	error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
625 			has_capability_noaudit(current, CAP_FOWNER), &tp);
626 	if (error)
627 		goto out_error;
628 
629 	if (xfs_has_wsync(mp))
630 		xfs_trans_set_sync(tp);
631 
632 	return tp;
633 
634 out_error:
635 	return ERR_PTR(error);
636 }
637 
638 /*
639  * Validate a proposed extent size hint.  For regular files, the hint can only
640  * be changed if no extents are allocated.
641  */
642 static int
643 xfs_ioctl_setattr_check_extsize(
644 	struct xfs_inode	*ip,
645 	struct fileattr		*fa)
646 {
647 	struct xfs_mount	*mp = ip->i_mount;
648 	xfs_failaddr_t		failaddr;
649 	uint16_t		new_diflags;
650 
651 	if (!fa->fsx_valid)
652 		return 0;
653 
654 	if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
655 	    XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
656 		return -EINVAL;
657 
658 	if (fa->fsx_extsize & mp->m_blockmask)
659 		return -EINVAL;
660 
661 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
662 
663 	/*
664 	 * Inode verifiers do not check that the extent size hint is an integer
665 	 * multiple of the rt extent size on a directory with both rtinherit
666 	 * and extszinherit flags set.  Don't let sysadmins misconfigure
667 	 * directories.
668 	 */
669 	if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
670 	    (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
671 		unsigned int	rtextsize_bytes;
672 
673 		rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
674 		if (fa->fsx_extsize % rtextsize_bytes)
675 			return -EINVAL;
676 	}
677 
678 	failaddr = xfs_inode_validate_extsize(ip->i_mount,
679 			XFS_B_TO_FSB(mp, fa->fsx_extsize),
680 			VFS_I(ip)->i_mode, new_diflags);
681 	return failaddr != NULL ? -EINVAL : 0;
682 }
683 
684 static int
685 xfs_ioctl_setattr_check_cowextsize(
686 	struct xfs_inode	*ip,
687 	struct fileattr		*fa)
688 {
689 	struct xfs_mount	*mp = ip->i_mount;
690 	xfs_failaddr_t		failaddr;
691 	uint64_t		new_diflags2;
692 	uint16_t		new_diflags;
693 
694 	if (!fa->fsx_valid)
695 		return 0;
696 
697 	if (fa->fsx_cowextsize & mp->m_blockmask)
698 		return -EINVAL;
699 
700 	new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
701 	new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
702 
703 	failaddr = xfs_inode_validate_cowextsize(ip->i_mount,
704 			XFS_B_TO_FSB(mp, fa->fsx_cowextsize),
705 			VFS_I(ip)->i_mode, new_diflags, new_diflags2);
706 	return failaddr != NULL ? -EINVAL : 0;
707 }
708 
709 static int
710 xfs_ioctl_setattr_check_projid(
711 	struct xfs_inode	*ip,
712 	struct fileattr		*fa)
713 {
714 	if (!fa->fsx_valid)
715 		return 0;
716 
717 	/* Disallow 32bit project ids if 32bit IDs are not enabled. */
718 	if (fa->fsx_projid > (uint16_t)-1 &&
719 	    !xfs_has_projid32(ip->i_mount))
720 		return -EINVAL;
721 	return 0;
722 }
723 
724 int
725 xfs_fileattr_set(
726 	struct mnt_idmap	*idmap,
727 	struct dentry		*dentry,
728 	struct fileattr		*fa)
729 {
730 	struct xfs_inode	*ip = XFS_I(d_inode(dentry));
731 	struct xfs_mount	*mp = ip->i_mount;
732 	struct xfs_trans	*tp;
733 	struct xfs_dquot	*pdqp = NULL;
734 	struct xfs_dquot	*olddquot = NULL;
735 	int			error;
736 
737 	trace_xfs_ioctl_setattr(ip);
738 
739 	if (d_is_special(dentry))
740 		return -ENOTTY;
741 
742 	if (!fa->fsx_valid) {
743 		if (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL |
744 				  FS_NOATIME_FL | FS_NODUMP_FL |
745 				  FS_SYNC_FL | FS_DAX_FL | FS_PROJINHERIT_FL))
746 			return -EOPNOTSUPP;
747 	}
748 
749 	error = xfs_ioctl_setattr_check_projid(ip, fa);
750 	if (error)
751 		return error;
752 
753 	/*
754 	 * If disk quotas is on, we make sure that the dquots do exist on disk,
755 	 * before we start any other transactions. Trying to do this later
756 	 * is messy. We don't care to take a readlock to look at the ids
757 	 * in inode here, because we can't hold it across the trans_reserve.
