1 // SPDX-License-Identifier: GPL-2.0
3 * Copyright (c) 2000-2005 Silicon Graphics, Inc.
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"
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"
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"
41 #include "xfs_exchrange.h"
42 #include "xfs_handle.h"
43 #include "xfs_rtgroup.h"
45 #include <linux/mount.h>
46 #include <linux/fileattr.h>
48 /* Return 0 on success or positive error */
50 xfs_fsbulkstat_one_fmt(
51 struct xfs_ibulk *breq,
52 const struct xfs_bulkstat *bstat)
56 xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
57 if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
59 return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
64 struct xfs_ibulk *breq,
65 const struct xfs_inumbers *igrp)
67 struct xfs_inogrp ig1;
69 xfs_inumbers_to_inogrp(&ig1, igrp);
70 if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
72 return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
81 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
82 struct xfs_fsop_bulkreq bulkreq;
83 struct xfs_ibulk breq = {
85 .idmap = file_mnt_idmap(file),
91 /* done = 1 if there are more stats to get and if bulkstat */
92 /* should be called again (unused here, but used in dmapi) */
94 if (!capable(CAP_SYS_ADMIN))
97 if (xfs_is_shutdown(mp))
100 if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
103 if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
106 if (bulkreq.icount <= 0)
109 if (bulkreq.ubuffer == NULL)
112 breq.ubuffer = bulkreq.ubuffer;
113 breq.icount = bulkreq.icount;
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
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;
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;
143 if (bulkreq.lastip != NULL &&
144 copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
147 if (bulkreq.ocount != NULL &&
148 copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
154 /* Return 0 on success or positive error */
157 struct xfs_ibulk *breq,
158 const struct xfs_bulkstat *bstat)
160 if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
162 return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
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.
173 struct xfs_mount *mp,
174 const struct xfs_bulk_ireq *hdr,
175 struct xfs_ibulk *breq,
176 void __user *ubuffer)
178 if (hdr->icount == 0 ||
179 (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
180 memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
183 breq->startino = hdr->ino;
184 breq->ubuffer = ubuffer;
185 breq->icount = hdr->icount;
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
194 if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
195 if (hdr->flags & XFS_BULK_IREQ_AGNO)
199 case XFS_BULK_IREQ_SPECIAL_ROOT:
200 breq->startino = mp->m_sb.sb_rootino;
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.
213 if (hdr->flags & XFS_BULK_IREQ_AGNO) {
214 if (hdr->agno >= mp->m_sb.sb_agcount)
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)
222 breq->flags |= XFS_IBULK_SAME_AG;
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)
227 } else if (hdr->agno)
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)
234 if (hdr->flags & XFS_BULK_IREQ_NREXT64)
235 breq->flags |= XFS_IBULK_NREXT64;
237 /* Caller wants to see metadata directories in bulkstat output. */
238 if (hdr->flags & XFS_BULK_IREQ_METADIR)
239 breq->flags |= XFS_IBULK_METADIR;
245 * Update the userspace bulk request @hdr to reflect the end state of the
246 * internal bulk request @breq.
