Loading...
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#include "xfs_rtgroup.h"
44
45#include <linux/mount.h>
46#include <linux/fileattr.h>
47
48/* Return 0 on success or positive error */
49int
50xfs_fsbulkstat_one_fmt(
51 struct xfs_ibulk *breq,
52 const struct xfs_bulkstat *bstat)
53{
54 struct xfs_bstat bs1;
55
56 xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
57 if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
58 return -EFAULT;
59 return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
60}
61
62int
63xfs_fsinumbers_fmt(
64 struct xfs_ibulk *breq,
65 const struct xfs_inumbers *igrp)
66{
67 struct xfs_inogrp ig1;
68
69 xfs_inumbers_to_inogrp(&ig1, igrp);
70 if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
71 return -EFAULT;
72 return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
73}
74
75STATIC int
76xfs_ioc_fsbulkstat(
77 struct file *file,
78 unsigned int cmd,
79 void __user *arg)
80{
81 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
82 struct xfs_fsop_bulkreq bulkreq;
83 struct xfs_ibulk breq = {
84 .mp = mp,
85 .idmap = file_mnt_idmap(file),
86 .ocount = 0,
87 };
88 xfs_ino_t lastino;
89 int error;
90
91 /* done = 1 if there are more stats to get and if bulkstat */
92 /* should be called again (unused here, but used in dmapi) */
93
94 if (!capable(CAP_SYS_ADMIN))
95 return -EPERM;
96
97 if (xfs_is_shutdown(mp))
98 return -EIO;
99
100 if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
101 return -EFAULT;
102
103 if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
104 return -EFAULT;
105
106 if (bulkreq.icount <= 0)
107 return -EINVAL;
108
109 if (bulkreq.ubuffer == NULL)
110 return -EINVAL;
111
112 breq.ubuffer = bulkreq.ubuffer;
113 breq.icount = bulkreq.icount;
114
115 /*
116 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
117 * that we want to stat. However, FSINUMBERS and FSBULKSTAT expect
118 * that *lastip contains either zero or the number of the last inode to
119 * be examined by the previous call and return results starting with
120 * the next inode after that. The new bulk request back end functions
121 * take the inode to start with, so we have to compute the startino
122 * parameter from lastino to maintain correct function. lastino == 0
123 * is a special case because it has traditionally meant "first inode
124 * in filesystem".
125 */
126 if (cmd == XFS_IOC_FSINUMBERS) {
127 breq.startino = lastino ? lastino + 1 : 0;
128 error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
129 lastino = breq.startino - 1;
130 } else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
131 breq.startino = lastino;
132 breq.icount = 1;
133 error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
134 } else { /* XFS_IOC_FSBULKSTAT */
135 breq.startino = lastino ? lastino + 1 : 0;
136 error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
137 lastino = breq.startino - 1;
138 }
139
140 if (error)
141 return error;
142
143 if (bulkreq.lastip != NULL &&
144 copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
145 return -EFAULT;
146
147 if (bulkreq.ocount != NULL &&
148 copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
149 return -EFAULT;
150
151 return 0;
152}
153
154/* Return 0 on success or positive error */
155static int
156xfs_bulkstat_fmt(
157 struct xfs_ibulk *breq,
158 const struct xfs_bulkstat *bstat)
159{
160 if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
161 return -EFAULT;
162 return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
163}
164
165/*
166 * Check the incoming bulk request @hdr from userspace and initialize the
167 * internal @breq bulk request appropriately. Returns 0 if the bulk request
168 * should proceed; -ECANCELED if there's nothing to do; or the usual
169 * negative error code.
170 */
171static int
172xfs_bulk_ireq_setup(
173 struct xfs_mount *mp,
174 const struct xfs_bulk_ireq *hdr,
175 struct xfs_ibulk *breq,
176 void __user *ubuffer)
177{
178 if (hdr->icount == 0 ||
179 (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
180 memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
181 return -EINVAL;
182
183 breq->startino = hdr->ino;
184 breq->ubuffer = ubuffer;
185 breq->icount = hdr->icount;
186 breq->ocount = 0;
187 breq->flags = 0;
188
189 /*
190 * The @ino parameter is a special value, so we must look it up here.
191 * We're not allowed to have IREQ_AGNO, and we only return one inode
192 * worth of data.
193 */
194 if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
195 if (hdr->flags & XFS_BULK_IREQ_AGNO)
196 return -EINVAL;
197
198 switch (hdr->ino) {
199 case XFS_BULK_IREQ_SPECIAL_ROOT:
200 breq->startino = mp->m_sb.sb_rootino;
201 break;
202 default:
203 return -EINVAL;
204 }
205 breq->icount = 1;
206 }
207
208 /*
209 * The IREQ_AGNO flag means that we only want results from a given AG.
210 * If @hdr->ino is zero, we start iterating in that AG. If @hdr->ino is
211 * beyond the specified AG then we return no results.
212 */
213 if (hdr->flags & XFS_BULK_IREQ_AGNO) {
214 if (hdr->agno >= mp->m_sb.sb_agcount)
215 return -EINVAL;
216
217 if (breq->startino == 0)
218 breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
219 else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
220 return -EINVAL;
221
222 breq->flags |= XFS_IBULK_SAME_AG;
223
224 /* Asking for an inode past the end of the AG? We're done! */
225 if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
226 return -ECANCELED;
227 } else if (hdr->agno)
228 return -EINVAL;
229
230 /* Asking for an inode past the end of the FS? We're done! */
231 if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
232 return -ECANCELED;
233
234 if (hdr->flags & XFS_BULK_IREQ_NREXT64)
235 breq->flags |= XFS_IBULK_NREXT64;
236
237 /* Caller wants to see metadata directories in bulkstat output. */
238 if (hdr->flags & XFS_BULK_IREQ_METADIR)
239 breq->flags |= XFS_IBULK_METADIR;
240
241 return 0;
242}
243
244/*
245 * Update the userspace bulk request @hdr to reflect the end state of the
246 * internal bulk request @breq.
247 */
248static void
249xfs_bulk_ireq_teardown(
250 struct xfs_bulk_ireq *hdr,
251 struct xfs_ibulk *breq)
252{
253 hdr->ino = breq->startino;
254 hdr->ocount = breq->ocount;
255}
256
257/* Handle the v5 bulkstat ioctl. */
258STATIC int
259xfs_ioc_bulkstat(
260 struct file *file,
261 unsigned int cmd,
262 struct xfs_bulkstat_req __user *arg)
263{
264 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
265 struct xfs_bulk_ireq hdr;
266 struct xfs_ibulk breq = {
267 .mp = mp,
268 .idmap = file_mnt_idmap(file),
269 };
270 int error;
271
272 if (!capable(CAP_SYS_ADMIN))
273 return -EPERM;
274
275 if (xfs_is_shutdown(mp))
276 return -EIO;
277
278 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
279 return -EFAULT;
280
281 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
282 if (error == -ECANCELED)
283 goto out_teardown;
284 if (error < 0)
285 return error;
286
287 error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
288 if (error)
289 return error;
290
291out_teardown:
292 xfs_bulk_ireq_teardown(&hdr, &breq);
293 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
294 return -EFAULT;
295
296 return 0;
297}
298
299STATIC int
300xfs_inumbers_fmt(
301 struct xfs_ibulk *breq,
302 const struct xfs_inumbers *igrp)
303{
304 if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
305 return -EFAULT;
306 return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
307}
308
309/* Handle the v5 inumbers ioctl. */
310STATIC int
311xfs_ioc_inumbers(
312 struct xfs_mount *mp,
313 unsigned int cmd,
314 struct xfs_inumbers_req __user *arg)
315{
316 struct xfs_bulk_ireq hdr;
317 struct xfs_ibulk breq = {
318 .mp = mp,
319 };
320 int error;
321
322 if (!capable(CAP_SYS_ADMIN))
323 return -EPERM;
324
325 if (xfs_is_shutdown(mp))
326 return -EIO;
327
328 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
329 return -EFAULT;
330
331 if (hdr.flags & XFS_BULK_IREQ_METADIR)
332 return -EINVAL;
333
334 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
335 if (error == -ECANCELED)
336 goto out_teardown;
337 if (error < 0)
338 return error;
339
340 error = xfs_inumbers(&breq, xfs_inumbers_fmt);
341 if (error)
342 return error;
343
344out_teardown:
345 xfs_bulk_ireq_teardown(&hdr, &breq);
346 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
347 return -EFAULT;
348
349 return 0;
350}
351
352STATIC int
353xfs_ioc_fsgeometry(
354 struct xfs_mount *mp,
355 void __user *arg,
356 int struct_version)
357{
358 struct xfs_fsop_geom fsgeo;
359 size_t len;
360
361 xfs_fs_geometry(mp, &fsgeo, struct_version);
362
363 if (struct_version <= 3)
364 len = sizeof(struct xfs_fsop_geom_v1);
365 else if (struct_version == 4)
366 len = sizeof(struct xfs_fsop_geom_v4);
367 else {
368 xfs_fsop_geom_health(mp, &fsgeo);
369 len = sizeof(fsgeo);
370 }
371
372 if (copy_to_user(arg, &fsgeo, len))
373 return -EFAULT;
374 return 0;
375}
376
377STATIC int
378xfs_ioc_ag_geometry(
379 struct xfs_mount *mp,
380 void __user *arg)
381{
382 struct xfs_perag *pag;
383 struct xfs_ag_geometry ageo;
384 int error;
385
386 if (copy_from_user(&ageo, arg, sizeof(ageo)))
387 return -EFAULT;
388 if (ageo.ag_flags)
389 return -EINVAL;
390 if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
391 return -EINVAL;
392
393 pag = xfs_perag_get(mp, ageo.ag_number);
394 if (!pag)
395 return -EINVAL;
396
397 error = xfs_ag_get_geometry(pag, &ageo);
398 xfs_perag_put(pag);
399 if (error)
400 return error;
401
402 if (copy_to_user(arg, &ageo, sizeof(ageo)))
403 return -EFAULT;
404 return 0;
405}
406
407STATIC int
408xfs_ioc_rtgroup_geometry(
409 struct xfs_mount *mp,
410 void __user *arg)
411{
412 struct xfs_rtgroup *rtg;
413 struct xfs_rtgroup_geometry rgeo;
414 int error;
415
416 if (copy_from_user(&rgeo, arg, sizeof(rgeo)))
417 return -EFAULT;
418 if (rgeo.rg_flags)
419 return -EINVAL;
420 if (memchr_inv(&rgeo.rg_reserved, 0, sizeof(rgeo.rg_reserved)))
421 return -EINVAL;
422 if (!xfs_has_rtgroups(mp))
423 return -EINVAL;
424
425 rtg = xfs_rtgroup_get(mp, rgeo.rg_number);
426 if (!rtg)
427 return -EINVAL;
428
429 error = xfs_rtgroup_get_geometry(rtg, &rgeo);
430 xfs_rtgroup_put(rtg);
431 if (error)
432 return error;
433
434 if (copy_to_user(arg, &rgeo, sizeof(rgeo)))
435 return -EFAULT;
436 return 0;
437}
438
439/*
440 * Linux extended inode flags interface.
441 */
442
443static void
444xfs_fill_fsxattr(
445 struct xfs_inode *ip,
446 int whichfork,
447 struct fileattr *fa)
448{
449 struct xfs_mount *mp = ip->i_mount;
450 struct xfs_ifork *ifp = xfs_ifork_ptr(ip, whichfork);
451
452 fileattr_fill_xflags(fa, xfs_ip2xflags(ip));
453
454 if (ip->i_diflags & XFS_DIFLAG_EXTSIZE) {
455 fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
456 } else if (ip->i_diflags & XFS_DIFLAG_EXTSZINHERIT) {
457 /*
458 * Don't let a misaligned extent size hint on a directory
459 * escape to userspace if it won't pass the setattr checks
460 * later.
461 */
462 if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
463 xfs_extlen_to_rtxmod(mp, ip->i_extsize) > 0) {
464 fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE |
465 FS_XFLAG_EXTSZINHERIT);
466 fa->fsx_extsize = 0;
467 } else {
468 fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
469 }
470 }
471
472 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
473 fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
474 fa->fsx_projid = ip->i_projid;
475 if (ifp && !xfs_need_iread_extents(ifp))
476 fa->fsx_nextents = xfs_iext_count(ifp);
477 else
478 fa->fsx_nextents = xfs_ifork_nextents(ifp);
479}
480
481STATIC int
482xfs_ioc_fsgetxattra(
483 xfs_inode_t *ip,
484 void __user *arg)
485{
486 struct fileattr fa;
487
488 xfs_ilock(ip, XFS_ILOCK_SHARED);
489 xfs_fill_fsxattr(ip, XFS_ATTR_FORK, &fa);
490 xfs_iunlock(ip, XFS_ILOCK_SHARED);
491
492 return copy_fsxattr_to_user(&fa, arg);
493}
494
495int
496xfs_fileattr_get(
497 struct dentry *dentry,
498 struct fileattr *fa)
499{
500 struct xfs_inode *ip = XFS_I(d_inode(dentry));
501
502 if (d_is_special(dentry))
503 return -ENOTTY;
504
505 xfs_ilock(ip, XFS_ILOCK_SHARED);
506 xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
507 xfs_iunlock(ip, XFS_ILOCK_SHARED);
508
509 return 0;
510}
511
512static int
513xfs_ioctl_setattr_xflags(
514 struct xfs_trans *tp,
515 struct xfs_inode *ip,
516 struct fileattr *fa)
517{
518 struct xfs_mount *mp = ip->i_mount;
519 bool rtflag = (fa->fsx_xflags & FS_XFLAG_REALTIME);
520 uint64_t i_flags2;
521
522 if (rtflag != XFS_IS_REALTIME_INODE(ip)) {
523 /* Can't change realtime flag if any extents are allocated. */
524 if (xfs_inode_has_filedata(ip))
525 return -EINVAL;
526
527 /*
528 * If S_DAX is enabled on this file, we can only switch the
529 * device if both support fsdax. We can't update S_DAX because
530 * there might be other threads walking down the access paths.