758 	 * If the IDs do change before we take the ilock, we're covered
759 	 * because the i_*dquot fields will get updated anyway.
760 	 */
761 	if (fa->fsx_valid && XFS_IS_QUOTA_ON(mp)) {
762 		error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
763 				VFS_I(ip)->i_gid, fa->fsx_projid,
764 				XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
765 		if (error)
766 			return error;
767 	}
768 
769 	xfs_ioctl_setattr_prepare_dax(ip, fa);
770 
771 	tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
772 	if (IS_ERR(tp)) {
773 		error = PTR_ERR(tp);
774 		goto error_free_dquots;
775 	}
776 
777 	error = xfs_ioctl_setattr_check_extsize(ip, fa);
778 	if (error)
779 		goto error_trans_cancel;
780 
781 	error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
782 	if (error)
783 		goto error_trans_cancel;
784 
785 	error = xfs_ioctl_setattr_xflags(tp, ip, fa);
786 	if (error)
787 		goto error_trans_cancel;
788 
789 	if (!fa->fsx_valid)
790 		goto skip_xattr;
791 	/*
792 	 * Change file ownership.  Must be the owner or privileged.  CAP_FSETID
793 	 * overrides the following restrictions:
794 	 *
795 	 * The set-user-ID and set-group-ID bits of a file will be cleared upon
796 	 * successful return from chown()
797 	 */
798 
799 	if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
800 	    !capable_wrt_inode_uidgid(idmap, VFS_I(ip), CAP_FSETID))
801 		VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
802 
803 	/* Change the ownerships and register project quota modifications */
804 	if (ip->i_projid != fa->fsx_projid) {
805 		if (XFS_IS_PQUOTA_ON(mp)) {
806 			olddquot = xfs_qm_vop_chown(tp, ip,
807 						&ip->i_pdquot, pdqp);
808 		}
809 		ip->i_projid = fa->fsx_projid;
810 	}
811 
812 	/*
813 	 * Only set the extent size hint if we've already determined that the
814 	 * extent size hint should be set on the inode. If no extent size flags
815 	 * are set on the inode then unconditionally clear the extent size hint.
816 	 */
817 	if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
818 		ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
819 	else
820 		ip->i_extsize = 0;
821 
822 	if (xfs_has_v3inodes(mp)) {
823 		if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
824 			ip->i_cowextsize = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
825 		else
826 			ip->i_cowextsize = 0;
827 	}
828 
829 skip_xattr:
830 	error = xfs_trans_commit(tp);
831 
832 	/*
833 	 * Release any dquot(s) the inode had kept before chown.
834 	 */
835 	xfs_qm_dqrele(olddquot);
836 	xfs_qm_dqrele(pdqp);
837 
838 	return error;
839 
840 error_trans_cancel:
841 	xfs_trans_cancel(tp);
842 error_free_dquots:
843 	xfs_qm_dqrele(pdqp);
844 	return error;
845 }
846 
847 static bool
848 xfs_getbmap_format(
849 	struct kgetbmap		*p,
850 	struct getbmapx __user	*u,
851 	size_t			recsize)
852 {
853 	if (put_user(p->bmv_offset, &u->bmv_offset) ||
854 	    put_user(p->bmv_block, &u->bmv_block) ||
855 	    put_user(p->bmv_length, &u->bmv_length) ||
856 	    put_user(0, &u->bmv_count) ||
857 	    put_user(0, &u->bmv_entries))
858 		return false;
859 	if (recsize < sizeof(struct getbmapx))
860 		return true;
861 	if (put_user(0, &u->bmv_iflags) ||
862 	    put_user(p->bmv_oflags, &u->bmv_oflags) ||
863 	    put_user(0, &u->bmv_unused1) ||
864 	    put_user(0, &u->bmv_unused2))
865 		return false;
866 	return true;
867 }
868 
869 STATIC int
870 xfs_ioc_getbmap(
871 	struct file		*file,
872 	unsigned int		cmd,
873 	void			__user *arg)
874 {
875 	struct getbmapx		bmx = { 0 };
876 	struct kgetbmap		*buf;
877 	size_t			recsize;
878 	int			error, i;
879 
880 	switch (cmd) {
881 	case XFS_IOC_GETBMAPA:
882 		bmx.bmv_iflags = BMV_IF_ATTRFORK;
883 		fallthrough;
884 	case XFS_IOC_GETBMAP:
885 		/* struct getbmap is a strict subset of struct getbmapx. */
886 		recsize = sizeof(struct getbmap);
887 		break;
888 	case XFS_IOC_GETBMAPX:
889 		recsize = sizeof(struct getbmapx);
890 		break;
891 	default:
892 		return -EINVAL;
893 	}
894 
895 	if (copy_from_user(&bmx, arg, recsize))
896 		return -EFAULT;
897 
898 	if (bmx.bmv_count < 2)
899 		return -EINVAL;
900 	if (bmx.bmv_count >= INT_MAX / recsize)
901 		return -ENOMEM;
902 
903 	buf = kvcalloc(bmx.bmv_count, sizeof(*buf), GFP_KERNEL);
904 	if (!buf)
905 		return -ENOMEM;
906 
907 	error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
908 	if (error)
909 		goto out_free_buf;
910 
911 	error = -EFAULT;
912 	if (copy_to_user(arg, &bmx, recsize))
913 		goto out_free_buf;
914 	arg += recsize;
915 
916 	for (i = 0; i < bmx.bmv_entries; i++) {
917 		if (!xfs_getbmap_format(buf + i, arg, recsize))
918 			goto out_free_buf;
919 		arg += recsize;
920 	}
921 
922 	error = 0;
923 out_free_buf:
924 	kvfree(buf);
925 	return error;
926 }
927 
928 STATIC int
929 xfs_ioc_getfsmap(
930 	struct xfs_inode	*ip,
931 	struct fsmap_head	__user *arg)
932 {
933 	struct xfs_fsmap_head	xhead = {0};
934 	struct fsmap_head	head;
935 	struct fsmap		*recs;
936 	unsigned int		count;
937 	__u32			last_flags = 0;
938 	bool			done = false;
939 	int			error;
940 
941 	if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
942 		return -EFAULT;
943 	if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
944 	    memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
945 		       sizeof(head.fmh_keys[0].fmr_reserved)) ||
946 	    memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
947 		       sizeof(head.fmh_keys[1].fmr_reserved)))
948 		return -EINVAL;
949 
950 	/*
951 	 * Use an internal memory buffer so that we don't have to copy fsmap
952 	 * data to userspace while holding locks.  Start by trying to allocate
953 	 * up to 128k for the buffer, but fall back to a single page if needed.
954 	 */
955 	count = min_t(unsigned int, head.fmh_count,
956 			131072 / sizeof(struct fsmap));
957 	recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
958 	if (!recs) {
959 		count = min_t(unsigned int, head.fmh_count,
960 				PAGE_SIZE / sizeof(struct fsmap));
961 		recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
962 		if (!recs)
963 			return -ENOMEM;
964 	}
965 
966 	xhead.fmh_iflags = head.fmh_iflags;
967 	xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
968 	xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
969 
970 	trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
971 	trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
972 
973 	head.fmh_entries = 0;
974 	do {
975 		struct fsmap __user	*user_recs;
976 		struct fsmap		*last_rec;
977 
978 		user_recs = &arg->fmh_recs[head.fmh_entries];
979 		xhead.fmh_entries = 0;
980 		xhead.fmh_count = min_t(unsigned int, count,
981 					head.fmh_count - head.fmh_entries);
982 
983 		/* Run query, record how many entries we got. */
984 		error = xfs_getfsmap(ip->i_mount, &xhead, recs);
985 		switch (error) {
986 		case 0:
987 			/*
988 			 * There are no more records in the result set.  Copy
989 			 * whatever we got to userspace and break out.
990 			 */
991 			done = true;
992 			break;
993 		case -ECANCELED:
994 			/*
995 			 * The internal memory buffer is full.  Copy whatever
996 			 * records we got to userspace and go again if we have
997 			 * not yet filled the userspace buffer.
998 			 */
999 			error = 0;
1000 			break;
1001 		default:
1002 			goto out_free;
1003 		}
1004 		head.fmh_entries += xhead.fmh_entries;
1005 		head.fmh_oflags = xhead.fmh_oflags;
1006 
1007 		/*
1008 		 * If the caller wanted a record count or there aren't any
1009 		 * new records to return, we're done.