249 xfs_bulk_ireq_teardown(
250 struct xfs_bulk_ireq *hdr,
251 struct xfs_ibulk *breq)
253 hdr->ino = breq->startino;
254 hdr->ocount = breq->ocount;
257 /* Handle the v5 bulkstat ioctl. */
262 struct xfs_bulkstat_req __user *arg)
264 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
265 struct xfs_bulk_ireq hdr;
266 struct xfs_ibulk breq = {
268 .idmap = file_mnt_idmap(file),
272 if (!capable(CAP_SYS_ADMIN))
275 if (xfs_is_shutdown(mp))
278 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
281 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
282 if (error == -ECANCELED)
287 error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
292 xfs_bulk_ireq_teardown(&hdr, &breq);
293 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
301 struct xfs_ibulk *breq,
302 const struct xfs_inumbers *igrp)
304 if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
306 return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
309 /* Handle the v5 inumbers ioctl. */
312 struct xfs_mount *mp,
314 struct xfs_inumbers_req __user *arg)
316 struct xfs_bulk_ireq hdr;
317 struct xfs_ibulk breq = {
322 if (!capable(CAP_SYS_ADMIN))
325 if (xfs_is_shutdown(mp))
328 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
331 if (hdr.flags & XFS_BULK_IREQ_METADIR)
334 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
335 if (error == -ECANCELED)
340 error = xfs_inumbers(&breq, xfs_inumbers_fmt);
345 xfs_bulk_ireq_teardown(&hdr, &breq);
346 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
354 struct xfs_mount *mp,
358 struct xfs_fsop_geom fsgeo;
361 xfs_fs_geometry(mp, &fsgeo, struct_version);
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);
368 xfs_fsop_geom_health(mp, &fsgeo);
372 if (copy_to_user(arg, &fsgeo, len))
379 struct xfs_mount *mp,
382 struct xfs_perag *pag;
383 struct xfs_ag_geometry ageo;
386 if (copy_from_user(&ageo, arg, sizeof(ageo)))
390 if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
393 pag = xfs_perag_get(mp, ageo.ag_number);
397 error = xfs_ag_get_geometry(pag, &ageo);
402 if (copy_to_user(arg, &ageo, sizeof(ageo)))
408 xfs_ioc_rtgroup_geometry(
409 struct xfs_mount *mp,
412 struct xfs_rtgroup *rtg;
413 struct xfs_rtgroup_geometry rgeo;
416 if (copy_from_user(&rgeo, arg, sizeof(rgeo)))
420 if (memchr_inv(&rgeo.rg_reserved, 0, sizeof(rgeo.rg_reserved)))
422 if (!xfs_has_rtgroups(mp))
425 rtg = xfs_rtgroup_get(mp, rgeo.rg_number);
429 error = xfs_rtgroup_get_geometry(rtg, &rgeo);
430 xfs_rtgroup_put(rtg);
434 if (copy_to_user(arg, &rgeo, sizeof(rgeo)))
440 * Linux extended inode flags interface.
445 struct xfs_inode *ip,
449 struct xfs_mount *mp = ip->i_mount;
450 struct xfs_ifork *ifp = xfs_ifork_ptr(ip, whichfork);
452 fileattr_fill_xflags(fa, xfs_ip2xflags(ip));
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) {
458 * Don't let a misaligned extent size hint on a directory
459 * escape to userspace if it won't pass the setattr checks
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);
468 fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
472 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE) {
474 * Don't let a misaligned CoW extent size hint on a directory
475 * escape to userspace if it won't pass the setattr checks
478 if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
479 ip->i_cowextsize % mp->m_sb.sb_rextsize > 0) {
480 fa->fsx_xflags &= ~FS_XFLAG_COWEXTSIZE;
481 fa->fsx_cowextsize = 0;
483 fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
487 fa->fsx_projid = ip->i_projid;
488 if (ifp && !xfs_need_iread_extents(ifp))
489 fa->fsx_nextents = xfs_iext_count(ifp);
491 fa->fsx_nextents = xfs_ifork_nextents(ifp);
501 xfs_ilock(ip, XFS_ILOCK_SHARED);
502 xfs_fill_fsxattr(ip, XFS_ATTR_FORK, &fa);
503 xfs_iunlock(ip, XFS_ILOCK_SHARED);
505 return copy_fsxattr_to_user(&fa, arg);
510 struct dentry *dentry,
513 struct xfs_inode *ip = XFS_I(d_inode(dentry));
515 if (d_is_special(dentry))
518 xfs_ilock(ip, XFS_ILOCK_SHARED);
519 xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
520 xfs_iunlock(ip, XFS_ILOCK_SHARED);
526 xfs_ioctl_setattr_xflags(
527 struct xfs_trans *tp,
528 struct xfs_inode *ip,
531 struct xfs_mount *mp = ip->i_mount;
532 bool rtflag = (fa->fsx_xflags & FS_XFLAG_REALTIME);
535 if (rtflag != XFS_IS_REALTIME_INODE(ip)) {
536 /* Can't change realtime flag if any extents are allocated. */
537 if (xfs_inode_has_filedata(ip))
541 * If S_DAX is enabled on this file, we can only switch the
542 * device if both support fsdax. We can't update S_DAX because
543 * there might be other threads walking down the access paths.