531 */
532 if (IS_DAX(VFS_I(ip)) &&
533 (mp->m_ddev_targp->bt_daxdev == NULL ||
534 (mp->m_rtdev_targp &&
535 mp->m_rtdev_targp->bt_daxdev == NULL)))
536 return -EINVAL;
537 }
538
539 if (rtflag) {
540 /* If realtime flag is set then must have realtime device */
541 if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
542 xfs_extlen_to_rtxmod(mp, ip->i_extsize))
543 return -EINVAL;
544
545 /* Clear reflink if we are actually able to set the rt flag. */
546 if (xfs_is_reflink_inode(ip))
547 ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
548 }
549
550 /* diflags2 only valid for v3 inodes. */
551 i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
552 if (i_flags2 && !xfs_has_v3inodes(mp))
553 return -EINVAL;
554
555 ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
556 ip->i_diflags2 = i_flags2;
557
558 xfs_diflags_to_iflags(ip, false);
559
560 /*
561 * Make the stable writes flag match that of the device the inode
562 * resides on when flipping the RT flag.
563 */
564 if (rtflag != XFS_IS_REALTIME_INODE(ip) && S_ISREG(VFS_I(ip)->i_mode))
565 xfs_update_stable_writes(ip);
566
567 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
568 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
569 XFS_STATS_INC(mp, xs_ig_attrchg);
570 return 0;
571}
572
573static void
574xfs_ioctl_setattr_prepare_dax(
575 struct xfs_inode *ip,
576 struct fileattr *fa)
577{
578 struct xfs_mount *mp = ip->i_mount;
579 struct inode *inode = VFS_I(ip);
580
581 if (S_ISDIR(inode->i_mode))
582 return;
583
584 if (xfs_has_dax_always(mp) || xfs_has_dax_never(mp))
585 return;
586
587 if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
588 !(ip->i_diflags2 & XFS_DIFLAG2_DAX)) ||
589 (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
590 (ip->i_diflags2 & XFS_DIFLAG2_DAX)))
591 d_mark_dontcache(inode);
592}
593
594/*
595 * Set up the transaction structure for the setattr operation, checking that we
596 * have permission to do so. On success, return a clean transaction and the
597 * inode locked exclusively ready for further operation specific checks. On
598 * failure, return an error without modifying or locking the inode.
599 */
600static struct xfs_trans *
601xfs_ioctl_setattr_get_trans(
602 struct xfs_inode *ip,
603 struct xfs_dquot *pdqp)
604{
605 struct xfs_mount *mp = ip->i_mount;
606 struct xfs_trans *tp;
607 int error = -EROFS;
608
609 if (xfs_is_readonly(mp))
610 goto out_error;
611 error = -EIO;
612 if (xfs_is_shutdown(mp))
613 goto out_error;
614
615 error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
616 has_capability_noaudit(current, CAP_FOWNER), &tp);
617 if (error)
618 goto out_error;
619
620 if (xfs_has_wsync(mp))
621 xfs_trans_set_sync(tp);
622
623 return tp;
624
625out_error:
626 return ERR_PTR(error);
627}
628
629/*
630 * Validate a proposed extent size hint. For regular files, the hint can only
631 * be changed if no extents are allocated.
632 */
633static int
634xfs_ioctl_setattr_check_extsize(
635 struct xfs_inode *ip,
636 struct fileattr *fa)
637{
638 struct xfs_mount *mp = ip->i_mount;
639 xfs_failaddr_t failaddr;
640 uint16_t new_diflags;
641
642 if (!fa->fsx_valid)
643 return 0;
644
645 if (S_ISREG(VFS_I(ip)->i_mode) && xfs_inode_has_filedata(ip) &&
646 XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
647 return -EINVAL;
648
649 if (fa->fsx_extsize & mp->m_blockmask)
650 return -EINVAL;
651
652 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
653
654 /*
655 * Inode verifiers do not check that the extent size hint is an integer
656 * multiple of the rt extent size on a directory with both rtinherit
657 * and extszinherit flags set. Don't let sysadmins misconfigure
658 * directories.
659 */
660 if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
661 (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
662 unsigned int rtextsize_bytes;
663
664 rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
665 if (fa->fsx_extsize % rtextsize_bytes)
666 return -EINVAL;
667 }
668
669 failaddr = xfs_inode_validate_extsize(ip->i_mount,
670 XFS_B_TO_FSB(mp, fa->fsx_extsize),
671 VFS_I(ip)->i_mode, new_diflags);
672 return failaddr != NULL ? -EINVAL : 0;
673}
674
675static int
676xfs_ioctl_setattr_check_cowextsize(
677 struct xfs_inode *ip,
678 struct fileattr *fa)
679{
680 struct xfs_mount *mp = ip->i_mount;
681 xfs_failaddr_t failaddr;
682 uint64_t new_diflags2;
683 uint16_t new_diflags;
684
685 if (!fa->fsx_valid)
686 return 0;
687
688 if (fa->fsx_cowextsize & mp->m_blockmask)
689 return -EINVAL;
690
691 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
692 new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
693
694 failaddr = xfs_inode_validate_cowextsize(ip->i_mount,
695 XFS_B_TO_FSB(mp, fa->fsx_cowextsize),
696 VFS_I(ip)->i_mode, new_diflags, new_diflags2);
697 return failaddr != NULL ? -EINVAL : 0;
698}
699
700static int
701xfs_ioctl_setattr_check_projid(
702 struct xfs_inode *ip,
703 struct fileattr *fa)
704{
705 if (!fa->fsx_valid)
706 return 0;
707
708 /* Disallow 32bit project ids if 32bit IDs are not enabled. */
709 if (fa->fsx_projid > (uint16_t)-1 &&
710 !xfs_has_projid32(ip->i_mount))
711 return -EINVAL;
712 return 0;
713}
714
715int
716xfs_fileattr_set(
717 struct mnt_idmap *idmap,
718 struct dentry *dentry,
719 struct fileattr *fa)
720{
721 struct xfs_inode *ip = XFS_I(d_inode(dentry));
722 struct xfs_mount *mp = ip->i_mount;
723 struct xfs_trans *tp;
724 struct xfs_dquot *pdqp = NULL;
725 struct xfs_dquot *olddquot = NULL;
726 int error;
727
728 trace_xfs_ioctl_setattr(ip);
729
730 if (d_is_special(dentry))
731 return -ENOTTY;
732
733 if (!fa->fsx_valid) {
734 if (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL |
735 FS_NOATIME_FL | FS_NODUMP_FL |
736 FS_SYNC_FL | FS_DAX_FL | FS_PROJINHERIT_FL))
737 return -EOPNOTSUPP;
738 }
739
740 error = xfs_ioctl_setattr_check_projid(ip, fa);
741 if (error)
742 return error;
743
744 /*
745 * If disk quotas is on, we make sure that the dquots do exist on disk,
746 * before we start any other transactions. Trying to do this later
747 * is messy. We don't care to take a readlock to look at the ids
748 * in inode here, because we can't hold it across the trans_reserve.
749 * If the IDs do change before we take the ilock, we're covered
750 * because the i_*dquot fields will get updated anyway.
751 */
752 if (fa->fsx_valid && XFS_IS_QUOTA_ON(mp)) {
753 error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
754 VFS_I(ip)->i_gid, fa->fsx_projid,
755 XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
756 if (error)
757 return error;
758 }
759
760 xfs_ioctl_setattr_prepare_dax(ip, fa);
761
762 tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
763 if (IS_ERR(tp)) {
764 error = PTR_ERR(tp);
765 goto error_free_dquots;
766 }
767
768 error = xfs_ioctl_setattr_check_extsize(ip, fa);
769 if (error)
770 goto error_trans_cancel;
771
772 error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
773 if (error)
774 goto error_trans_cancel;
775
776 error = xfs_ioctl_setattr_xflags(tp, ip, fa);
777 if (error)
778 goto error_trans_cancel;
779
780 if (!fa->fsx_valid)
781 goto skip_xattr;
782 /*
783 * Change file ownership. Must be the owner or privileged. CAP_FSETID
784 * overrides the following restrictions:
785 *
786 * The set-user-ID and set-group-ID bits of a file will be cleared upon
787 * successful return from chown()
788 */
789
790 if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
791 !capable_wrt_inode_uidgid(idmap, VFS_I(ip), CAP_FSETID))
792 VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
793
794 /* Change the ownerships and register project quota modifications */
795 if (ip->i_projid != fa->fsx_projid) {
796 if (XFS_IS_PQUOTA_ON(mp)) {
797 olddquot = xfs_qm_vop_chown(tp, ip,
798 &ip->i_pdquot, pdqp);
799 }
800 ip->i_projid = fa->fsx_projid;
801 }
802
803 /*
804 * Only set the extent size hint if we've already determined that the
805 * extent size hint should be set on the inode. If no extent size flags
806 * are set on the inode then unconditionally clear the extent size hint.
807 */
808 if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
809 ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
810 else
811 ip->i_extsize = 0;
812
813 if (xfs_has_v3inodes(mp)) {
814 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
815 ip->i_cowextsize = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
816 else
817 ip->i_cowextsize = 0;
818 }
819
820skip_xattr:
821 error = xfs_trans_commit(tp);
822
823 /*
824 * Release any dquot(s) the inode had kept before chown.
825 */
826 xfs_qm_dqrele(olddquot);
827 xfs_qm_dqrele(pdqp);
828
829 return error;
830
831error_trans_cancel:
832 xfs_trans_cancel(tp);
833error_free_dquots:
834 xfs_qm_dqrele(pdqp);
835 return error;
836}
837
838static bool
839xfs_getbmap_format(
840 struct kgetbmap *p,
841 struct getbmapx __user *u,
842 size_t recsize)
843{
844 if (put_user(p->bmv_offset, &u->bmv_offset) ||
845 put_user(p->bmv_block, &u->bmv_block) ||
846 put_user(p->bmv_length, &u->bmv_length) ||
847 put_user(0, &u->bmv_count) ||
848 put_user(0, &u->bmv_entries))
849 return false;
850 if (recsize < sizeof(struct getbmapx))
851 return true;
852 if (put_user(0, &u->bmv_iflags) ||
853 put_user(p->bmv_oflags, &u->bmv_oflags) ||
854 put_user(0, &u->bmv_unused1) ||
855 put_user(0, &u->bmv_unused2))
856 return false;
857 return true;
858}
859
860STATIC int
861xfs_ioc_getbmap(
862 struct file *file,
863 unsigned int cmd,
864 void __user *arg)
865{
866 struct getbmapx bmx = { 0 };
867 struct kgetbmap *buf;
868 size_t recsize;
869 int error, i;
870
871 switch (cmd) {
872 case XFS_IOC_GETBMAPA:
873 bmx.bmv_iflags = BMV_IF_ATTRFORK;
874 fallthrough;
875 case XFS_IOC_GETBMAP:
876 /* struct getbmap is a strict subset of struct getbmapx. */
877 recsize = sizeof(struct getbmap);
878 break;
879 case XFS_IOC_GETBMAPX:
880 recsize = sizeof(struct getbmapx);
881 break;
882 default:
883 return -EINVAL;
884 }
885
886 if (copy_from_user(&bmx, arg, recsize))
887 return -EFAULT;
888
889 if (bmx.bmv_count < 2)
890 return -EINVAL;
891 if (bmx.bmv_count >= INT_MAX / recsize)
892 return -ENOMEM;
893
894 buf = kvcalloc(bmx.bmv_count, sizeof(*buf), GFP_KERNEL);
895 if (!buf)
896 return -ENOMEM;
897
898 error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
899 if (error)
900 goto out_free_buf;
901
902 error = -EFAULT;
903 if (copy_to_user(arg, &bmx, recsize))
904 goto out_free_buf;
905 arg += recsize;
906
907 for (i = 0; i < bmx.bmv_entries; i++) {
908 if (!xfs_getbmap_format(buf + i, arg, recsize))
909 goto out_free_buf;
910 arg += recsize;
911 }
912
913 error = 0;
914out_free_buf:
915 kvfree(buf);
916 return error;
917}
918
919int
920xfs_ioc_swapext(
921 xfs_swapext_t *sxp)
922{
923 xfs_inode_t *ip, *tip;
924
925 /* Pull information for the target fd */
926 CLASS(fd, f)((int)sxp->sx_fdtarget);
927 if (fd_empty(f))
928 return -EINVAL;
929
930 if (!(fd_file(f)->f_mode & FMODE_WRITE) ||
931 !(fd_file(f)->f_mode & FMODE_READ) ||
932 (fd_file(f)->f_flags & O_APPEND))
933 return -EBADF;
934
935 CLASS(fd, tmp)((int)sxp->sx_fdtmp);
936 if (fd_empty(tmp))
937 return -EINVAL;
938
939 if (!(fd_file(tmp)->f_mode & FMODE_WRITE) ||
940 !(fd_file(tmp)->f_mode & FMODE_READ) ||
941 (fd_file(tmp)->f_flags & O_APPEND))
942 return -EBADF;
943
944 if (IS_SWAPFILE(file_inode(fd_file(f))) ||
945 IS_SWAPFILE(file_inode(fd_file(tmp))))
946 return -EINVAL;
947
948 /*
949 * We need to ensure that the fds passed in point to XFS inodes
950 * before we cast and access them as XFS structures as we have no
951 * control over what the user passes us here.