1010 		 */
1011 		if (head.fmh_count == 0 || xhead.fmh_entries == 0)
1012 			break;
1013 
1014 		/* Copy all the records we got out to userspace. */
1015 		if (copy_to_user(user_recs, recs,
1016 				 xhead.fmh_entries * sizeof(struct fsmap))) {
1017 			error = -EFAULT;
1018 			goto out_free;
1019 		}
1020 
1021 		/* Remember the last record flags we copied to userspace. */
1022 		last_rec = &recs[xhead.fmh_entries - 1];
1023 		last_flags = last_rec->fmr_flags;
1024 
1025 		/* Set up the low key for the next iteration. */
1026 		xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
1027 		trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1028 	} while (!done && head.fmh_entries < head.fmh_count);
1029 
1030 	/*
1031 	 * If there are no more records in the query result set and we're not
1032 	 * in counting mode, mark the last record returned with the LAST flag.
1033 	 */
1034 	if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
1035 		struct fsmap __user	*user_rec;
1036 
1037 		last_flags |= FMR_OF_LAST;
1038 		user_rec = &arg->fmh_recs[head.fmh_entries - 1];
1039 
1040 		if (copy_to_user(&user_rec->fmr_flags, &last_flags,
1041 					sizeof(last_flags))) {
1042 			error = -EFAULT;
1043 			goto out_free;
1044 		}
1045 	}
1046 
1047 	/* copy back header */
1048 	if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1049 		error = -EFAULT;
1050 		goto out_free;
1051 	}
1052 
1053 out_free:
1054 	kvfree(recs);
1055 	return error;
1056 }
1057 
1058 int
1059 xfs_ioc_swapext(
1060 	xfs_swapext_t	*sxp)
1061 {
1062 	xfs_inode_t     *ip, *tip;
1063 	struct fd	f, tmp;
1064 	int		error = 0;
1065 
1066 	/* Pull information for the target fd */
1067 	f = fdget((int)sxp->sx_fdtarget);
1068 	if (!f.file) {
1069 		error = -EINVAL;
1070 		goto out;
1071 	}
1072 
1073 	if (!(f.file->f_mode & FMODE_WRITE) ||
1074 	    !(f.file->f_mode & FMODE_READ) ||
1075 	    (f.file->f_flags & O_APPEND)) {
1076 		error = -EBADF;
1077 		goto out_put_file;
1078 	}
1079 
1080 	tmp = fdget((int)sxp->sx_fdtmp);
1081 	if (!tmp.file) {
1082 		error = -EINVAL;
1083 		goto out_put_file;
1084 	}
1085 
1086 	if (!(tmp.file->f_mode & FMODE_WRITE) ||
1087 	    !(tmp.file->f_mode & FMODE_READ) ||
1088 	    (tmp.file->f_flags & O_APPEND)) {
1089 		error = -EBADF;
1090 		goto out_put_tmp_file;
1091 	}
1092 
1093 	if (IS_SWAPFILE(file_inode(f.file)) ||
1094 	    IS_SWAPFILE(file_inode(tmp.file))) {
1095 		error = -EINVAL;
1096 		goto out_put_tmp_file;
1097 	}
1098 
1099 	/*
1100 	 * We need to ensure that the fds passed in point to XFS inodes
1101 	 * before we cast and access them as XFS structures as we have no
1102 	 * control over what the user passes us here.
1103 	 */
1104 	if (f.file->f_op != &xfs_file_operations ||
1105 	    tmp.file->f_op != &xfs_file_operations) {
1106 		error = -EINVAL;
1107 		goto out_put_tmp_file;
1108 	}
1109 
1110 	ip = XFS_I(file_inode(f.file));
1111 	tip = XFS_I(file_inode(tmp.file));
1112 
1113 	if (ip->i_mount != tip->i_mount) {
1114 		error = -EINVAL;
1115 		goto out_put_tmp_file;
1116 	}
1117 
1118 	if (ip->i_ino == tip->i_ino) {
1119 		error = -EINVAL;
1120 		goto out_put_tmp_file;
1121 	}
1122 
1123 	if (xfs_is_shutdown(ip->i_mount)) {
1124 		error = -EIO;
1125 		goto out_put_tmp_file;
1126 	}
1127 
1128 	error = xfs_swap_extents(ip, tip, sxp);
1129 
1130  out_put_tmp_file:
1131 	fdput(tmp);
1132  out_put_file:
1133 	fdput(f);
1134  out:
1135 	return error;
1136 }
1137 
1138 static int
1139 xfs_ioc_getlabel(
1140 	struct xfs_mount	*mp,
1141 	char			__user *user_label)
1142 {
1143 	struct xfs_sb		*sbp = &mp->m_sb;
1144 	char			label[XFSLABEL_MAX + 1];
1145 
1146 	/* Paranoia */
1147 	BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1148 
1149 	/* 1 larger than sb_fname, so this ensures a trailing NUL char */
1150 	memset(label, 0, sizeof(label));
1151 	spin_lock(&mp->m_sb_lock);
1152 	strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1153 	spin_unlock(&mp->m_sb_lock);
1154 
1155 	if (copy_to_user(user_label, label, sizeof(label)))
1156 		return -EFAULT;
1157 	return 0;
1158 }
1159 
1160 static int
1161 xfs_ioc_setlabel(
1162 	struct file		*filp,
1163 	struct xfs_mount	*mp,
1164 	char			__user *newlabel)
1165 {
1166 	struct xfs_sb		*sbp = &mp->m_sb;
1167 	char			label[XFSLABEL_MAX + 1];
1168 	size_t			len;
1169 	int			error;
1170 
1171 	if (!capable(CAP_SYS_ADMIN))
1172 		return -EPERM;
1173 	/*
1174 	 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1175 	 * smaller, at 12 bytes.  We copy one more to be sure we find the
1176 	 * (required) NULL character to test the incoming label length.