545 if (IS_DAX(VFS_I(ip)) &&
546 (mp->m_ddev_targp->bt_daxdev == NULL ||
547 (mp->m_rtdev_targp &&
548 mp->m_rtdev_targp->bt_daxdev == NULL)))
553 /* If realtime flag is set then must have realtime device */
554 if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
555 xfs_extlen_to_rtxmod(mp, ip->i_extsize))
559 /* diflags2 only valid for v3 inodes. */
560 i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
561 if (i_flags2 && !xfs_has_v3inodes(mp))
564 ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
565 ip->i_diflags2 = i_flags2;
567 xfs_diflags_to_iflags(ip, false);
570 * Make the stable writes flag match that of the device the inode
571 * resides on when flipping the RT flag.
573 if (rtflag != XFS_IS_REALTIME_INODE(ip) && S_ISREG(VFS_I(ip)->i_mode))
574 xfs_update_stable_writes(ip);
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);
583 xfs_ioctl_setattr_prepare_dax(
584 struct xfs_inode *ip,
587 struct xfs_mount *mp = ip->i_mount;
588 struct inode *inode = VFS_I(ip);
590 if (S_ISDIR(inode->i_mode))
593 if (xfs_has_dax_always(mp) || xfs_has_dax_never(mp))
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);
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.
609 static struct xfs_trans *
610 xfs_ioctl_setattr_get_trans(
611 struct xfs_inode *ip,
612 struct xfs_dquot *pdqp)
614 struct xfs_mount *mp = ip->i_mount;
615 struct xfs_trans *tp;
618 if (xfs_is_readonly(mp))
621 if (xfs_is_shutdown(mp))
624 error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
625 has_capability_noaudit(current, CAP_FOWNER), &tp);
629 if (xfs_has_wsync(mp))
630 xfs_trans_set_sync(tp);
635 return ERR_PTR(error);
639 * Validate a proposed extent size hint. For regular files, the hint can only
640 * be changed if no extents are allocated.
643 xfs_ioctl_setattr_check_extsize(
644 struct xfs_inode *ip,
647 struct xfs_mount *mp = ip->i_mount;
648 xfs_failaddr_t failaddr;
649 uint16_t new_diflags;
654 if (S_ISREG(VFS_I(ip)->i_mode) && xfs_inode_has_filedata(ip) &&
655 XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
658 if (fa->fsx_extsize & mp->m_blockmask)
661 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
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
669 if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
670 (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
671 unsigned int rtextsize_bytes;
673 rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
674 if (fa->fsx_extsize % rtextsize_bytes)
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;
685 xfs_ioctl_setattr_check_cowextsize(
686 struct xfs_inode *ip,
689 struct xfs_mount *mp = ip->i_mount;
690 xfs_failaddr_t failaddr;
691 uint64_t new_diflags2;
692 uint16_t new_diflags;
697 if (fa->fsx_cowextsize & mp->m_blockmask)
700 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
701 new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
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;
710 xfs_ioctl_setattr_check_projid(
711 struct xfs_inode *ip,
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))
726 struct mnt_idmap *idmap,
727 struct dentry *dentry,
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;
737 trace_xfs_ioctl_setattr(ip);
739 if (d_is_special(dentry))
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))
749 error = xfs_ioctl_setattr_check_projid(ip, fa);
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.
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);
769 xfs_ioctl_setattr_prepare_dax(ip, fa);
771 tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
774 goto error_free_dquots;
777 error = xfs_ioctl_setattr_check_extsize(ip, fa);
779 goto error_trans_cancel;
781 error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
783 goto error_trans_cancel;
785 error = xfs_ioctl_setattr_xflags(tp, ip, fa);
787 goto error_trans_cancel;
792 * Change file ownership. Must be the owner or privileged. CAP_FSETID
793 * overrides the following restrictions:
795 * The set-user-ID and set-group-ID bits of a file will be cleared upon
796 * successful return from chown()
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);
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);
809 ip->i_projid = fa->fsx_projid;
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.
817 if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
818 ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
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);
826 ip->i_cowextsize = 0;
830 error = xfs_trans_commit(tp);
833 * Release any dquot(s) the inode had kept before chown.
