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"
44 #include <linux/mount.h>
45 #include <linux/fileattr.h>
47 /* Return 0 on success or positive error */
49 xfs_fsbulkstat_one_fmt(
50 struct xfs_ibulk *breq,
51 const struct xfs_bulkstat *bstat)
55 xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
56 if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
58 return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
63 struct xfs_ibulk *breq,
64 const struct xfs_inumbers *igrp)
66 struct xfs_inogrp ig1;
68 xfs_inumbers_to_inogrp(&ig1, igrp);
69 if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
71 return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
80 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
81 struct xfs_fsop_bulkreq bulkreq;
82 struct xfs_ibulk breq = {
84 .idmap = file_mnt_idmap(file),
90 /* done = 1 if there are more stats to get and if bulkstat */
91 /* should be called again (unused here, but used in dmapi) */
93 if (!capable(CAP_SYS_ADMIN))
96 if (xfs_is_shutdown(mp))
99 if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
102 if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
105 if (bulkreq.icount <= 0)
108 if (bulkreq.ubuffer == NULL)
111 breq.ubuffer = bulkreq.ubuffer;
112 breq.icount = bulkreq.icount;
115 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
116 * that we want to stat. However, FSINUMBERS and FSBULKSTAT expect
117 * that *lastip contains either zero or the number of the last inode to
118 * be examined by the previous call and return results starting with
119 * the next inode after that. The new bulk request back end functions
120 * take the inode to start with, so we have to compute the startino
121 * parameter from lastino to maintain correct function. lastino == 0
122 * is a special case because it has traditionally meant "first inode
125 if (cmd == XFS_IOC_FSINUMBERS) {
126 breq.startino = lastino ? lastino + 1 : 0;
127 error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
128 lastino = breq.startino - 1;
129 } else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
130 breq.startino = lastino;
132 error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
133 } else { /* XFS_IOC_FSBULKSTAT */
134 breq.startino = lastino ? lastino + 1 : 0;
135 error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
136 lastino = breq.startino - 1;
142 if (bulkreq.lastip != NULL &&
143 copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
146 if (bulkreq.ocount != NULL &&
147 copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
153 /* Return 0 on success or positive error */
156 struct xfs_ibulk *breq,
157 const struct xfs_bulkstat *bstat)
159 if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
161 return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
165 * Check the incoming bulk request @hdr from userspace and initialize the
166 * internal @breq bulk request appropriately. Returns 0 if the bulk request
167 * should proceed; -ECANCELED if there's nothing to do; or the usual
168 * negative error code.
172 struct xfs_mount *mp,
173 const struct xfs_bulk_ireq *hdr,
174 struct xfs_ibulk *breq,
175 void __user *ubuffer)
177 if (hdr->icount == 0 ||
178 (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
179 memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
182 breq->startino = hdr->ino;
183 breq->ubuffer = ubuffer;
184 breq->icount = hdr->icount;
189 * The @ino parameter is a special value, so we must look it up here.
190 * We're not allowed to have IREQ_AGNO, and we only return one inode
193 if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
194 if (hdr->flags & XFS_BULK_IREQ_AGNO)
198 case XFS_BULK_IREQ_SPECIAL_ROOT:
199 breq->startino = mp->m_sb.sb_rootino;
208 * The IREQ_AGNO flag means that we only want results from a given AG.
209 * If @hdr->ino is zero, we start iterating in that AG. If @hdr->ino is
210 * beyond the specified AG then we return no results.
212 if (hdr->flags & XFS_BULK_IREQ_AGNO) {
213 if (hdr->agno >= mp->m_sb.sb_agcount)
216 if (breq->startino == 0)
217 breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
218 else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
221 breq->flags |= XFS_IBULK_SAME_AG;
223 /* Asking for an inode past the end of the AG? We're done! */
224 if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
226 } else if (hdr->agno)
229 /* Asking for an inode past the end of the FS? We're done! */
230 if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
233 if (hdr->flags & XFS_BULK_IREQ_NREXT64)
234 breq->flags |= XFS_IBULK_NREXT64;
240 * Update the userspace bulk request @hdr to reflect the end state of the
241 * internal bulk request @breq.