1177 	 * NB: The on disk label doesn't need to be null terminated.
1178 	 */
1179 	if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1180 		return -EFAULT;
1181 	len = strnlen(label, XFSLABEL_MAX + 1);
1182 	if (len > sizeof(sbp->sb_fname))
1183 		return -EINVAL;
1184 
1185 	error = mnt_want_write_file(filp);
1186 	if (error)
1187 		return error;
1188 
1189 	spin_lock(&mp->m_sb_lock);
1190 	memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1191 	memcpy(sbp->sb_fname, label, len);
1192 	spin_unlock(&mp->m_sb_lock);
1193 
1194 	/*
1195 	 * Now we do several things to satisfy userspace.
1196 	 * In addition to normal logging of the primary superblock, we also
1197 	 * immediately write these changes to sector zero for the primary, then
1198 	 * update all backup supers (as xfs_db does for a label change), then
1199 	 * invalidate the block device page cache.  This is so that any prior
1200 	 * buffered reads from userspace (i.e. from blkid) are invalidated,
1201 	 * and userspace will see the newly-written label.
1202 	 */
1203 	error = xfs_sync_sb_buf(mp);
1204 	if (error)
1205 		goto out;
1206 	/*
1207 	 * growfs also updates backup supers so lock against that.
1208 	 */
1209 	mutex_lock(&mp->m_growlock);
1210 	error = xfs_update_secondary_sbs(mp);
1211 	mutex_unlock(&mp->m_growlock);
1212 
1213 	invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1214 
1215 out:
1216 	mnt_drop_write_file(filp);
1217 	return error;
1218 }
1219 
1220 static inline int
1221 xfs_fs_eofblocks_from_user(
1222 	struct xfs_fs_eofblocks		*src,
1223 	struct xfs_icwalk		*dst)
1224 {
1225 	if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1226 		return -EINVAL;
1227 
1228 	if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1229 		return -EINVAL;
1230 
1231 	if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1232 	    memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1233 		return -EINVAL;
1234 
1235 	dst->icw_flags = 0;
1236 	if (src->eof_flags & XFS_EOF_FLAGS_SYNC)
1237 		dst->icw_flags |= XFS_ICWALK_FLAG_SYNC;
1238 	if (src->eof_flags & XFS_EOF_FLAGS_UID)
1239 		dst->icw_flags |= XFS_ICWALK_FLAG_UID;
1240 	if (src->eof_flags & XFS_EOF_FLAGS_GID)
1241 		dst->icw_flags |= XFS_ICWALK_FLAG_GID;
1242 	if (src->eof_flags & XFS_EOF_FLAGS_PRID)
1243 		dst->icw_flags |= XFS_ICWALK_FLAG_PRID;
1244 	if (src->eof_flags & XFS_EOF_FLAGS_MINFILESIZE)
1245 		dst->icw_flags |= XFS_ICWALK_FLAG_MINFILESIZE;
1246 
1247 	dst->icw_prid = src->eof_prid;
1248 	dst->icw_min_file_size = src->eof_min_file_size;
1249 
1250 	dst->icw_uid = INVALID_UID;
1251 	if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1252 		dst->icw_uid = make_kuid(current_user_ns(), src->eof_uid);