835 xfs_qm_dqrele(olddquot);
841 xfs_trans_cancel(tp);
850 struct getbmapx __user *u,
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))
859 if (recsize < sizeof(struct getbmapx))
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))
875 struct getbmapx bmx = { 0 };
876 struct kgetbmap *buf;
881 case XFS_IOC_GETBMAPA:
882 bmx.bmv_iflags = BMV_IF_ATTRFORK;
884 case XFS_IOC_GETBMAP:
885 /* struct getbmap is a strict subset of struct getbmapx. */
886 recsize = sizeof(struct getbmap);
888 case XFS_IOC_GETBMAPX:
889 recsize = sizeof(struct getbmapx);
895 if (copy_from_user(&bmx, arg, recsize))
898 if (bmx.bmv_count < 2)
900 if (bmx.bmv_count >= INT_MAX / recsize)
903 buf = kvcalloc(bmx.bmv_count, sizeof(*buf), GFP_KERNEL);
907 error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
912 if (copy_to_user(arg, &bmx, recsize))
916 for (i = 0; i < bmx.bmv_entries; i++) {
917 if (!xfs_getbmap_format(buf + i, arg, recsize))
932 xfs_inode_t *ip, *tip;
934 /* Pull information for the target fd */
935 CLASS(fd, f)((int)sxp->sx_fdtarget);
939 if (!(fd_file(f)->f_mode & FMODE_WRITE) ||
940 !(fd_file(f)->f_mode & FMODE_READ) ||
941 (fd_file(f)->f_flags & O_APPEND))
944 CLASS(fd, tmp)((int)sxp->sx_fdtmp);
948 if (!(fd_file(tmp)->f_mode & FMODE_WRITE) ||
949 !(fd_file(tmp)->f_mode & FMODE_READ) ||
950 (fd_file(tmp)->f_flags & O_APPEND))
953 if (IS_SWAPFILE(file_inode(fd_file(f))) ||
954 IS_SWAPFILE(file_inode(fd_file(tmp))))
958 * We need to ensure that the fds passed in point to XFS inodes
959 * before we cast and access them as XFS structures as we have no
960 * control over what the user passes us here.
962 if (fd_file(f)->f_op != &xfs_file_operations ||
963 fd_file(tmp)->f_op != &xfs_file_operations)
966 ip = XFS_I(file_inode(fd_file(f)));
967 tip = XFS_I(file_inode(fd_file(tmp)));
969 if (ip->i_mount != tip->i_mount)
972 if (ip->i_ino == tip->i_ino)
975 if (xfs_is_shutdown(ip->i_mount))
978 return xfs_swap_extents(ip, tip, sxp);
983 struct xfs_mount *mp,
984 char __user *user_label)
986 struct xfs_sb *sbp = &mp->m_sb;
987 char label[XFSLABEL_MAX + 1];
990 BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
992 /* 1 larger than sb_fname, so this ensures a trailing NUL char */
993 memset(label, 0, sizeof(label));
994 spin_lock(&mp->m_sb_lock);
995 strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
996 spin_unlock(&mp->m_sb_lock);
998 if (copy_to_user(user_label, label, sizeof(label)))
1006 struct xfs_mount *mp,
1007 char __user *newlabel)
1009 struct xfs_sb *sbp = &mp->m_sb;
1010 char label[XFSLABEL_MAX + 1];
1014 if (!capable(CAP_SYS_ADMIN))
1017 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1018 * smaller, at 12 bytes. We copy one more to be sure we find the
1019 * (required) NULL character to test the incoming label length.
1020 * NB: The on disk label doesn't need to be null terminated.
1022 if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1024 len = strnlen(label, XFSLABEL_MAX + 1);
1025 if (len > sizeof(sbp->sb_fname))
1028 error = mnt_want_write_file(filp);
1032 spin_lock(&mp->m_sb_lock);
1033 memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1034 memcpy(sbp->sb_fname, label, len);
1035 spin_unlock(&mp->m_sb_lock);
1038 * Now we do several things to satisfy userspace.
1039 * In addition to normal logging of the primary superblock, we also
1040 * immediately write these changes to sector zero for the primary, then
1041 * update all backup supers (as xfs_db does for a label change), then
1042 * invalidate the block device page cache. This is so that any prior
1043 * buffered reads from userspace (i.e. from blkid) are invalidated,
1044 * and userspace will see the newly-written label.