952 */
953 if (fd_file(f)->f_op != &xfs_file_operations ||
954 fd_file(tmp)->f_op != &xfs_file_operations)
955 return -EINVAL;
956
957 ip = XFS_I(file_inode(fd_file(f)));
958 tip = XFS_I(file_inode(fd_file(tmp)));
959
960 if (ip->i_mount != tip->i_mount)
961 return -EINVAL;
962
963 if (ip->i_ino == tip->i_ino)
964 return -EINVAL;
965
966 if (xfs_is_shutdown(ip->i_mount))
967 return -EIO;
968
969 return xfs_swap_extents(ip, tip, sxp);
970}
971
972static int
973xfs_ioc_getlabel(
974 struct xfs_mount *mp,
975 char __user *user_label)
976{
977 struct xfs_sb *sbp = &mp->m_sb;
978 char label[XFSLABEL_MAX + 1];
979
980 /* Paranoia */
981 BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
982
983 /* 1 larger than sb_fname, so this ensures a trailing NUL char */
984 memset(label, 0, sizeof(label));
985 spin_lock(&mp->m_sb_lock);
986 strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
987 spin_unlock(&mp->m_sb_lock);
988
989 if (copy_to_user(user_label, label, sizeof(label)))
990 return -EFAULT;
991 return 0;
992}
993
994static int
995xfs_ioc_setlabel(
996 struct file *filp,
997 struct xfs_mount *mp,
998 char __user *newlabel)
999{
1000 struct xfs_sb *sbp = &mp->m_sb;
1001 char label[XFSLABEL_MAX + 1];
1002 size_t len;
1003 int error;
1004
1005 if (!capable(CAP_SYS_ADMIN))
1006 return -EPERM;
1007 /*
1008 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1009 * smaller, at 12 bytes. We copy one more to be sure we find the
1010 * (required) NULL character to test the incoming label length.
1011 * NB: The on disk label doesn't need to be null terminated.
1012 */
1013 if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1014 return -EFAULT;
1015 len = strnlen(label, XFSLABEL_MAX + 1);
1016 if (len > sizeof(sbp->sb_fname))
1017 return -EINVAL;
1018
1019 error = mnt_want_write_file(filp);
1020 if (error)
1021 return error;
1022
1023 spin_lock(&mp->m_sb_lock);
1024 memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1025 memcpy(sbp->sb_fname, label, len);
1026 spin_unlock(&mp->m_sb_lock);
1027
1028 /*
1029 * Now we do several things to satisfy userspace.
1030 * In addition to normal logging of the primary superblock, we also
1031 * immediately write these changes to sector zero for the primary, then
1032 * update all backup supers (as xfs_db does for a label change), then
1033 * invalidate the block device page cache. This is so that any prior
1034 * buffered reads from userspace (i.e. from blkid) are invalidated,
1035 * and userspace will see the newly-written label.
1036 */
1037 error = xfs_sync_sb_buf(mp, true);
1038 if (error)
1039 goto out;
1040 /*
1041 * growfs also updates backup supers so lock against that.
1042 */
1043 mutex_lock(&mp->m_growlock);
1044 error = xfs_update_secondary_sbs(mp);
1045 mutex_unlock(&mp->m_growlock);
1046
1047 invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1048 if (xfs_has_rtsb(mp) && mp->m_rtdev_targp)
1049 invalidate_bdev(mp->m_rtdev_targp->bt_bdev);
1050
1051out:
1052 mnt_drop_write_file(filp);
1053 return error;
1054}
1055
1056static inline int
1057xfs_fs_eofblocks_from_user(
1058 struct xfs_fs_eofblocks *src,
1059 struct xfs_icwalk *dst)
1060{
1061 if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1062 return -EINVAL;
1063
1064 if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1065 return -EINVAL;
1066
1067 if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1068 memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1069 return -EINVAL;
1070
1071 dst->icw_flags = 0;
1072 if (src->eof_flags & XFS_EOF_FLAGS_SYNC)
1073 dst->icw_flags |= XFS_ICWALK_FLAG_SYNC;
1074 if (src->eof_flags & XFS_EOF_FLAGS_UID)
1075 dst->icw_flags |= XFS_ICWALK_FLAG_UID;
1076 if (src->eof_flags & XFS_EOF_FLAGS_GID)
1077 dst->icw_flags |= XFS_ICWALK_FLAG_GID;
1078 if (src->eof_flags & XFS_EOF_FLAGS_PRID)
1079 dst->icw_flags |= XFS_ICWALK_FLAG_PRID;
1080 if (src->eof_flags & XFS_EOF_FLAGS_MINFILESIZE)
1081 dst->icw_flags |= XFS_ICWALK_FLAG_MINFILESIZE;
1082
1083 dst->icw_prid = src->eof_prid;
1084 dst->icw_min_file_size = src->eof_min_file_size;
1085
1086 dst->icw_uid = INVALID_UID;
1087 if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1088 dst->icw_uid = make_kuid(current_user_ns(), src->eof_uid);
1089 if (!uid_valid(dst->icw_uid))
1090 return -EINVAL;
1091 }
1092
1093 dst->icw_gid = INVALID_GID;
1094 if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1095 dst->icw_gid = make_kgid(current_user_ns(), src->eof_gid);
1096 if (!gid_valid(dst->icw_gid))
1097 return -EINVAL;
1098 }
1099 return 0;
1100}
1101
1102static int
1103xfs_ioctl_getset_resblocks(
1104 struct file *filp,
1105 unsigned int cmd,
1106 void __user *arg)
1107{
1108 struct xfs_mount *mp = XFS_I(file_inode(filp))->i_mount;
1109 struct xfs_fsop_resblks fsop = { };
1110 int error;
1111
1112 if (!capable(CAP_SYS_ADMIN))
1113 return -EPERM;
1114
1115 if (cmd == XFS_IOC_SET_RESBLKS) {
1116 if (xfs_is_readonly(mp))
1117 return -EROFS;
1118
1119 if (copy_from_user(&fsop, arg, sizeof(fsop)))
1120 return -EFAULT;
1121
1122 error = mnt_want_write_file(filp);
1123 if (error)
1124 return error;
1125 error = xfs_reserve_blocks(mp, fsop.resblks);
1126 mnt_drop_write_file(filp);
1127 if (error)
1128 return error;
1129 }
1130
1131 spin_lock(&mp->m_sb_lock);
1132 fsop.resblks = mp->m_resblks;
1133 fsop.resblks_avail = mp->m_resblks_avail;
1134 spin_unlock(&mp->m_sb_lock);
1135
1136 if (copy_to_user(arg, &fsop, sizeof(fsop)))
1137 return -EFAULT;
1138 return 0;
1139}
1140
1141static int
1142xfs_ioctl_fs_counts(
1143 struct xfs_mount *mp,
1144 struct xfs_fsop_counts __user *uarg)
1145{
1146 struct xfs_fsop_counts out = {
1147 .allocino = percpu_counter_read_positive(&mp->m_icount),
1148 .freeino = percpu_counter_read_positive(&mp->m_ifree),
1149 .freedata = percpu_counter_read_positive(&mp->m_fdblocks) -
1150 xfs_fdblocks_unavailable(mp),
1151 .freertx = percpu_counter_read_positive(&mp->m_frextents),
1152 };
1153
1154 if (copy_to_user(uarg, &out, sizeof(out)))
1155 return -EFAULT;
1156 return 0;
1157}
1158
1159/*
1160 * These long-unused ioctls were removed from the official ioctl API in 5.17,
1161 * but retain these definitions so that we can log warnings about them.
1162 */
1163#define XFS_IOC_ALLOCSP _IOW ('X', 10, struct xfs_flock64)
1164#define XFS_IOC_FREESP _IOW ('X', 11, struct xfs_flock64)
1165#define XFS_IOC_ALLOCSP64 _IOW ('X', 36, struct xfs_flock64)
1166#define XFS_IOC_FREESP64 _IOW ('X', 37, struct xfs_flock64)
1167
1168/*
1169 * Note: some of the ioctl's return positive numbers as a
1170 * byte count indicating success, such as readlink_by_handle.
1171 * So we don't "sign flip" like most other routines. This means
1172 * true errors need to be returned as a negative value.
1173 */
1174long
1175xfs_file_ioctl(
1176 struct file *filp,
1177 unsigned int cmd,
1178 unsigned long p)
1179{
1180 struct inode *inode = file_inode(filp);
1181 struct xfs_inode *ip = XFS_I(inode);
1182 struct xfs_mount *mp = ip->i_mount;
1183 void __user *arg = (void __user *)p;
1184 int error;
1185
1186 trace_xfs_file_ioctl(ip);
1187
1188 switch (cmd) {
1189 case FITRIM:
1190 return xfs_ioc_trim(mp, arg);
1191 case FS_IOC_GETFSLABEL:
1192 return xfs_ioc_getlabel(mp, arg);
1193 case FS_IOC_SETFSLABEL:
1194 return xfs_ioc_setlabel(filp, mp, arg);
1195 case XFS_IOC_ALLOCSP:
1196 case XFS_IOC_FREESP:
1197 case XFS_IOC_ALLOCSP64:
1198 case XFS_IOC_FREESP64:
1199 xfs_warn_once(mp,
1200 "%s should use fallocate; XFS_IOC_{ALLOC,FREE}SP ioctl unsupported",
1201 current->comm);
1202 return -ENOTTY;
1203 case XFS_IOC_DIOINFO: {
1204 struct xfs_buftarg *target = xfs_inode_buftarg(ip);
1205 struct dioattr da;
1206
1207 da.d_mem = da.d_miniosz = target->bt_logical_sectorsize;
1208 da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1209
1210 if (copy_to_user(arg, &da, sizeof(da)))
1211 return -EFAULT;
1212 return 0;
1213 }
1214
1215 case XFS_IOC_FSBULKSTAT_SINGLE:
1216 case XFS_IOC_FSBULKSTAT:
1217 case XFS_IOC_FSINUMBERS:
1218 return xfs_ioc_fsbulkstat(filp, cmd, arg);
1219
1220 case XFS_IOC_BULKSTAT:
1221 return xfs_ioc_bulkstat(filp, cmd, arg);
1222 case XFS_IOC_INUMBERS:
1223 return xfs_ioc_inumbers(mp, cmd, arg);
1224
1225 case XFS_IOC_FSGEOMETRY_V1:
1226 return xfs_ioc_fsgeometry(mp, arg, 3);
1227 case XFS_IOC_FSGEOMETRY_V4:
1228 return xfs_ioc_fsgeometry(mp, arg, 4);
1229 case XFS_IOC_FSGEOMETRY:
1230 return xfs_ioc_fsgeometry(mp, arg, 5);
1231
1232 case XFS_IOC_AG_GEOMETRY:
1233 return xfs_ioc_ag_geometry(mp, arg);
1234 case XFS_IOC_RTGROUP_GEOMETRY:
1235 return xfs_ioc_rtgroup_geometry(mp, arg);
1236
1237 case XFS_IOC_GETVERSION:
1238 return put_user(inode->i_generation, (int __user *)arg);
1239
1240 case XFS_IOC_FSGETXATTRA:
1241 return xfs_ioc_fsgetxattra(ip, arg);
1242 case XFS_IOC_GETPARENTS:
1243 return xfs_ioc_getparents(filp, arg);
1244 case XFS_IOC_GETPARENTS_BY_HANDLE:
1245 return xfs_ioc_getparents_by_handle(filp, arg);
1246 case XFS_IOC_GETBMAP:
1247 case XFS_IOC_GETBMAPA:
1248 case XFS_IOC_GETBMAPX:
1249 return xfs_ioc_getbmap(filp, cmd, arg);
1250
1251 case FS_IOC_GETFSMAP:
1252 return xfs_ioc_getfsmap(ip, arg);
1253
1254 case XFS_IOC_SCRUBV_METADATA:
1255 return xfs_ioc_scrubv_metadata(filp, arg);
1256 case XFS_IOC_SCRUB_METADATA:
1257 return xfs_ioc_scrub_metadata(filp, arg);
1258
1259 case XFS_IOC_FD_TO_HANDLE:
1260 case XFS_IOC_PATH_TO_HANDLE:
1261 case XFS_IOC_PATH_TO_FSHANDLE: {
1262 xfs_fsop_handlereq_t hreq;
1263
1264 if (copy_from_user(&hreq, arg, sizeof(hreq)))
1265 return -EFAULT;
1266 return xfs_find_handle(cmd, &hreq);
1267 }
1268 case XFS_IOC_OPEN_BY_HANDLE: {
1269 xfs_fsop_handlereq_t hreq;
1270
1271 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1272 return -EFAULT;
1273 return xfs_open_by_handle(filp, &hreq);
1274 }
1275
1276 case XFS_IOC_READLINK_BY_HANDLE: {
1277 xfs_fsop_handlereq_t hreq;
1278
1279 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1280 return -EFAULT;
1281 return xfs_readlink_by_handle(filp, &hreq);
1282 }
1283 case XFS_IOC_ATTRLIST_BY_HANDLE:
1284 return xfs_attrlist_by_handle(filp, arg);
1285
1286 case XFS_IOC_ATTRMULTI_BY_HANDLE:
1287 return xfs_attrmulti_by_handle(filp, arg);
1288
1289 case XFS_IOC_SWAPEXT: {
1290 struct xfs_swapext sxp;
1291
1292 if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1293 return -EFAULT;
1294 error = mnt_want_write_file(filp);
1295 if (error)
1296 return error;
1297 error = xfs_ioc_swapext(&sxp);
1298 mnt_drop_write_file(filp);
1299 return error;
1300 }
1301
1302 case XFS_IOC_FSCOUNTS:
1303 return xfs_ioctl_fs_counts(mp, arg);
1304
1305 case XFS_IOC_SET_RESBLKS:
1306 case XFS_IOC_GET_RESBLKS:
1307 return xfs_ioctl_getset_resblocks(filp, cmd, arg);
1308
1309 case XFS_IOC_FSGROWFSDATA: {
1310 struct xfs_growfs_data in;
1311
1312 if (copy_from_user(&in, arg, sizeof(in)))
1313 return -EFAULT;
1314
1315 error = mnt_want_write_file(filp);
1316 if (error)
1317 return error;
1318 error = xfs_growfs_data(mp, &in);
1319 mnt_drop_write_file(filp);
1320 return error;
1321 }
1322
1323 case XFS_IOC_FSGROWFSLOG: {
1324 struct xfs_growfs_log in;
1325
1326 if (copy_from_user(&in, arg, sizeof(in)))
1327 return -EFAULT;
1328
1329 error = mnt_want_write_file(filp);
1330 if (error)
1331 return error;
1332 error = xfs_growfs_log(mp, &in);
1333 mnt_drop_write_file(filp);
1334 return error;
1335 }
1336
1337 case XFS_IOC_FSGROWFSRT: {
1338 xfs_growfs_rt_t in;