244 xfs_bulk_ireq_teardown(
245 struct xfs_bulk_ireq *hdr,
246 struct xfs_ibulk *breq)
248 hdr->ino = breq->startino;
249 hdr->ocount = breq->ocount;
252 /* Handle the v5 bulkstat ioctl. */
257 struct xfs_bulkstat_req __user *arg)
259 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
260 struct xfs_bulk_ireq hdr;
261 struct xfs_ibulk breq = {
263 .idmap = file_mnt_idmap(file),
267 if (!capable(CAP_SYS_ADMIN))
270 if (xfs_is_shutdown(mp))
273 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
276 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
277 if (error == -ECANCELED)
282 error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
287 xfs_bulk_ireq_teardown(&hdr, &breq);
288 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
296 struct xfs_ibulk *breq,
297 const struct xfs_inumbers *igrp)
299 if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
301 return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
304 /* Handle the v5 inumbers ioctl. */
307 struct xfs_mount *mp,
309 struct xfs_inumbers_req __user *arg)
311 struct xfs_bulk_ireq hdr;
312 struct xfs_ibulk breq = {
317 if (!capable(CAP_SYS_ADMIN))
320 if (xfs_is_shutdown(mp))
323 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
326 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
327 if (error == -ECANCELED)
332 error = xfs_inumbers(&breq, xfs_inumbers_fmt);
337 xfs_bulk_ireq_teardown(&hdr, &breq);
338 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
346 struct xfs_mount *mp,
350 struct xfs_fsop_geom fsgeo;
353 xfs_fs_geometry(mp, &fsgeo, struct_version);
355 if (struct_version <= 3)
356 len = sizeof(struct xfs_fsop_geom_v1);
357 else if (struct_version == 4)
358 len = sizeof(struct xfs_fsop_geom_v4);
360 xfs_fsop_geom_health(mp, &fsgeo);
364 if (copy_to_user(arg, &fsgeo, len))
371 struct xfs_mount *mp,
374 struct xfs_perag *pag;
375 struct xfs_ag_geometry ageo;
378 if (copy_from_user(&ageo, arg, sizeof(ageo)))
382 if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
385 pag = xfs_perag_get(mp, ageo.ag_number);
389 error = xfs_ag_get_geometry(pag, &ageo);
394 if (copy_to_user(arg, &ageo, sizeof(ageo)))
400 * Linux extended inode flags interface.
405 struct xfs_inode *ip,
409 struct xfs_mount *mp = ip->i_mount;
410 struct xfs_ifork *ifp = xfs_ifork_ptr(ip, whichfork);
412 fileattr_fill_xflags(fa, xfs_ip2xflags(ip));
414 if (ip->i_diflags & XFS_DIFLAG_EXTSIZE) {
415 fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
416 } else if (ip->i_diflags & XFS_DIFLAG_EXTSZINHERIT) {
418 * Don't let a misaligned extent size hint on a directory
419 * escape to userspace if it won't pass the setattr checks
422 if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
423 xfs_extlen_to_rtxmod(mp, ip->i_extsize) > 0) {
424 fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE |
425 FS_XFLAG_EXTSZINHERIT);
428 fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
432 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
433 fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
434 fa->fsx_projid = ip->i_projid;
435 if (ifp && !xfs_need_iread_extents(ifp))
436 fa->fsx_nextents = xfs_iext_count(ifp);
438 fa->fsx_nextents = xfs_ifork_nextents(ifp);
448 xfs_ilock(ip, XFS_ILOCK_SHARED);
449 xfs_fill_fsxattr(ip, XFS_ATTR_FORK, &fa);
450 xfs_iunlock(ip, XFS_ILOCK_SHARED);
452 return copy_fsxattr_to_user(&fa, arg);
457 struct dentry *dentry,
460 struct xfs_inode *ip = XFS_I(d_inode(dentry));
462 if (d_is_special(dentry))
465 xfs_ilock(ip, XFS_ILOCK_SHARED);
466 xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
467 xfs_iunlock(ip, XFS_ILOCK_SHARED);
473 xfs_ioctl_setattr_xflags(
474 struct xfs_trans *tp,