1046 error = xfs_sync_sb_buf(mp, true);
1050 * growfs also updates backup supers so lock against that.
1052 mutex_lock(&mp->m_growlock);
1053 error = xfs_update_secondary_sbs(mp);
1054 mutex_unlock(&mp->m_growlock);
1056 invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1057 if (xfs_has_rtsb(mp) && mp->m_rtdev_targp)
1058 invalidate_bdev(mp->m_rtdev_targp->bt_bdev);
1061 mnt_drop_write_file(filp);
1066 xfs_fs_eofblocks_from_user(
1067 struct xfs_fs_eofblocks *src,
1068 struct xfs_icwalk *dst)
1070 if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1073 if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1076 if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1077 memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1081 if (src->eof_flags & XFS_EOF_FLAGS_SYNC)
1082 dst->icw_flags |= XFS_ICWALK_FLAG_SYNC;
1083 if (src->eof_flags & XFS_EOF_FLAGS_UID)
1084 dst->icw_flags |= XFS_ICWALK_FLAG_UID;
1085 if (src->eof_flags & XFS_EOF_FLAGS_GID)
1086 dst->icw_flags |= XFS_ICWALK_FLAG_GID;
1087 if (src->eof_flags & XFS_EOF_FLAGS_PRID)
1088 dst->icw_flags |= XFS_ICWALK_FLAG_PRID;
1089 if (src->eof_flags & XFS_EOF_FLAGS_MINFILESIZE)
1090 dst->icw_flags |= XFS_ICWALK_FLAG_MINFILESIZE;
1092 dst->icw_prid = src->eof_prid;
1093 dst->icw_min_file_size = src->eof_min_file_size;
1095 dst->icw_uid = INVALID_UID;
1096 if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1097 dst->icw_uid = make_kuid(current_user_ns(), src->eof_uid);
1098 if (!uid_valid(dst->icw_uid))
1102 dst->icw_gid = INVALID_GID;
1103 if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1104 dst->icw_gid = make_kgid(current_user_ns(), src->eof_gid);
1105 if (!gid_valid(dst->icw_gid))
1112 xfs_ioctl_getset_resblocks(
1117 struct xfs_mount *mp = XFS_I(file_inode(filp))->i_mount;
1118 struct xfs_fsop_resblks fsop = { };
1121 if (!capable(CAP_SYS_ADMIN))
1124 if (cmd == XFS_IOC_SET_RESBLKS) {
1125 if (xfs_is_readonly(mp))
1128 if (copy_from_user(&fsop, arg, sizeof(fsop)))
1131 error = mnt_want_write_file(filp);
1134 error = xfs_reserve_blocks(mp, XC_FREE_BLOCKS, fsop.resblks);
1135 mnt_drop_write_file(filp);
1140 spin_lock(&mp->m_sb_lock);
1141 fsop.resblks = mp->m_free[XC_FREE_BLOCKS].res_total;
1142 fsop.resblks_avail = mp->m_free[XC_FREE_BLOCKS].res_avail;
1143 spin_unlock(&mp->m_sb_lock);
1145 if (copy_to_user(arg, &fsop, sizeof(fsop)))
1151 xfs_ioctl_fs_counts(
1152 struct xfs_mount *mp,
1153 struct xfs_fsop_counts __user *uarg)
1155 struct xfs_fsop_counts out = {
1156 .allocino = percpu_counter_read_positive(&mp->m_icount),
1157 .freeino = percpu_counter_read_positive(&mp->m_ifree),
1158 .freedata = xfs_estimate_freecounter(mp, XC_FREE_BLOCKS) -
1159 xfs_freecounter_unavailable(mp, XC_FREE_BLOCKS),
1160 .freertx = xfs_estimate_freecounter(mp, XC_FREE_RTEXTENTS),
1163 if (copy_to_user(uarg, &out, sizeof(out)))
1169 * These long-unused ioctls were removed from the official ioctl API in 5.17,
1170 * but retain these definitions so that we can log warnings about them.