1339
1340 if (copy_from_user(&in, arg, sizeof(in)))
1341 return -EFAULT;
1342
1343 error = mnt_want_write_file(filp);
1344 if (error)
1345 return error;
1346 error = xfs_growfs_rt(mp, &in);
1347 mnt_drop_write_file(filp);
1348 return error;
1349 }
1350
1351 case XFS_IOC_GOINGDOWN: {
1352 uint32_t in;
1353
1354 if (!capable(CAP_SYS_ADMIN))
1355 return -EPERM;
1356
1357 if (get_user(in, (uint32_t __user *)arg))
1358 return -EFAULT;
1359
1360 return xfs_fs_goingdown(mp, in);
1361 }
1362
1363 case XFS_IOC_ERROR_INJECTION: {
1364 xfs_error_injection_t in;
1365
1366 if (!capable(CAP_SYS_ADMIN))
1367 return -EPERM;
1368
1369 if (copy_from_user(&in, arg, sizeof(in)))
1370 return -EFAULT;
1371
1372 return xfs_errortag_add(mp, in.errtag);
1373 }
1374
1375 case XFS_IOC_ERROR_CLEARALL:
1376 if (!capable(CAP_SYS_ADMIN))
1377 return -EPERM;
1378
1379 return xfs_errortag_clearall(mp);
1380
1381 case XFS_IOC_FREE_EOFBLOCKS: {
1382 struct xfs_fs_eofblocks eofb;
1383 struct xfs_icwalk icw;
1384
1385 if (!capable(CAP_SYS_ADMIN))
1386 return -EPERM;
1387
1388 if (xfs_is_readonly(mp))
1389 return -EROFS;
1390
1391 if (copy_from_user(&eofb, arg, sizeof(eofb)))
1392 return -EFAULT;
1393
1394 error = xfs_fs_eofblocks_from_user(&eofb, &icw);
1395 if (error)
1396 return error;
1397
1398 trace_xfs_ioc_free_eofblocks(mp, &icw, _RET_IP_);
1399
1400 sb_start_write(mp->m_super);
1401 error = xfs_blockgc_free_space(mp, &icw);
1402 sb_end_write(mp->m_super);
1403 return error;
1404 }
1405
1406 case XFS_IOC_EXCHANGE_RANGE:
1407 return xfs_ioc_exchange_range(filp, arg);
1408 case XFS_IOC_START_COMMIT:
1409 return xfs_ioc_start_commit(filp, arg);
1410 case XFS_IOC_COMMIT_RANGE:
1411 return xfs_ioc_commit_range(filp, arg);
1412
1413 default:
1414 return -ENOTTY;
1415 }
1416}
1/*
2 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
3 * All Rights Reserved.
4 *
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18#include "xfs.h"
19#include "xfs_fs.h"
20#include "xfs_shared.h"
21#include "xfs_format.h"
22#include "xfs_log_format.h"
23#include "xfs_trans_resv.h"
24#include "xfs_mount.h"
25#include "xfs_inode.h"
26#include "xfs_ioctl.h"
27#include "xfs_alloc.h"
28#include "xfs_rtalloc.h"
29#include "xfs_itable.h"
30#include "xfs_error.h"
31#include "xfs_attr.h"
32#include "xfs_bmap.h"
33#include "xfs_bmap_util.h"
34#include "xfs_fsops.h"
35#include "xfs_discard.h"
36#include "xfs_quota.h"
37#include "xfs_export.h"
38#include "xfs_trace.h"
39#include "xfs_icache.h"
40#include "xfs_symlink.h"
41#include "xfs_trans.h"
42#include "xfs_pnfs.h"
43#include "xfs_acl.h"
44
45#include <linux/capability.h>
46#include <linux/dcache.h>
47#include <linux/mount.h>
48#include <linux/namei.h>
49#include <linux/pagemap.h>
50#include <linux/slab.h>
51#include <linux/exportfs.h>
52
53/*
54 * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
55 * a file or fs handle.
56 *
57 * XFS_IOC_PATH_TO_FSHANDLE
58 * returns fs handle for a mount point or path within that mount point
59 * XFS_IOC_FD_TO_HANDLE
60 * returns full handle for a FD opened in user space
61 * XFS_IOC_PATH_TO_HANDLE
62 * returns full handle for a path
63 */
64int
65xfs_find_handle(
66 unsigned int cmd,
67 xfs_fsop_handlereq_t *hreq)
68{
69 int hsize;
70 xfs_handle_t handle;
71 struct inode *inode;
72 struct fd f = {NULL};
73 struct path path;
74 int error;
75 struct xfs_inode *ip;
76
77 if (cmd == XFS_IOC_FD_TO_HANDLE) {
78 f = fdget(hreq->fd);
79 if (!f.file)
80 return -EBADF;
81 inode = file_inode(f.file);
82 } else {
83 error = user_lpath((const char __user *)hreq->path, &path);
84 if (error)
85 return error;
86 inode = d_inode(path.dentry);
87 }
88 ip = XFS_I(inode);
89
90 /*
91 * We can only generate handles for inodes residing on a XFS filesystem,
92 * and only for regular files, directories or symbolic links.
93 */
94 error = -EINVAL;
95 if (inode->i_sb->s_magic != XFS_SB_MAGIC)
96 goto out_put;
97
98 error = -EBADF;
99 if (!S_ISREG(inode->i_mode) &&
100 !S_ISDIR(inode->i_mode) &&
101 !S_ISLNK(inode->i_mode))
102 goto out_put;
103
104
105 memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
106
107 if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
108 /*
109 * This handle only contains an fsid, zero the rest.
110 */
111 memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
112 hsize = sizeof(xfs_fsid_t);
113 } else {
114 handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
115 sizeof(handle.ha_fid.fid_len);
116 handle.ha_fid.fid_pad = 0;
117 handle.ha_fid.fid_gen = inode->i_generation;
118 handle.ha_fid.fid_ino = ip->i_ino;
119
120 hsize = XFS_HSIZE(handle);
121 }
122
123 error = -EFAULT;
124 if (copy_to_user(hreq->ohandle, &handle, hsize) ||
125 copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
126 goto out_put;
127
128 error = 0;
129
130 out_put:
131 if (cmd == XFS_IOC_FD_TO_HANDLE)
132 fdput(f);
133 else
134 path_put(&path);
135 return error;
136}
137
138/*
139 * No need to do permission checks on the various pathname components
140 * as the handle operations are privileged.
141 */
142STATIC int
143xfs_handle_acceptable(
144 void *context,
145 struct dentry *dentry)
146{
147 return 1;
148}
149
150/*
151 * Convert userspace handle data into a dentry.
152 */
153struct dentry *
154xfs_handle_to_dentry(
155 struct file *parfilp,
156 void __user *uhandle,
157 u32 hlen)
158{
159 xfs_handle_t handle;
160 struct xfs_fid64 fid;
161
162 /*
163 * Only allow handle opens under a directory.
164 */
165 if (!S_ISDIR(file_inode(parfilp)->i_mode))
166 return ERR_PTR(-ENOTDIR);
167
168 if (hlen != sizeof(xfs_handle_t))
169 return ERR_PTR(-EINVAL);
170 if (copy_from_user(&handle, uhandle, hlen))
171 return ERR_PTR(-EFAULT);
172 if (handle.ha_fid.fid_len !=
173 sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
174 return ERR_PTR(-EINVAL);
175
176 memset(&fid, 0, sizeof(struct fid));
177 fid.ino = handle.ha_fid.fid_ino;
178 fid.gen = handle.ha_fid.fid_gen;
179
180 return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
181 FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
182 xfs_handle_acceptable, NULL);
183}
184
185STATIC struct dentry *
186xfs_handlereq_to_dentry(
187 struct file *parfilp,
188 xfs_fsop_handlereq_t *hreq)
189{
190 return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
191}
192
193int
194xfs_open_by_handle(
195 struct file *parfilp,
196 xfs_fsop_handlereq_t *hreq)
197{
198 const struct cred *cred = current_cred();
199 int error;
200 int fd;
201 int permflag;
202 struct file *filp;
203 struct inode *inode;
204 struct dentry *dentry;
205 fmode_t fmode;
206 struct path path;
207
208 if (!capable(CAP_SYS_ADMIN))
209 return -EPERM;
210
211 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
212 if (IS_ERR(dentry))
213 return PTR_ERR(dentry);
214 inode = d_inode(dentry);
215
216 /* Restrict xfs_open_by_handle to directories & regular files. */
217 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
218 error = -EPERM;
219 goto out_dput;
220 }
221
222#if BITS_PER_LONG != 32
223 hreq->oflags |= O_LARGEFILE;
224#endif
225
226 permflag = hreq->oflags;
227 fmode = OPEN_FMODE(permflag);
228 if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
229 (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
230 error = -EPERM;
231 goto out_dput;
232 }
233
234 if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
235 error = -EACCES;
236 goto out_dput;
237 }
238
239 /* Can't write directories. */
240 if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
241 error = -EISDIR;
242 goto out_dput;
243 }
244
245 fd = get_unused_fd_flags(0);
246 if (fd < 0) {
247 error = fd;
248 goto out_dput;
249 }
250
251 path.mnt = parfilp->f_path.mnt;
252 path.dentry = dentry;
253 filp = dentry_open(&path, hreq->oflags, cred);
254 dput(dentry);
255 if (IS_ERR(filp)) {
256 put_unused_fd(fd);
257 return PTR_ERR(filp);
258 }
259
260 if (S_ISREG(inode->i_mode)) {
261 filp->f_flags |= O_NOATIME;
262 filp->f_mode |= FMODE_NOCMTIME;
263 }
264
265 fd_install(fd, filp);
266 return fd;
267
268 out_dput:
269 dput(dentry);
270 return error;
271}
272
273int
274xfs_readlink_by_handle(
275 struct file *parfilp,
276 xfs_fsop_handlereq_t *hreq)
277{
278 struct dentry *dentry;
279 __u32 olen;
280 void *link;
281 int error;
282
283 if (!capable(CAP_SYS_ADMIN))
284 return -EPERM;
285
286 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
287 if (IS_ERR(dentry))
288 return PTR_ERR(dentry);
289
290 /* Restrict this handle operation to symlinks only. */
291 if (!d_is_symlink(dentry)) {
292 error = -EINVAL;
293 goto out_dput;
294 }
295
296 if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
297 error = -EFAULT;
298 goto out_dput;
299 }
300
301 link = kmalloc(MAXPATHLEN+1, GFP_KERNEL);
302 if (!link) {
303 error = -ENOMEM;
304 goto out_dput;
305 }
306
307 error = xfs_readlink(XFS_I(d_inode(dentry)), link);
308 if (error)
309 goto out_kfree;
310 error = readlink_copy(hreq->ohandle, olen, link);
311 if (error)
312 goto out_kfree;
313
314 out_kfree:
315 kfree(link);
316 out_dput:
317 dput(dentry);
318 return error;
319}
320
321int
322xfs_set_dmattrs(
323 xfs_inode_t *ip,
324 u_int evmask,
325 u_int16_t state)
326{
327 xfs_mount_t *mp = ip->i_mount;
328 xfs_trans_t *tp;
329 int error;
330
331 if (!capable(CAP_SYS_ADMIN))
332 return -EPERM;
333
334 if (XFS_FORCED_SHUTDOWN(mp))
335 return -EIO;
336
337 tp = xfs_trans_alloc(mp, XFS_TRANS_SET_DMATTRS);
338 error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
339 if (error) {
340 xfs_trans_cancel(tp);
341 return error;
342 }
343 xfs_ilock(ip, XFS_ILOCK_EXCL);
344 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
345
346 ip->i_d.di_dmevmask = evmask;
347 ip->i_d.di_dmstate = state;
348
349 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
350 error = xfs_trans_commit(tp);
351
352 return error;
353}
354
355STATIC int
356xfs_fssetdm_by_handle(
357 struct file *parfilp,
358 void __user *arg)
359{
360 int error;
361 struct fsdmidata fsd;
362 xfs_fsop_setdm_handlereq_t dmhreq;
363 struct dentry *dentry;
364
365 if (!capable(CAP_MKNOD))
366 return -EPERM;
367 if (copy_from_user(&dmhreq, arg, sizeof(xfs_fsop_setdm_handlereq_t)))
368 return -EFAULT;
369
370 error = mnt_want_write_file(parfilp);
371 if (error)
372 return error;
373
374 dentry = xfs_handlereq_to_dentry(parfilp, &dmhreq.hreq);
375 if (IS_ERR(dentry)) {
376 mnt_drop_write_file(parfilp);
377 return PTR_ERR(dentry);
378 }
379
380 if (IS_IMMUTABLE(d_inode(dentry)) || IS_APPEND(d_inode(dentry))) {
381 error = -EPERM;
382 goto out;
383 }
384
385 if (copy_from_user(&fsd, dmhreq.data, sizeof(fsd))) {
386 error = -EFAULT;
387 goto out;
388 }
389
390 error = xfs_set_dmattrs(XFS_I(d_inode(dentry)), fsd.fsd_dmevmask,
391 fsd.fsd_dmstate);
392
393 out:
394 mnt_drop_write_file(parfilp);
395 dput(dentry);
396 return error;
397}
398
399STATIC int
400xfs_attrlist_by_handle(
401 struct file *parfilp,
402 void __user *arg)
403{
404 int error = -ENOMEM;
405 attrlist_cursor_kern_t *cursor;
406 xfs_fsop_attrlist_handlereq_t al_hreq;
407 struct dentry *dentry;
408 char *kbuf;
409
410 if (!capable(CAP_SYS_ADMIN))
411 return -EPERM;
412 if (copy_from_user(&al_hreq, arg, sizeof(xfs_fsop_attrlist_handlereq_t)))
413 return -EFAULT;
414 if (al_hreq.buflen < sizeof(struct attrlist) ||
415 al_hreq.buflen > XFS_XATTR_LIST_MAX)
416 return -EINVAL;
417
418 /*
419 * Reject flags, only allow namespaces.