1253 		if (!uid_valid(dst->icw_uid))
1254 			return -EINVAL;
1255 	}
1256 
1257 	dst->icw_gid = INVALID_GID;
1258 	if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1259 		dst->icw_gid = make_kgid(current_user_ns(), src->eof_gid);
1260 		if (!gid_valid(dst->icw_gid))
1261 			return -EINVAL;
1262 	}
1263 	return 0;
1264 }
1265 
1266 static int
1267 xfs_ioctl_getset_resblocks(
1268 	struct file		*filp,
1269 	unsigned int		cmd,
1270 	void __user		*arg)
1271 {
1272 	struct xfs_mount	*mp = XFS_I(file_inode(filp))->i_mount;
1273 	struct xfs_fsop_resblks	fsop = { };
1274 	int			error;
1275 
1276 	if (!capable(CAP_SYS_ADMIN))
1277 		return -EPERM;
1278 
1279 	if (cmd == XFS_IOC_SET_RESBLKS) {
1280 		if (xfs_is_readonly(mp))
1281 			return -EROFS;
1282 
1283 		if (copy_from_user(&fsop, arg, sizeof(fsop)))
1284 			return -EFAULT;
1285 
1286 		error = mnt_want_write_file(filp);
1287 		if (error)
1288 			return error;
1289 		error = xfs_reserve_blocks(mp, fsop.resblks);
1290 		mnt_drop_write_file(filp);
1291 		if (error)
1292 			return error;
1293 	}
1294 
1295 	spin_lock(&mp->m_sb_lock);
1296 	fsop.resblks = mp->m_resblks;
1297 	fsop.resblks_avail = mp->m_resblks_avail;
1298 	spin_unlock(&mp->m_sb_lock);
1299 
1300 	if (copy_to_user(arg, &fsop, sizeof(fsop)))
1301 		return -EFAULT;
1302 	return 0;
1303 }
1304 
1305 static int
1306 xfs_ioctl_fs_counts(
1307 	struct xfs_mount	*mp,
1308 	struct xfs_fsop_counts __user	*uarg)
1309 {
1310 	struct xfs_fsop_counts	out = {
1311 		.allocino = percpu_counter_read_positive(&mp->m_icount),
1312 		.freeino  = percpu_counter_read_positive(&mp->m_ifree),
1313 		.freedata = percpu_counter_read_positive(&mp->m_fdblocks) -
1314 				xfs_fdblocks_unavailable(mp),
1315 		.freertx  = percpu_counter_read_positive(&mp->m_frextents),
1316 	};
1317 
1318 	if (copy_to_user(uarg, &out, sizeof(out)))
1319 		return -EFAULT;
1320 	return 0;
1321 }
1322 
1323 /*
1324  * These long-unused ioctls were removed from the official ioctl API in 5.17,
1325  * but retain these definitions so that we can log warnings about them.
1326  */
1327 #define XFS_IOC_ALLOCSP		_IOW ('X', 10, struct xfs_flock64)
1328 #define XFS_IOC_FREESP		_IOW ('X', 11, struct xfs_flock64)
1329 #define XFS_IOC_ALLOCSP64	_IOW ('X', 36, struct xfs_flock64)
1330 #define XFS_IOC_FREESP64	_IOW ('X', 37, struct xfs_flock64)
1331 
1332 /*
1333  * Note: some of the ioctl's return positive numbers as a
1334  * byte count indicating success, such as readlink_by_handle.
1335  * So we don't "sign flip" like most other routines.  This means
1336  * true errors need to be returned as a negative value.