1172 #define XFS_IOC_ALLOCSP _IOW ('X', 10, struct xfs_flock64)
1173 #define XFS_IOC_FREESP _IOW ('X', 11, struct xfs_flock64)
1174 #define XFS_IOC_ALLOCSP64 _IOW ('X', 36, struct xfs_flock64)
1175 #define XFS_IOC_FREESP64 _IOW ('X', 37, struct xfs_flock64)
1178 * Note: some of the ioctl's return positive numbers as a
1179 * byte count indicating success, such as readlink_by_handle.
1180 * So we don't "sign flip" like most other routines. This means
1181 * true errors need to be returned as a negative value.
1189 struct inode *inode = file_inode(filp);
1190 struct xfs_inode *ip = XFS_I(inode);
1191 struct xfs_mount *mp = ip->i_mount;
1192 void __user *arg = (void __user *)p;
1195 trace_xfs_file_ioctl(ip);
1199 return xfs_ioc_trim(mp, arg);
1200 case FS_IOC_GETFSLABEL:
1201 return xfs_ioc_getlabel(mp, arg);
1202 case FS_IOC_SETFSLABEL:
1203 return xfs_ioc_setlabel(filp, mp, arg);
1204 case XFS_IOC_ALLOCSP:
1205 case XFS_IOC_FREESP:
1206 case XFS_IOC_ALLOCSP64:
1207 case XFS_IOC_FREESP64:
1209 "%s should use fallocate; XFS_IOC_{ALLOC,FREE}SP ioctl unsupported",
1212 case XFS_IOC_DIOINFO: {
1213 struct xfs_buftarg *target = xfs_inode_buftarg(ip);
1216 da.d_mem = target->bt_logical_sectorsize;
1219 * See xfs_report_dioalign() for an explanation about why this
1220 * reports a value larger than the sector size for COW inodes.
1222 if (xfs_is_cow_inode(ip))
1223 da.d_miniosz = xfs_inode_alloc_unitsize(ip);
1225 da.d_miniosz = target->bt_logical_sectorsize;
1226 da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1228 if (copy_to_user(arg, &da, sizeof(da)))
1233 case XFS_IOC_FSBULKSTAT_SINGLE:
1234 case XFS_IOC_FSBULKSTAT:
1235 case XFS_IOC_FSINUMBERS:
1236 return xfs_ioc_fsbulkstat(filp, cmd, arg);
1238 case XFS_IOC_BULKSTAT:
1239 return xfs_ioc_bulkstat(filp, cmd, arg);
1240 case XFS_IOC_INUMBERS:
1241 return xfs_ioc_inumbers(mp, cmd, arg);
1243 case XFS_IOC_FSGEOMETRY_V1:
1244 return xfs_ioc_fsgeometry(mp, arg, 3);
1245 case XFS_IOC_FSGEOMETRY_V4:
1246 return xfs_ioc_fsgeometry(mp, arg, 4);
1247 case XFS_IOC_FSGEOMETRY:
1248 return xfs_ioc_fsgeometry(mp, arg, 5);
1250 case XFS_IOC_AG_GEOMETRY:
1251 return xfs_ioc_ag_geometry(mp, arg);
1252 case XFS_IOC_RTGROUP_GEOMETRY:
1253 return xfs_ioc_rtgroup_geometry(mp, arg);
1255 case XFS_IOC_GETVERSION:
1256 return put_user(inode->i_generation, (int __user *)arg);
1258 case XFS_IOC_FSGETXATTRA:
1259 return xfs_ioc_fsgetxattra(ip, arg);
1260 case XFS_IOC_GETPARENTS:
1261 return xfs_ioc_getparents(filp, arg);
1262 case XFS_IOC_GETPARENTS_BY_HANDLE:
1263 return xfs_ioc_getparents_by_handle(filp, arg);
1264 case XFS_IOC_GETBMAP:
1265 case XFS_IOC_GETBMAPA:
1266 case XFS_IOC_GETBMAPX:
1267 return xfs_ioc_getbmap(filp, cmd, arg);
1269 case FS_IOC_GETFSMAP:
1270 return xfs_ioc_getfsmap(ip, arg);
1272 case XFS_IOC_SCRUBV_METADATA:
1273 return xfs_ioc_scrubv_metadata(filp, arg);
1274 case XFS_IOC_SCRUB_METADATA:
1275 return xfs_ioc_scrub_metadata(filp, arg);