420 */
421 if (al_hreq.flags & ~(ATTR_ROOT | ATTR_SECURE))
422 return -EINVAL;
423
424 dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
425 if (IS_ERR(dentry))
426 return PTR_ERR(dentry);
427
428 kbuf = kmem_zalloc_large(al_hreq.buflen, KM_SLEEP);
429 if (!kbuf)
430 goto out_dput;
431
432 cursor = (attrlist_cursor_kern_t *)&al_hreq.pos;
433 error = xfs_attr_list(XFS_I(d_inode(dentry)), kbuf, al_hreq.buflen,
434 al_hreq.flags, cursor);
435 if (error)
436 goto out_kfree;
437
438 if (copy_to_user(al_hreq.buffer, kbuf, al_hreq.buflen))
439 error = -EFAULT;
440
441out_kfree:
442 kmem_free(kbuf);
443out_dput:
444 dput(dentry);
445 return error;
446}
447
448int
449xfs_attrmulti_attr_get(
450 struct inode *inode,
451 unsigned char *name,
452 unsigned char __user *ubuf,
453 __uint32_t *len,
454 __uint32_t flags)
455{
456 unsigned char *kbuf;
457 int error = -EFAULT;
458
459 if (*len > XFS_XATTR_SIZE_MAX)
460 return -EINVAL;
461 kbuf = kmem_zalloc_large(*len, KM_SLEEP);
462 if (!kbuf)
463 return -ENOMEM;
464
465 error = xfs_attr_get(XFS_I(inode), name, kbuf, (int *)len, flags);
466 if (error)
467 goto out_kfree;
468
469 if (copy_to_user(ubuf, kbuf, *len))
470 error = -EFAULT;
471
472out_kfree:
473 kmem_free(kbuf);
474 return error;
475}
476
477int
478xfs_attrmulti_attr_set(
479 struct inode *inode,
480 unsigned char *name,
481 const unsigned char __user *ubuf,
482 __uint32_t len,
483 __uint32_t flags)
484{
485 unsigned char *kbuf;
486 int error;
487
488 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
489 return -EPERM;
490 if (len > XFS_XATTR_SIZE_MAX)
491 return -EINVAL;
492
493 kbuf = memdup_user(ubuf, len);
494 if (IS_ERR(kbuf))
495 return PTR_ERR(kbuf);
496
497 error = xfs_attr_set(XFS_I(inode), name, kbuf, len, flags);
498 if (!error)
499 xfs_forget_acl(inode, name, flags);
500 kfree(kbuf);
501 return error;
502}
503
504int
505xfs_attrmulti_attr_remove(
506 struct inode *inode,
507 unsigned char *name,
508 __uint32_t flags)
509{
510 int error;
511
512 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
513 return -EPERM;
514 error = xfs_attr_remove(XFS_I(inode), name, flags);
515 if (!error)
516 xfs_forget_acl(inode, name, flags);
517 return error;
518}
519
520STATIC int
521xfs_attrmulti_by_handle(
522 struct file *parfilp,
523 void __user *arg)
524{
525 int error;
526 xfs_attr_multiop_t *ops;
527 xfs_fsop_attrmulti_handlereq_t am_hreq;
528 struct dentry *dentry;
529 unsigned int i, size;
530 unsigned char *attr_name;
531
532 if (!capable(CAP_SYS_ADMIN))
533 return -EPERM;
534 if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
535 return -EFAULT;
536
537 /* overflow check */
538 if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
539 return -E2BIG;
540
541 dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
542 if (IS_ERR(dentry))
543 return PTR_ERR(dentry);
544
545 error = -E2BIG;
546 size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
547 if (!size || size > 16 * PAGE_SIZE)
548 goto out_dput;
549
550 ops = memdup_user(am_hreq.ops, size);
551 if (IS_ERR(ops)) {
552 error = PTR_ERR(ops);
553 goto out_dput;
554 }
555
556 error = -ENOMEM;
557 attr_name = kmalloc(MAXNAMELEN, GFP_KERNEL);
558 if (!attr_name)
559 goto out_kfree_ops;
560
561 error = 0;
562 for (i = 0; i < am_hreq.opcount; i++) {
563 ops[i].am_error = strncpy_from_user((char *)attr_name,
564 ops[i].am_attrname, MAXNAMELEN);
565 if (ops[i].am_error == 0 || ops[i].am_error == MAXNAMELEN)
566 error = -ERANGE;
567 if (ops[i].am_error < 0)
568 break;
569
570 switch (ops[i].am_opcode) {
571 case ATTR_OP_GET:
572 ops[i].am_error = xfs_attrmulti_attr_get(
573 d_inode(dentry), attr_name,
574 ops[i].am_attrvalue, &ops[i].am_length,
575 ops[i].am_flags);
576 break;
577 case ATTR_OP_SET:
578 ops[i].am_error = mnt_want_write_file(parfilp);
579 if (ops[i].am_error)
580 break;
581 ops[i].am_error = xfs_attrmulti_attr_set(
582 d_inode(dentry), attr_name,
583 ops[i].am_attrvalue, ops[i].am_length,
584 ops[i].am_flags);
585 mnt_drop_write_file(parfilp);
586 break;
587 case ATTR_OP_REMOVE:
588 ops[i].am_error = mnt_want_write_file(parfilp);
589 if (ops[i].am_error)
590 break;
591 ops[i].am_error = xfs_attrmulti_attr_remove(
592 d_inode(dentry), attr_name,
593 ops[i].am_flags);
594 mnt_drop_write_file(parfilp);
595 break;
596 default:
597 ops[i].am_error = -EINVAL;
598 }
599 }
600
601 if (copy_to_user(am_hreq.ops, ops, size))
602 error = -EFAULT;
603
604 kfree(attr_name);
605 out_kfree_ops:
606 kfree(ops);
607 out_dput:
608 dput(dentry);
609 return error;
610}
611
612int
613xfs_ioc_space(
614 struct xfs_inode *ip,
615 struct inode *inode,
616 struct file *filp,
617 int ioflags,
618 unsigned int cmd,
619 xfs_flock64_t *bf)
620{
621 struct iattr iattr;
622 enum xfs_prealloc_flags flags = 0;
623 uint iolock = XFS_IOLOCK_EXCL;
624 int error;
625
626 /*
627 * Only allow the sys admin to reserve space unless
628 * unwritten extents are enabled.
629 */
630 if (!xfs_sb_version_hasextflgbit(&ip->i_mount->m_sb) &&
631 !capable(CAP_SYS_ADMIN))
632 return -EPERM;
633
634 if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
635 return -EPERM;
636
637 if (!(filp->f_mode & FMODE_WRITE))
638 return -EBADF;
639
640 if (!S_ISREG(inode->i_mode))
641 return -EINVAL;
642
643 if (filp->f_flags & O_DSYNC)
644 flags |= XFS_PREALLOC_SYNC;
645 if (ioflags & XFS_IO_INVIS)
646 flags |= XFS_PREALLOC_INVISIBLE;
647
648 error = mnt_want_write_file(filp);
649 if (error)
650 return error;
651
652 xfs_ilock(ip, iolock);
653 error = xfs_break_layouts(inode, &iolock, false);
654 if (error)
655 goto out_unlock;
656
657 xfs_ilock(ip, XFS_MMAPLOCK_EXCL);
658 iolock |= XFS_MMAPLOCK_EXCL;
659
660 switch (bf->l_whence) {
661 case 0: /*SEEK_SET*/
662 break;
663 case 1: /*SEEK_CUR*/
664 bf->l_start += filp->f_pos;
665 break;
666 case 2: /*SEEK_END*/
667 bf->l_start += XFS_ISIZE(ip);
668 break;
669 default:
670 error = -EINVAL;
671 goto out_unlock;
672 }
673
674 /*
675 * length of <= 0 for resv/unresv/zero is invalid. length for
676 * alloc/free is ignored completely and we have no idea what userspace
677 * might have set it to, so set it to zero to allow range
678 * checks to pass.
679 */
680 switch (cmd) {
681 case XFS_IOC_ZERO_RANGE:
682 case XFS_IOC_RESVSP:
683 case XFS_IOC_RESVSP64:
684 case XFS_IOC_UNRESVSP:
685 case XFS_IOC_UNRESVSP64:
686 if (bf->l_len <= 0) {
687 error = -EINVAL;
688 goto out_unlock;
689 }
690 break;
691 default:
692 bf->l_len = 0;
693 break;
694 }
695
696 if (bf->l_start < 0 ||
697 bf->l_start > inode->i_sb->s_maxbytes ||
698 bf->l_start + bf->l_len < 0 ||
699 bf->l_start + bf->l_len >= inode->i_sb->s_maxbytes) {
700 error = -EINVAL;
701 goto out_unlock;
702 }
703
704 switch (cmd) {
705 case XFS_IOC_ZERO_RANGE:
706 flags |= XFS_PREALLOC_SET;
707 error = xfs_zero_file_space(ip, bf->l_start, bf->l_len);
708 break;
709 case XFS_IOC_RESVSP:
710 case XFS_IOC_RESVSP64:
711 flags |= XFS_PREALLOC_SET;
712 error = xfs_alloc_file_space(ip, bf->l_start, bf->l_len,
713 XFS_BMAPI_PREALLOC);
714 break;
715 case XFS_IOC_UNRESVSP:
716 case XFS_IOC_UNRESVSP64:
717 error = xfs_free_file_space(ip, bf->l_start, bf->l_len);
718 break;
719 case XFS_IOC_ALLOCSP:
720 case XFS_IOC_ALLOCSP64:
721 case XFS_IOC_FREESP:
722 case XFS_IOC_FREESP64:
723 flags |= XFS_PREALLOC_CLEAR;
724 if (bf->l_start > XFS_ISIZE(ip)) {
725 error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
726 bf->l_start - XFS_ISIZE(ip), 0);
727 if (error)
728 goto out_unlock;
729 }
730
731 iattr.ia_valid = ATTR_SIZE;
732 iattr.ia_size = bf->l_start;
733
734 error = xfs_setattr_size(ip, &iattr);
735 break;
736 default:
737 ASSERT(0);
738 error = -EINVAL;
739 }
740
741 if (error)
742 goto out_unlock;
743
744 error = xfs_update_prealloc_flags(ip, flags);
745
746out_unlock:
747 xfs_iunlock(ip, iolock);
748 mnt_drop_write_file(filp);
749 return error;
750}
751
752STATIC int
753xfs_ioc_bulkstat(
754 xfs_mount_t *mp,
755 unsigned int cmd,
756 void __user *arg)
757{
758 xfs_fsop_bulkreq_t bulkreq;
759 int count; /* # of records returned */
760 xfs_ino_t inlast; /* last inode number */
761 int done;
762 int error;
763
764 /* done = 1 if there are more stats to get and if bulkstat */
765 /* should be called again (unused here, but used in dmapi) */
766
767 if (!capable(CAP_SYS_ADMIN))
768 return -EPERM;
769
770 if (XFS_FORCED_SHUTDOWN(mp))
771 return -EIO;
772
773 if (copy_from_user(&bulkreq, arg, sizeof(xfs_fsop_bulkreq_t)))
774 return -EFAULT;
775
776 if (copy_from_user(&inlast, bulkreq.lastip, sizeof(__s64)))
777 return -EFAULT;
778
779 if ((count = bulkreq.icount) <= 0)
780 return -EINVAL;
781
782 if (bulkreq.ubuffer == NULL)
783 return -EINVAL;
784
785 if (cmd == XFS_IOC_FSINUMBERS)
786 error = xfs_inumbers(mp, &inlast, &count,
787 bulkreq.ubuffer, xfs_inumbers_fmt);
788 else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE)
789 error = xfs_bulkstat_one(mp, inlast, bulkreq.ubuffer,
790 sizeof(xfs_bstat_t), NULL, &done);
791 else /* XFS_IOC_FSBULKSTAT */
792 error = xfs_bulkstat(mp, &inlast, &count, xfs_bulkstat_one,
793 sizeof(xfs_bstat_t), bulkreq.ubuffer,
794 &done);
795
796 if (error)
797 return error;
798
799 if (bulkreq.ocount != NULL) {
800 if (copy_to_user(bulkreq.lastip, &inlast,
801 sizeof(xfs_ino_t)))
802 return -EFAULT;
803
804 if (copy_to_user(bulkreq.ocount, &count, sizeof(count)))
805 return -EFAULT;
806 }
807
808 return 0;
809}
810
811STATIC int
812xfs_ioc_fsgeometry_v1(
813 xfs_mount_t *mp,
814 void __user *arg)
815{
816 xfs_fsop_geom_t fsgeo;
817 int error;
818
819 error = xfs_fs_geometry(mp, &fsgeo, 3);
820 if (error)
821 return error;
822
823 /*
824 * Caller should have passed an argument of type
825 * xfs_fsop_geom_v1_t. This is a proper subset of the
826 * xfs_fsop_geom_t that xfs_fs_geometry() fills in.