1337  */
1338 long
1339 xfs_file_ioctl(
1340 	struct file		*filp,
1341 	unsigned int		cmd,
1342 	unsigned long		p)
1343 {
1344 	struct inode		*inode = file_inode(filp);
1345 	struct xfs_inode	*ip = XFS_I(inode);
1346 	struct xfs_mount	*mp = ip->i_mount;
1347 	void			__user *arg = (void __user *)p;
1348 	int			error;
1349 
1350 	trace_xfs_file_ioctl(ip);
1351 
1352 	switch (cmd) {
1353 	case FITRIM:
1354 		return xfs_ioc_trim(mp, arg);
1355 	case FS_IOC_GETFSLABEL:
1356 		return xfs_ioc_getlabel(mp, arg);
1357 	case FS_IOC_SETFSLABEL:
1358 		return xfs_ioc_setlabel(filp, mp, arg);
1359 	case XFS_IOC_ALLOCSP:
1360 	case XFS_IOC_FREESP:
1361 	case XFS_IOC_ALLOCSP64:
1362 	case XFS_IOC_FREESP64:
1363 		xfs_warn_once(mp,
1364 	"%s should use fallocate; XFS_IOC_{ALLOC,FREE}SP ioctl unsupported",
1365 				current->comm);
1366 		return -ENOTTY;
1367 	case XFS_IOC_DIOINFO: {
1368 		struct xfs_buftarg	*target = xfs_inode_buftarg(ip);
1369 		struct dioattr		da;
1370 
1371 		da.d_mem =  da.d_miniosz = target->bt_logical_sectorsize;
1372 		da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1373 
1374 		if (copy_to_user(arg, &da, sizeof(da)))
1375 			return -EFAULT;
1376 		return 0;
1377 	}
1378 
1379 	case XFS_IOC_FSBULKSTAT_SINGLE:
1380 	case XFS_IOC_FSBULKSTAT:
1381 	case XFS_IOC_FSINUMBERS:
1382 		return xfs_ioc_fsbulkstat(filp, cmd, arg);
1383 
1384 	case XFS_IOC_BULKSTAT:
1385 		return xfs_ioc_bulkstat(filp, cmd, arg);
1386 	case XFS_IOC_INUMBERS:
1387 		return xfs_ioc_inumbers(mp, cmd, arg);
1388 
1389 	case XFS_IOC_FSGEOMETRY_V1:
1390 		return xfs_ioc_fsgeometry(mp, arg, 3);
1391 	case XFS_IOC_FSGEOMETRY_V4:
1392 		return xfs_ioc_fsgeometry(mp, arg, 4);
1393 	case XFS_IOC_FSGEOMETRY:
1394 		return xfs_ioc_fsgeometry(mp, arg, 5);
1395 
1396 	case XFS_IOC_AG_GEOMETRY:
1397 		return xfs_ioc_ag_geometry(mp, arg);
1398 
1399 	case XFS_IOC_GETVERSION:
1400 		return put_user(inode->i_generation, (int __user *)arg);
1401 
1402 	case XFS_IOC_FSGETXATTRA:
1403 		return xfs_ioc_fsgetxattra(ip, arg);
1404 	case XFS_IOC_GETPARENTS:
1405 		return xfs_ioc_getparents(filp, arg);
1406 	case XFS_IOC_GETPARENTS_BY_HANDLE:
1407 		return xfs_ioc_getparents_by_handle(filp, arg);
1408 	case XFS_IOC_GETBMAP:
1409 	case XFS_IOC_GETBMAPA:
1410 	case XFS_IOC_GETBMAPX:
1411 		return xfs_ioc_getbmap(filp, cmd, arg);
1412 
1413 	case FS_IOC_GETFSMAP:
1414 		return xfs_ioc_getfsmap(ip, arg);
1415 
1416 	case XFS_IOC_SCRUBV_METADATA:
1417 		return xfs_ioc_scrubv_metadata(filp, arg);
1418 	case XFS_IOC_SCRUB_METADATA:
1419 		return xfs_ioc_scrub_metadata(filp, arg);
1420 
1421 	case XFS_IOC_FD_TO_HANDLE:
1422 	case XFS_IOC_PATH_TO_HANDLE:
1423 	case XFS_IOC_PATH_TO_FSHANDLE: {
1424 		xfs_fsop_handlereq_t	hreq;
1425 
1426 		if (copy_from_user(&hreq, arg, sizeof(hreq)))
1427 			return -EFAULT;
1428 		return xfs_find_handle(cmd, &hreq);
1429 	}
1430 	case XFS_IOC_OPEN_BY_HANDLE: {
1431 		xfs_fsop_handlereq_t	hreq;
1432 
1433 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1434 			return -EFAULT;
1435 		return xfs_open_by_handle(filp, &hreq);
1436 	}
1437 
1438 	case XFS_IOC_READLINK_BY_HANDLE: {
1439 		xfs_fsop_handlereq_t	hreq;
1440 
1441 		if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1442 			return -EFAULT;