1277 case XFS_IOC_FD_TO_HANDLE:
1278 case XFS_IOC_PATH_TO_HANDLE:
1279 case XFS_IOC_PATH_TO_FSHANDLE: {
1280 xfs_fsop_handlereq_t hreq;
1282 if (copy_from_user(&hreq, arg, sizeof(hreq)))
1284 return xfs_find_handle(cmd, &hreq);
1286 case XFS_IOC_OPEN_BY_HANDLE: {
1287 xfs_fsop_handlereq_t hreq;
1289 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1291 return xfs_open_by_handle(filp, &hreq);
1294 case XFS_IOC_READLINK_BY_HANDLE: {
1295 xfs_fsop_handlereq_t hreq;
1297 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1299 return xfs_readlink_by_handle(filp, &hreq);
1301 case XFS_IOC_ATTRLIST_BY_HANDLE:
1302 return xfs_attrlist_by_handle(filp, arg);
1304 case XFS_IOC_ATTRMULTI_BY_HANDLE:
1305 return xfs_attrmulti_by_handle(filp, arg);
1307 case XFS_IOC_SWAPEXT: {
1308 struct xfs_swapext sxp;
1310 if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1312 error = mnt_want_write_file(filp);
1315 error = xfs_ioc_swapext(&sxp);
1316 mnt_drop_write_file(filp);
1320 case XFS_IOC_FSCOUNTS:
1321 return xfs_ioctl_fs_counts(mp, arg);
1323 case XFS_IOC_SET_RESBLKS:
1324 case XFS_IOC_GET_RESBLKS:
1325 return xfs_ioctl_getset_resblocks(filp, cmd, arg);
1327 case XFS_IOC_FSGROWFSDATA: {
1328 struct xfs_growfs_data in;
1330 if (copy_from_user(&in, arg, sizeof(in)))
1333 error = mnt_want_write_file(filp);
1336 error = xfs_growfs_data(mp, &in);
1337 mnt_drop_write_file(filp);
1341 case XFS_IOC_FSGROWFSLOG: {
1342 struct xfs_growfs_log in;
1344 if (copy_from_user(&in, arg, sizeof(in)))
1347 error = mnt_want_write_file(filp);
1350 error = xfs_growfs_log(mp, &in);
1351 mnt_drop_write_file(filp);
1355 case XFS_IOC_FSGROWFSRT: {
1358 if (copy_from_user(&in, arg, sizeof(in)))
1361 error = mnt_want_write_file(filp);
1364 error = xfs_growfs_rt(mp, &in);
1365 mnt_drop_write_file(filp);
1369 case XFS_IOC_GOINGDOWN: {
1372 if (!capable(CAP_SYS_ADMIN))
1375 if (get_user(in, (uint32_t __user *)arg))
1378 return xfs_fs_goingdown(mp, in);
1381 case XFS_IOC_ERROR_INJECTION: {
1382 xfs_error_injection_t in;
1384 if (!capable(CAP_SYS_ADMIN))
1387 if (copy_from_user(&in, arg, sizeof(in)))
1390 return xfs_errortag_add(mp, in.errtag);
1393 case XFS_IOC_ERROR_CLEARALL:
1394 if (!capable(CAP_SYS_ADMIN))
1397 return xfs_errortag_clearall(mp);
1399 case XFS_IOC_FREE_EOFBLOCKS: {
1400 struct xfs_fs_eofblocks eofb;
1401 struct xfs_icwalk icw;
1403 if (!capable(CAP_SYS_ADMIN))
1406 if (xfs_is_readonly(mp))
1409 if (copy_from_user(&eofb, arg, sizeof(eofb)))
1412 error = xfs_fs_eofblocks_from_user(&eofb, &icw);
1416 trace_xfs_ioc_free_eofblocks(mp, &icw, _RET_IP_);
1418 sb_start_write(mp->m_super);
1419 error = xfs_blockgc_free_space(mp, &icw);
1420 sb_end_write(mp->m_super);
1424 case XFS_IOC_EXCHANGE_RANGE:
1425 return xfs_ioc_exchange_range(filp, arg);
1426 case XFS_IOC_START_COMMIT:
1427 return xfs_ioc_start_commit(filp, arg);
1428 case XFS_IOC_COMMIT_RANGE:
1429 return xfs_ioc_commit_range(filp, arg);