827 */
828 if (copy_to_user(arg, &fsgeo, sizeof(xfs_fsop_geom_v1_t)))
829 return -EFAULT;
830 return 0;
831}
832
833STATIC int
834xfs_ioc_fsgeometry(
835 xfs_mount_t *mp,
836 void __user *arg)
837{
838 xfs_fsop_geom_t fsgeo;
839 int error;
840
841 error = xfs_fs_geometry(mp, &fsgeo, 4);
842 if (error)
843 return error;
844
845 if (copy_to_user(arg, &fsgeo, sizeof(fsgeo)))
846 return -EFAULT;
847 return 0;
848}
849
850/*
851 * Linux extended inode flags interface.
852 */
853
854STATIC unsigned int
855xfs_merge_ioc_xflags(
856 unsigned int flags,
857 unsigned int start)
858{
859 unsigned int xflags = start;
860
861 if (flags & FS_IMMUTABLE_FL)
862 xflags |= FS_XFLAG_IMMUTABLE;
863 else
864 xflags &= ~FS_XFLAG_IMMUTABLE;
865 if (flags & FS_APPEND_FL)
866 xflags |= FS_XFLAG_APPEND;
867 else
868 xflags &= ~FS_XFLAG_APPEND;
869 if (flags & FS_SYNC_FL)
870 xflags |= FS_XFLAG_SYNC;
871 else
872 xflags &= ~FS_XFLAG_SYNC;
873 if (flags & FS_NOATIME_FL)
874 xflags |= FS_XFLAG_NOATIME;
875 else
876 xflags &= ~FS_XFLAG_NOATIME;
877 if (flags & FS_NODUMP_FL)
878 xflags |= FS_XFLAG_NODUMP;
879 else
880 xflags &= ~FS_XFLAG_NODUMP;
881
882 return xflags;
883}
884
885STATIC unsigned int
886xfs_di2lxflags(
887 __uint16_t di_flags)
888{
889 unsigned int flags = 0;
890
891 if (di_flags & XFS_DIFLAG_IMMUTABLE)
892 flags |= FS_IMMUTABLE_FL;
893 if (di_flags & XFS_DIFLAG_APPEND)
894 flags |= FS_APPEND_FL;
895 if (di_flags & XFS_DIFLAG_SYNC)
896 flags |= FS_SYNC_FL;
897 if (di_flags & XFS_DIFLAG_NOATIME)
898 flags |= FS_NOATIME_FL;
899 if (di_flags & XFS_DIFLAG_NODUMP)
900 flags |= FS_NODUMP_FL;
901 return flags;
902}
903
904STATIC int
905xfs_ioc_fsgetxattr(
906 xfs_inode_t *ip,
907 int attr,
908 void __user *arg)
909{
910 struct fsxattr fa;
911
912 memset(&fa, 0, sizeof(struct fsxattr));
913
914 xfs_ilock(ip, XFS_ILOCK_SHARED);
915 fa.fsx_xflags = xfs_ip2xflags(ip);
916 fa.fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
917 fa.fsx_projid = xfs_get_projid(ip);
918
919 if (attr) {
920 if (ip->i_afp) {
921 if (ip->i_afp->if_flags & XFS_IFEXTENTS)
922 fa.fsx_nextents = ip->i_afp->if_bytes /
923 sizeof(xfs_bmbt_rec_t);
924 else
925 fa.fsx_nextents = ip->i_d.di_anextents;
926 } else
927 fa.fsx_nextents = 0;
928 } else {
929 if (ip->i_df.if_flags & XFS_IFEXTENTS)
930 fa.fsx_nextents = ip->i_df.if_bytes /
931 sizeof(xfs_bmbt_rec_t);
932 else
933 fa.fsx_nextents = ip->i_d.di_nextents;
934 }
935 xfs_iunlock(ip, XFS_ILOCK_SHARED);
936
937 if (copy_to_user(arg, &fa, sizeof(fa)))
938 return -EFAULT;
939 return 0;
940}
941
942STATIC void
943xfs_set_diflags(
944 struct xfs_inode *ip,
945 unsigned int xflags)
946{
947 unsigned int di_flags;
948 uint64_t di_flags2;
949
950 /* can't set PREALLOC this way, just preserve it */
951 di_flags = (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
952 if (xflags & FS_XFLAG_IMMUTABLE)
953 di_flags |= XFS_DIFLAG_IMMUTABLE;
954 if (xflags & FS_XFLAG_APPEND)
955 di_flags |= XFS_DIFLAG_APPEND;
956 if (xflags & FS_XFLAG_SYNC)
957 di_flags |= XFS_DIFLAG_SYNC;
958 if (xflags & FS_XFLAG_NOATIME)
959 di_flags |= XFS_DIFLAG_NOATIME;
960 if (xflags & FS_XFLAG_NODUMP)
961 di_flags |= XFS_DIFLAG_NODUMP;
962 if (xflags & FS_XFLAG_NODEFRAG)
963 di_flags |= XFS_DIFLAG_NODEFRAG;
964 if (xflags & FS_XFLAG_FILESTREAM)
965 di_flags |= XFS_DIFLAG_FILESTREAM;
966 if (S_ISDIR(VFS_I(ip)->i_mode)) {
967 if (xflags & FS_XFLAG_RTINHERIT)
968 di_flags |= XFS_DIFLAG_RTINHERIT;
969 if (xflags & FS_XFLAG_NOSYMLINKS)
970 di_flags |= XFS_DIFLAG_NOSYMLINKS;
971 if (xflags & FS_XFLAG_EXTSZINHERIT)
972 di_flags |= XFS_DIFLAG_EXTSZINHERIT;
973 if (xflags & FS_XFLAG_PROJINHERIT)
974 di_flags |= XFS_DIFLAG_PROJINHERIT;
975 } else if (S_ISREG(VFS_I(ip)->i_mode)) {
976 if (xflags & FS_XFLAG_REALTIME)
977 di_flags |= XFS_DIFLAG_REALTIME;
978 if (xflags & FS_XFLAG_EXTSIZE)
979 di_flags |= XFS_DIFLAG_EXTSIZE;
980 }
981 ip->i_d.di_flags = di_flags;
982
983 /* diflags2 only valid for v3 inodes. */
984 if (ip->i_d.di_version < 3)
985 return;
986
987 di_flags2 = 0;
988 if (xflags & FS_XFLAG_DAX)
989 di_flags2 |= XFS_DIFLAG2_DAX;
990
991 ip->i_d.di_flags2 = di_flags2;
992
993}
994
995STATIC void
996xfs_diflags_to_linux(
997 struct xfs_inode *ip)
998{
999 struct inode *inode = VFS_I(ip);
1000 unsigned int xflags = xfs_ip2xflags(ip);
1001
1002 if (xflags & FS_XFLAG_IMMUTABLE)
1003 inode->i_flags |= S_IMMUTABLE;
1004 else
1005 inode->i_flags &= ~S_IMMUTABLE;
1006 if (xflags & FS_XFLAG_APPEND)
1007 inode->i_flags |= S_APPEND;
1008 else
1009 inode->i_flags &= ~S_APPEND;
1010 if (xflags & FS_XFLAG_SYNC)
1011 inode->i_flags |= S_SYNC;
1012 else
1013 inode->i_flags &= ~S_SYNC;
1014 if (xflags & FS_XFLAG_NOATIME)
1015 inode->i_flags |= S_NOATIME;
1016 else
1017 inode->i_flags &= ~S_NOATIME;
1018 if (xflags & FS_XFLAG_DAX)
1019 inode->i_flags |= S_DAX;
1020 else
1021 inode->i_flags &= ~S_DAX;
1022
1023}
1024
1025static int
1026xfs_ioctl_setattr_xflags(
1027 struct xfs_trans *tp,
1028 struct xfs_inode *ip,
1029 struct fsxattr *fa)
1030{
1031 struct xfs_mount *mp = ip->i_mount;
1032
1033 /* Can't change realtime flag if any extents are allocated. */
1034 if ((ip->i_d.di_nextents || ip->i_delayed_blks) &&
1035 XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1036 return -EINVAL;
1037
1038 /* If realtime flag is set then must have realtime device */
1039 if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1040 if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1041 (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1042 return -EINVAL;
1043 }
1044
1045 /*
1046 * Can't modify an immutable/append-only file unless
1047 * we have appropriate permission.
1048 */
1049 if (((ip->i_d.di_flags & (XFS_DIFLAG_IMMUTABLE | XFS_DIFLAG_APPEND)) ||
1050 (fa->fsx_xflags & (FS_XFLAG_IMMUTABLE | FS_XFLAG_APPEND))) &&
1051 !capable(CAP_LINUX_IMMUTABLE))
1052 return -EPERM;
1053
1054 xfs_set_diflags(ip, fa->fsx_xflags);
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 */
1070static int
1071xfs_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 int error;
1078
1079 *join_flags = 0;
1080
1081 /*
1082 * It is only valid to set the DAX flag on regular files and
1083 * directories on filesystems where the block size is equal to the page
1084 * size. On directories it serves as an inherit hint.
1085 */
1086 if (fa->fsx_xflags & FS_XFLAG_DAX) {
1087 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode)))
1088 return -EINVAL;
1089 if (ip->i_mount->m_sb.sb_blocksize != PAGE_SIZE)
1090 return -EINVAL;
1091 }
1092
1093 /* If the DAX state is not changing, we have nothing to do here. */
1094 if ((fa->fsx_xflags & FS_XFLAG_DAX) && IS_DAX(inode))
1095 return 0;
1096 if (!(fa->fsx_xflags & FS_XFLAG_DAX) && !IS_DAX(inode))
1097 return 0;
1098
1099 /* lock, flush and invalidate mapping in preparation for flag change */
1100 xfs_ilock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1101 error = filemap_write_and_wait(inode->i_mapping);
1102 if (error)
1103 goto out_unlock;
1104 error = invalidate_inode_pages2(inode->i_mapping);
1105 if (error)
1106 goto out_unlock;
1107
1108 *join_flags = XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL;
1109 return 0;
1110
1111out_unlock:
1112 xfs_iunlock(ip, XFS_MMAPLOCK_EXCL | XFS_IOLOCK_EXCL);
1113 return error;
1114
1115}
1116
1117/*
1118 * Set up the transaction structure for the setattr operation, checking that we
1119 * have permission to do so. On success, return a clean transaction and the
1120 * inode locked exclusively ready for further operation specific checks. On
1121 * failure, return an error without modifying or locking the inode.
1122 *
1123 * The inode might already be IO locked on call. If this is the case, it is
1124 * indicated in @join_flags and we take full responsibility for ensuring they
1125 * are unlocked from now on. Hence if we have an error here, we still have to
1126 * unlock them. Otherwise, once they are joined to the transaction, they will
1127 * be unlocked on commit/cancel.