1443 		return xfs_readlink_by_handle(filp, &hreq);
1444 	}
1445 	case XFS_IOC_ATTRLIST_BY_HANDLE:
1446 		return xfs_attrlist_by_handle(filp, arg);
1447 
1448 	case XFS_IOC_ATTRMULTI_BY_HANDLE:
1449 		return xfs_attrmulti_by_handle(filp, arg);
1450 
1451 	case XFS_IOC_SWAPEXT: {
1452 		struct xfs_swapext	sxp;
1453 
1454 		if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1455 			return -EFAULT;
1456 		error = mnt_want_write_file(filp);
1457 		if (error)
1458 			return error;
1459 		error = xfs_ioc_swapext(&sxp);
1460 		mnt_drop_write_file(filp);
1461 		return error;
1462 	}
1463 
1464 	case XFS_IOC_FSCOUNTS:
1465 		return xfs_ioctl_fs_counts(mp, arg);
1466 
1467 	case XFS_IOC_SET_RESBLKS:
1468 	case XFS_IOC_GET_RESBLKS:
1469 		return xfs_ioctl_getset_resblocks(filp, cmd, arg);
1470 
1471 	case XFS_IOC_FSGROWFSDATA: {
1472 		struct xfs_growfs_data in;
1473 
1474 		if (copy_from_user(&in, arg, sizeof(in)))
1475 			return -EFAULT;
1476 
1477 		error = mnt_want_write_file(filp);
1478 		if (error)
1479 			return error;
1480 		error = xfs_growfs_data(mp, &in);
1481 		mnt_drop_write_file(filp);
1482 		return error;
1483 	}
1484 
1485 	case XFS_IOC_FSGROWFSLOG: {
1486 		struct xfs_growfs_log in;
1487 
1488 		if (copy_from_user(&in, arg, sizeof(in)))
1489 			return -EFAULT;
1490 
1491 		error = mnt_want_write_file(filp);
1492 		if (error)
1493 			return error;
1494 		error = xfs_growfs_log(mp, &in);
1495 		mnt_drop_write_file(filp);
1496 		return error;
1497 	}
1498 
1499 	case XFS_IOC_FSGROWFSRT: {
1500 		xfs_growfs_rt_t in;
1501 
1502 		if (copy_from_user(&in, arg, sizeof(in)))
1503 			return -EFAULT;
1504 
1505 		error = mnt_want_write_file(filp);
1506 		if (error)
1507 			return error;
1508 		error = xfs_growfs_rt(mp, &in);
1509 		mnt_drop_write_file(filp);
1510 		return error;
1511 	}
1512 
1513 	case XFS_IOC_GOINGDOWN: {
1514 		uint32_t in;
1515 
1516 		if (!capable(CAP_SYS_ADMIN))
1517 			return -EPERM;
1518 
1519 		if (get_user(in, (uint32_t __user *)arg))
1520 			return -EFAULT;
1521 
1522 		return xfs_fs_goingdown(mp, in);
1523 	}
1524 
1525 	case XFS_IOC_ERROR_INJECTION: {
1526 		xfs_error_injection_t in;
1527 
1528 		if (!capable(CAP_SYS_ADMIN))
1529 			return -EPERM;
1530 
1531 		if (copy_from_user(&in, arg, sizeof(in)))
1532 			return -EFAULT;
1533 
1534 		return xfs_errortag_add(mp, in.errtag);
1535 	}
1536 
1537 	case XFS_IOC_ERROR_CLEARALL:
1538 		if (!capable(CAP_SYS_ADMIN))
1539 			return -EPERM;
1540 
1541 		return xfs_errortag_clearall(mp);
1542 
1543 	case XFS_IOC_FREE_EOFBLOCKS: {
1544 		struct xfs_fs_eofblocks	eofb;
1545 		struct xfs_icwalk	icw;
1546 
1547 		if (!capable(CAP_SYS_ADMIN))
1548 			return -EPERM;
1549 
1550 		if (xfs_is_readonly(mp))
1551 			return -EROFS;
1552 
1553 		if (copy_from_user(&eofb, arg, sizeof(eofb)))
1554 			return -EFAULT;
1555 
1556 		error = xfs_fs_eofblocks_from_user(&eofb, &icw);
1557 		if (error)
1558 			return error;
1559 
1560 		trace_xfs_ioc_free_eofblocks(mp, &icw, _RET_IP_);
1561 
1562 		sb_start_write(mp->m_super);
1563 		error = xfs_blockgc_free_space(mp, &icw);
1564 		sb_end_write(mp->m_super);
1565 		return error;
1566 	}
1567 
1568 	case XFS_IOC_EXCHANGE_RANGE:
1569 		return xfs_ioc_exchange_range(filp, arg);
1570 
1571 	default:
1572 		return -ENOTTY;
1573 	}
1574 }
1575