1128 */
1129static struct xfs_trans *
1130xfs_ioctl_setattr_get_trans(
1131 struct xfs_inode *ip,
1132 int join_flags)
1133{
1134 struct xfs_mount *mp = ip->i_mount;
1135 struct xfs_trans *tp;
1136 int error = -EROFS;
1137
1138 if (mp->m_flags & XFS_MOUNT_RDONLY)
1139 goto out_unlock;
1140 error = -EIO;
1141 if (XFS_FORCED_SHUTDOWN(mp))
1142 goto out_unlock;
1143
1144 tp = xfs_trans_alloc(mp, XFS_TRANS_SETATTR_NOT_SIZE);
1145 error = xfs_trans_reserve(tp, &M_RES(mp)->tr_ichange, 0, 0);
1146 if (error)
1147 goto out_cancel;
1148
1149 xfs_ilock(ip, XFS_ILOCK_EXCL);
1150 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | join_flags);
1151 join_flags = 0;
1152
1153 /*
1154 * CAP_FOWNER overrides the following restrictions:
1155 *
1156 * The user ID of the calling process must be equal to the file owner
1157 * ID, except in cases where the CAP_FSETID capability is applicable.
1158 */
1159 if (!inode_owner_or_capable(VFS_I(ip))) {
1160 error = -EPERM;
1161 goto out_cancel;
1162 }
1163
1164 if (mp->m_flags & XFS_MOUNT_WSYNC)
1165 xfs_trans_set_sync(tp);
1166
1167 return tp;
1168
1169out_cancel:
1170 xfs_trans_cancel(tp);
1171out_unlock:
1172 if (join_flags)
1173 xfs_iunlock(ip, join_flags);
1174 return ERR_PTR(error);
1175}
1176
1177/*
1178 * extent size hint validation is somewhat cumbersome. Rules are:
1179 *
1180 * 1. extent size hint is only valid for directories and regular files
1181 * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1182 * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1183 * 4. can only be changed on regular files if no extents are allocated
1184 * 5. can be changed on directories at any time
1185 * 6. extsize hint of 0 turns off hints, clears inode flags.
1186 * 7. Extent size must be a multiple of the appropriate block size.
1187 * 8. for non-realtime files, the extent size hint must be limited
1188 * to half the AG size to avoid alignment extending the extent beyond the
1189 * limits of the AG.
1190 */
1191static int
1192xfs_ioctl_setattr_check_extsize(
1193 struct xfs_inode *ip,
1194 struct fsxattr *fa)
1195{
1196 struct xfs_mount *mp = ip->i_mount;
1197
1198 if ((fa->fsx_xflags & FS_XFLAG_EXTSIZE) && !S_ISREG(VFS_I(ip)->i_mode))
1199 return -EINVAL;
1200
1201 if ((fa->fsx_xflags & FS_XFLAG_EXTSZINHERIT) &&
1202 !S_ISDIR(VFS_I(ip)->i_mode))
1203 return -EINVAL;
1204
1205 if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_d.di_nextents &&
1206 ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1207 return -EINVAL;
1208
1209 if (fa->fsx_extsize != 0) {
1210 xfs_extlen_t size;
1211 xfs_fsblock_t extsize_fsb;
1212
1213 extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1214 if (extsize_fsb > MAXEXTLEN)
1215 return -EINVAL;
1216
1217 if (XFS_IS_REALTIME_INODE(ip) ||
1218 (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1219 size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1220 } else {
1221 size = mp->m_sb.sb_blocksize;
1222 if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1223 return -EINVAL;
1224 }
1225
1226 if (fa->fsx_extsize % size)
1227 return -EINVAL;
1228 } else
1229 fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE | FS_XFLAG_EXTSZINHERIT);
1230
1231 return 0;
1232}
1233
1234static int
1235xfs_ioctl_setattr_check_projid(
1236 struct xfs_inode *ip,
1237 struct fsxattr *fa)
1238{
1239 /* Disallow 32bit project ids if projid32bit feature is not enabled. */
1240 if (fa->fsx_projid > (__uint16_t)-1 &&
1241 !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1242 return -EINVAL;
1243
1244 /*
1245 * Project Quota ID state is only allowed to change from within the init
1246 * namespace. Enforce that restriction only if we are trying to change
1247 * the quota ID state. Everything else is allowed in user namespaces.
1248 */
1249 if (current_user_ns() == &init_user_ns)
1250 return 0;
1251
1252 if (xfs_get_projid(ip) != fa->fsx_projid)
1253 return -EINVAL;
1254 if ((fa->fsx_xflags & FS_XFLAG_PROJINHERIT) !=
1255 (ip->i_d.di_flags & XFS_DIFLAG_PROJINHERIT))
1256 return -EINVAL;
1257
1258 return 0;
1259}
1260
1261STATIC int
1262xfs_ioctl_setattr(
1263 xfs_inode_t *ip,
1264 struct fsxattr *fa)
1265{
1266 struct xfs_mount *mp = ip->i_mount;
1267 struct xfs_trans *tp;
1268 struct xfs_dquot *udqp = NULL;
1269 struct xfs_dquot *pdqp = NULL;
1270 struct xfs_dquot *olddquot = NULL;
1271 int code;
1272 int join_flags = 0;
1273
1274 trace_xfs_ioctl_setattr(ip);
1275
1276 code = xfs_ioctl_setattr_check_projid(ip, fa);
1277 if (code)
1278 return code;
1279
1280 /*
1281 * If disk quotas is on, we make sure that the dquots do exist on disk,
1282 * before we start any other transactions. Trying to do this later
1283 * is messy. We don't care to take a readlock to look at the ids
1284 * in inode here, because we can't hold it across the trans_reserve.
1285 * If the IDs do change before we take the ilock, we're covered
1286 * because the i_*dquot fields will get updated anyway.
1287 */
1288 if (XFS_IS_QUOTA_ON(mp)) {
1289 code = xfs_qm_vop_dqalloc(ip, ip->i_d.di_uid,
1290 ip->i_d.di_gid, fa->fsx_projid,
1291 XFS_QMOPT_PQUOTA, &udqp, NULL, &pdqp);
1292 if (code)
1293 return code;
1294 }
1295
1296 /*
1297 * Changing DAX config may require inode locking for mapping
1298 * invalidation. These need to be held all the way to transaction commit
1299 * or cancel time, so need to be passed through to
1300 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1301 * appropriately.
1302 */
1303 code = xfs_ioctl_setattr_dax_invalidate(ip, fa, &join_flags);
1304 if (code)
1305 goto error_free_dquots;
1306
1307 tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1308 if (IS_ERR(tp)) {
1309 code = PTR_ERR(tp);
1310 goto error_free_dquots;
1311 }
1312
1313
1314 if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1315 xfs_get_projid(ip) != fa->fsx_projid) {
1316 code = xfs_qm_vop_chown_reserve(tp, ip, udqp, NULL, pdqp,
1317 capable(CAP_FOWNER) ? XFS_QMOPT_FORCE_RES : 0);
1318 if (code) /* out of quota */
1319 goto error_trans_cancel;
1320 }
1321
1322 code = xfs_ioctl_setattr_check_extsize(ip, fa);
1323 if (code)
1324 goto error_trans_cancel;
1325
1326 code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1327 if (code)
1328 goto error_trans_cancel;
1329
1330 /*
1331 * Change file ownership. Must be the owner or privileged. CAP_FSETID
1332 * overrides the following restrictions:
1333 *
1334 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1335 * successful return from chown()
1336 */
1337
1338 if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1339 !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1340 VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1341
1342 /* Change the ownerships and register project quota modifications */
1343 if (xfs_get_projid(ip) != fa->fsx_projid) {
1344 if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1345 olddquot = xfs_qm_vop_chown(tp, ip,
1346 &ip->i_pdquot, pdqp);
1347 }
1348 ASSERT(ip->i_d.di_version > 1);
1349 xfs_set_projid(ip, fa->fsx_projid);
1350 }
1351
1352 /*
1353 * Only set the extent size hint if we've already determined that the
1354 * extent size hint should be set on the inode. If no extent size flags
1355 * are set on the inode then unconditionally clear the extent size hint.
1356 */
1357 if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1358 ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1359 else
1360 ip->i_d.di_extsize = 0;
1361
1362 code = xfs_trans_commit(tp);
1363
1364 /*
1365 * Release any dquot(s) the inode had kept before chown.
1366 */
1367 xfs_qm_dqrele(olddquot);
1368 xfs_qm_dqrele(udqp);
1369 xfs_qm_dqrele(pdqp);
1370
1371 return code;
1372
1373error_trans_cancel:
1374 xfs_trans_cancel(tp);
1375error_free_dquots:
1376 xfs_qm_dqrele(udqp);
1377 xfs_qm_dqrele(pdqp);
1378 return code;
1379}
1380
1381STATIC int
1382xfs_ioc_fssetxattr(
1383 xfs_inode_t *ip,
1384 struct file *filp,
1385 void __user *arg)
1386{
1387 struct fsxattr fa;
1388 int error;
1389
1390 if (copy_from_user(&fa, arg, sizeof(fa)))
1391 return -EFAULT;
1392
1393 error = mnt_want_write_file(filp);
1394 if (error)
1395 return error;
1396 error = xfs_ioctl_setattr(ip, &fa);
1397 mnt_drop_write_file(filp);
1398 return error;
1399}
1400
1401STATIC int
1402xfs_ioc_getxflags(
1403 xfs_inode_t *ip,
1404 void __user *arg)
1405{
1406 unsigned int flags;
1407
1408 flags = xfs_di2lxflags(ip->i_d.di_flags);
1409 if (copy_to_user(arg, &flags, sizeof(flags)))
1410 return -EFAULT;
1411 return 0;
1412}
1413
1414STATIC int
1415xfs_ioc_setxflags(
1416 struct xfs_inode *ip,
1417 struct file *filp,
1418 void __user *arg)
1419{
1420 struct xfs_trans *tp;
1421 struct fsxattr fa;
1422 unsigned int flags;
1423 int join_flags = 0;
1424 int error;
1425
1426 if (copy_from_user(&flags, arg, sizeof(flags)))
1427 return -EFAULT;
1428
1429 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1430 FS_NOATIME_FL | FS_NODUMP_FL | \
1431 FS_SYNC_FL))
1432 return -EOPNOTSUPP;
1433
1434 fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1435
1436 error = mnt_want_write_file(filp);
1437 if (error)
1438 return error;
1439
1440 /*
1441 * Changing DAX config may require inode locking for mapping
1442 * invalidation. These need to be held all the way to transaction commit
1443 * or cancel time, so need to be passed through to
1444 * xfs_ioctl_setattr_get_trans() so it can apply them to the join call
1445 * appropriately.
1446 */
1447 error = xfs_ioctl_setattr_dax_invalidate(ip, &fa, &join_flags);
1448 if (error)
1449 goto out_drop_write;
1450
1451 tp = xfs_ioctl_setattr_get_trans(ip, join_flags);
1452 if (IS_ERR(tp)) {
1453 error = PTR_ERR(tp);
1454 goto out_drop_write;
1455 }
1456
1457 error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1458 if (error) {
1459 xfs_trans_cancel(tp);
1460 goto out_drop_write;
1461 }
1462
1463 error = xfs_trans_commit(tp);
1464out_drop_write:
1465 mnt_drop_write_file(filp);
1466 return error;
1467}
1468
1469STATIC int
1470xfs_getbmap_format(void **ap, struct getbmapx *bmv, int *full)
1471{
1472 struct getbmap __user *base = (struct getbmap __user *)*ap;
1473
1474 /* copy only getbmap portion (not getbmapx) */
1475 if (copy_to_user(base, bmv, sizeof(struct getbmap)))
1476 return -EFAULT;
1477
1478 *ap += sizeof(struct getbmap);
1479 return 0;
1480}
1481
1482STATIC int
1483xfs_ioc_getbmap(
1484 struct xfs_inode *ip,
1485 int ioflags,
1486 unsigned int cmd,
1487 void __user *arg)
1488{
1489 struct getbmapx bmx;
1490 int error;
1491
1492 if (copy_from_user(&bmx, arg, sizeof(struct getbmapx)))
1493 return -EFAULT;
1494
1495 if (bmx.bmv_count < 2)
1496 return -EINVAL;
1497
1498 bmx.bmv_iflags = (cmd == XFS_IOC_GETBMAPA ? BMV_IF_ATTRFORK : 0);
1499 if (ioflags & XFS_IO_INVIS)
1500 bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1501
1502 error = xfs_getbmap(ip, &bmx, xfs_getbmap_format,
1503 (__force struct getbmap *)arg+1);
1504 if (error)
1505 return error;
1506
1507 /* copy back header - only size of getbmap */
1508 if (copy_to_user(arg, &bmx, sizeof(struct getbmap)))
1509 return -EFAULT;
1510 return 0;
1511}
1512
1513STATIC int
1514xfs_getbmapx_format(void **ap, struct getbmapx *bmv, int *full)
1515{
1516 struct getbmapx __user *base = (struct getbmapx __user *)*ap;
1517
1518 if (copy_to_user(base, bmv, sizeof(struct getbmapx)))
1519 return -EFAULT;
1520
1521 *ap += sizeof(struct getbmapx);
1522 return 0;
1523}
1524
1525STATIC int
1526xfs_ioc_getbmapx(
1527 struct xfs_inode *ip,
1528 void __user *arg)
1529{
1530 struct getbmapx bmx;
1531 int error;
1532
1533 if (copy_from_user(&bmx, arg, sizeof(bmx)))
1534 return -EFAULT;
1535
1536 if (bmx.bmv_count < 2)
1537 return -EINVAL;
1538
1539 if (bmx.bmv_iflags & (~BMV_IF_VALID))
1540 return -EINVAL;
1541
1542 error = xfs_getbmap(ip, &bmx, xfs_getbmapx_format,
1543 (__force struct getbmapx *)arg+1);
1544 if (error)
1545 return error;
1546
1547 /* copy back header */
1548 if (copy_to_user(arg, &bmx, sizeof(struct getbmapx)))
1549 return -EFAULT;
1550
1551 return 0;
1552}
1553
1554int
1555xfs_ioc_swapext(
1556 xfs_swapext_t *sxp)
1557{
1558 xfs_inode_t *ip, *tip;
1559 struct fd f, tmp;
1560 int error = 0;
1561
1562 /* Pull information for the target fd */
1563 f = fdget((int)sxp->sx_fdtarget);
1564 if (!f.file) {
1565 error = -EINVAL;
1566 goto out;
1567 }
1568
1569 if (!(f.file->f_mode & FMODE_WRITE) ||
1570 !(f.file->f_mode & FMODE_READ) ||
1571 (f.file->f_flags & O_APPEND)) {
1572 error = -EBADF;
1573 goto out_put_file;
1574 }
1575
1576 tmp = fdget((int)sxp->sx_fdtmp);
1577 if (!tmp.file) {
1578 error = -EINVAL;
1579 goto out_put_file;
1580 }
1581
1582 if (!(tmp.file->f_mode & FMODE_WRITE) ||
1583 !(tmp.file->f_mode & FMODE_READ) ||
1584 (tmp.file->f_flags & O_APPEND)) {
1585 error = -EBADF;
1586 goto out_put_tmp_file;
1587 }
1588
1589 if (IS_SWAPFILE(file_inode(f.file)) ||
1590 IS_SWAPFILE(file_inode(tmp.file))) {
1591 error = -EINVAL;
1592 goto out_put_tmp_file;
1593 }
1594
1595 ip = XFS_I(file_inode(f.file));
1596 tip = XFS_I(file_inode(tmp.file));
1597
1598 if (ip->i_mount != tip->i_mount) {
1599 error = -EINVAL;
1600 goto out_put_tmp_file;
1601 }
1602
1603 if (ip->i_ino == tip->i_ino) {
1604 error = -EINVAL;
1605 goto out_put_tmp_file;
1606 }
1607
1608 if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1609 error = -EIO;
1610 goto out_put_tmp_file;
1611 }
1612
1613 error = xfs_swap_extents(ip, tip, sxp);
1614
1615 out_put_tmp_file:
1616 fdput(tmp);
1617 out_put_file:
1618 fdput(f);
1619 out:
1620 return error;
1621}
1622
1623/*
1624 * Note: some of the ioctl's return positive numbers as a
1625 * byte count indicating success, such as readlink_by_handle.
1626 * So we don't "sign flip" like most other routines. This means
1627 * true errors need to be returned as a negative value.
1628 */
1629long
1630xfs_file_ioctl(
1631 struct file *filp,
1632 unsigned int cmd,
1633 unsigned long p)
1634{
1635 struct inode *inode = file_inode(filp);
1636 struct xfs_inode *ip = XFS_I(inode);
1637 struct xfs_mount *mp = ip->i_mount;
1638 void __user *arg = (void __user *)p;
1639 int ioflags = 0;
1640 int error;
1641
1642 if (filp->f_mode & FMODE_NOCMTIME)
1643 ioflags |= XFS_IO_INVIS;
1644
1645 trace_xfs_file_ioctl(ip);
1646
1647 switch (cmd) {
1648 case FITRIM:
1649 return xfs_ioc_trim(mp, arg);
1650 case XFS_IOC_ALLOCSP:
1651 case XFS_IOC_FREESP:
1652 case XFS_IOC_RESVSP:
1653 case XFS_IOC_UNRESVSP:
1654 case XFS_IOC_ALLOCSP64:
1655 case XFS_IOC_FREESP64:
1656 case XFS_IOC_RESVSP64:
1657 case XFS_IOC_UNRESVSP64:
1658 case XFS_IOC_ZERO_RANGE: {
1659 xfs_flock64_t bf;
1660
1661 if (copy_from_user(&bf, arg, sizeof(bf)))
1662 return -EFAULT;
1663 return xfs_ioc_space(ip, inode, filp, ioflags, cmd, &bf);
1664 }
1665 case XFS_IOC_DIOINFO: {
1666 struct dioattr da;
1667 xfs_buftarg_t *target =
1668 XFS_IS_REALTIME_INODE(ip) ?
1669 mp->m_rtdev_targp : mp->m_ddev_targp;
1670
1671 da.d_mem = da.d_miniosz = target->bt_logical_sectorsize;
1672 da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1673
1674 if (copy_to_user(arg, &da, sizeof(da)))
1675 return -EFAULT;
1676 return 0;
1677 }
1678
1679 case XFS_IOC_FSBULKSTAT_SINGLE:
1680 case XFS_IOC_FSBULKSTAT:
1681 case XFS_IOC_FSINUMBERS:
1682 return xfs_ioc_bulkstat(mp, cmd, arg);
1683
1684 case XFS_IOC_FSGEOMETRY_V1:
1685 return xfs_ioc_fsgeometry_v1(mp, arg);
1686
1687 case XFS_IOC_FSGEOMETRY:
1688 return xfs_ioc_fsgeometry(mp, arg);
1689
1690 case XFS_IOC_GETVERSION:
1691 return put_user(inode->i_generation, (int __user *)arg);
1692
1693 case XFS_IOC_FSGETXATTR:
1694 return xfs_ioc_fsgetxattr(ip, 0, arg);
1695 case XFS_IOC_FSGETXATTRA:
1696 return xfs_ioc_fsgetxattr(ip, 1, arg);
1697 case XFS_IOC_FSSETXATTR:
1698 return xfs_ioc_fssetxattr(ip, filp, arg);
1699 case XFS_IOC_GETXFLAGS:
1700 return xfs_ioc_getxflags(ip, arg);
1701 case XFS_IOC_SETXFLAGS:
1702 return xfs_ioc_setxflags(ip, filp, arg);
1703
1704 case XFS_IOC_FSSETDM: {
1705 struct fsdmidata dmi;
1706
1707 if (copy_from_user(&dmi, arg, sizeof(dmi)))
1708 return -EFAULT;
1709
1710 error = mnt_want_write_file(filp);
1711 if (error)
1712 return error;
1713
1714 error = xfs_set_dmattrs(ip, dmi.fsd_dmevmask,
1715 dmi.fsd_dmstate);
1716 mnt_drop_write_file(filp);
1717 return error;
1718 }
1719
1720 case XFS_IOC_GETBMAP:
1721 case XFS_IOC_GETBMAPA:
1722 return xfs_ioc_getbmap(ip, ioflags, cmd, arg);
1723
1724 case XFS_IOC_GETBMAPX:
1725 return xfs_ioc_getbmapx(ip, arg);
1726
1727 case XFS_IOC_FD_TO_HANDLE:
1728 case XFS_IOC_PATH_TO_HANDLE:
1729 case XFS_IOC_PATH_TO_FSHANDLE: {
1730 xfs_fsop_handlereq_t hreq;
1731
1732 if (copy_from_user(&hreq, arg, sizeof(hreq)))
1733 return -EFAULT;
1734 return xfs_find_handle(cmd, &hreq);
1735 }
1736 case XFS_IOC_OPEN_BY_HANDLE: {
1737 xfs_fsop_handlereq_t hreq;
1738
1739 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1740 return -EFAULT;
1741 return xfs_open_by_handle(filp, &hreq);
1742 }
1743 case XFS_IOC_FSSETDM_BY_HANDLE:
1744 return xfs_fssetdm_by_handle(filp, arg);
1745
1746 case XFS_IOC_READLINK_BY_HANDLE: {
1747 xfs_fsop_handlereq_t hreq;
1748
1749 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1750 return -EFAULT;
1751 return xfs_readlink_by_handle(filp, &hreq);
1752 }
1753 case XFS_IOC_ATTRLIST_BY_HANDLE:
1754 return xfs_attrlist_by_handle(filp, arg);
1755
1756 case XFS_IOC_ATTRMULTI_BY_HANDLE:
1757 return xfs_attrmulti_by_handle(filp, arg);
1758
1759 case XFS_IOC_SWAPEXT: {
1760 struct xfs_swapext sxp;
1761
1762 if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1763 return -EFAULT;
1764 error = mnt_want_write_file(filp);
1765 if (error)
1766 return error;
1767 error = xfs_ioc_swapext(&sxp);
1768 mnt_drop_write_file(filp);
1769 return error;
1770 }
1771
1772 case XFS_IOC_FSCOUNTS: {
1773 xfs_fsop_counts_t out;
1774
1775 error = xfs_fs_counts(mp, &out);
1776 if (error)
1777 return error;
1778
1779 if (copy_to_user(arg, &out, sizeof(out)))
1780 return -EFAULT;
1781 return 0;
1782 }
1783
1784 case XFS_IOC_SET_RESBLKS: {
1785 xfs_fsop_resblks_t inout;
1786 __uint64_t in;
1787
1788 if (!capable(CAP_SYS_ADMIN))
1789 return -EPERM;
1790
1791 if (mp->m_flags & XFS_MOUNT_RDONLY)
1792 return -EROFS;
1793
1794 if (copy_from_user(&inout, arg, sizeof(inout)))
1795 return -EFAULT;
1796
1797 error = mnt_want_write_file(filp);
1798 if (error)
1799 return error;
1800
1801 /* input parameter is passed in resblks field of structure */
1802 in = inout.resblks;
1803 error = xfs_reserve_blocks(mp, &in, &inout);
1804 mnt_drop_write_file(filp);
1805 if (error)
1806 return error;
1807
1808 if (copy_to_user(arg, &inout, sizeof(inout)))
1809 return -EFAULT;
1810 return 0;
1811 }
1812
1813 case XFS_IOC_GET_RESBLKS: {
1814 xfs_fsop_resblks_t out;
1815
1816 if (!capable(CAP_SYS_ADMIN))
1817 return -EPERM;
1818
1819 error = xfs_reserve_blocks(mp, NULL, &out);
1820 if (error)
1821 return error;
1822
1823 if (copy_to_user(arg, &out, sizeof(out)))
1824 return -EFAULT;
1825
1826 return 0;
1827 }
1828
1829 case XFS_IOC_FSGROWFSDATA: {
1830 xfs_growfs_data_t in;
1831
1832 if (copy_from_user(&in, arg, sizeof(in)))
1833 return -EFAULT;
1834
1835 error = mnt_want_write_file(filp);
1836 if (error)
1837 return error;
1838 error = xfs_growfs_data(mp, &in);
1839 mnt_drop_write_file(filp);
1840 return error;
1841 }
1842
1843 case XFS_IOC_FSGROWFSLOG: {
1844 xfs_growfs_log_t in;
1845
1846 if (copy_from_user(&in, arg, sizeof(in)))
1847 return -EFAULT;
1848
1849 error = mnt_want_write_file(filp);
1850 if (error)
1851 return error;
1852 error = xfs_growfs_log(mp, &in);
1853 mnt_drop_write_file(filp);
1854 return error;
1855 }
1856
1857 case XFS_IOC_FSGROWFSRT: {
1858 xfs_growfs_rt_t in;
1859
1860 if (copy_from_user(&in, arg, sizeof(in)))
1861 return -EFAULT;
1862
1863 error = mnt_want_write_file(filp);
1864 if (error)
1865 return error;
1866 error = xfs_growfs_rt(mp, &in);
1867 mnt_drop_write_file(filp);
1868 return error;
1869 }
1870
1871 case XFS_IOC_GOINGDOWN: {
1872 __uint32_t in;
1873
1874 if (!capable(CAP_SYS_ADMIN))
1875 return -EPERM;
1876
1877 if (get_user(in, (__uint32_t __user *)arg))
1878 return -EFAULT;
1879
1880 return xfs_fs_goingdown(mp, in);
1881 }
1882
1883 case XFS_IOC_ERROR_INJECTION: {
1884 xfs_error_injection_t in;
1885
1886 if (!capable(CAP_SYS_ADMIN))
1887 return -EPERM;
1888
1889 if (copy_from_user(&in, arg, sizeof(in)))
1890 return -EFAULT;
1891
1892 return xfs_errortag_add(in.errtag, mp);
1893 }
1894
1895 case XFS_IOC_ERROR_CLEARALL:
1896 if (!capable(CAP_SYS_ADMIN))
1897 return -EPERM;
1898
1899 return xfs_errortag_clearall(mp, 1);
1900
1901 case XFS_IOC_FREE_EOFBLOCKS: {
1902 struct xfs_fs_eofblocks eofb;
1903 struct xfs_eofblocks keofb;
1904
1905 if (!capable(CAP_SYS_ADMIN))
1906 return -EPERM;
1907
1908 if (mp->m_flags & XFS_MOUNT_RDONLY)
1909 return -EROFS;
1910
1911 if (copy_from_user(&eofb, arg, sizeof(eofb)))
1912 return -EFAULT;
1913
1914 error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
1915 if (error)
1916 return error;
1917
1918 return xfs_icache_free_eofblocks(mp, &keofb);
1919 }
1920
1921 default:
1922 return -ENOTTY;
1923 }
1924}