475 struct xfs_inode *ip,
478 struct xfs_mount *mp = ip->i_mount;
479 bool rtflag = (fa->fsx_xflags & FS_XFLAG_REALTIME);
482 if (rtflag != XFS_IS_REALTIME_INODE(ip)) {
483 /* Can't change realtime flag if any extents are allocated. */
484 if (ip->i_df.if_nextents || ip->i_delayed_blks)
489 /* If realtime flag is set then must have realtime device */
490 if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
491 xfs_extlen_to_rtxmod(mp, ip->i_extsize))
494 /* Clear reflink if we are actually able to set the rt flag. */
495 if (xfs_is_reflink_inode(ip))
496 ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
499 /* diflags2 only valid for v3 inodes. */
500 i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
501 if (i_flags2 && !xfs_has_v3inodes(mp))
504 ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
505 ip->i_diflags2 = i_flags2;
507 xfs_diflags_to_iflags(ip, false);
510 * Make the stable writes flag match that of the device the inode
511 * resides on when flipping the RT flag.
513 if (rtflag != XFS_IS_REALTIME_INODE(ip) && S_ISREG(VFS_I(ip)->i_mode))
514 xfs_update_stable_writes(ip);
516 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
517 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
518 XFS_STATS_INC(mp, xs_ig_attrchg);
523 xfs_ioctl_setattr_prepare_dax(
524 struct xfs_inode *ip,
527 struct xfs_mount *mp = ip->i_mount;
528 struct inode *inode = VFS_I(ip);
530 if (S_ISDIR(inode->i_mode))
533 if (xfs_has_dax_always(mp) || xfs_has_dax_never(mp))
536 if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
537 !(ip->i_diflags2 & XFS_DIFLAG2_DAX)) ||
538 (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
539 (ip->i_diflags2 & XFS_DIFLAG2_DAX)))
540 d_mark_dontcache(inode);
544 * Set up the transaction structure for the setattr operation, checking that we
545 * have permission to do so. On success, return a clean transaction and the
546 * inode locked exclusively ready for further operation specific checks. On
547 * failure, return an error without modifying or locking the inode.
549 static struct xfs_trans *
550 xfs_ioctl_setattr_get_trans(
551 struct xfs_inode *ip,
552 struct xfs_dquot *pdqp)
554 struct xfs_mount *mp = ip->i_mount;
555 struct xfs_trans *tp;
558 if (xfs_is_readonly(mp))
561 if (xfs_is_shutdown(mp))
564 error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
565 has_capability_noaudit(current, CAP_FOWNER), &tp);
569 if (xfs_has_wsync(mp))
570 xfs_trans_set_sync(tp);
575 return ERR_PTR(error);
579 * Validate a proposed extent size hint. For regular files, the hint can only
580 * be changed if no extents are allocated.
583 xfs_ioctl_setattr_check_extsize(
584 struct xfs_inode *ip,
587 struct xfs_mount *mp = ip->i_mount;
588 xfs_failaddr_t failaddr;
589 uint16_t new_diflags;
594 if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
595 XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
598 if (fa->fsx_extsize & mp->m_blockmask)
601 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
604 * Inode verifiers do not check that the extent size hint is an integer
605 * multiple of the rt extent size on a directory with both rtinherit
606 * and extszinherit flags set. Don't let sysadmins misconfigure
609 if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
610 (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
611 unsigned int rtextsize_bytes;
613 rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
614 if (fa->fsx_extsize % rtextsize_bytes)
618 failaddr = xfs_inode_validate_extsize(ip->i_mount,
619 XFS_B_TO_FSB(mp, fa->fsx_extsize),
620 VFS_I(ip)->i_mode, new_diflags);
621 return failaddr != NULL ? -EINVAL : 0;
625 xfs_ioctl_setattr_check_cowextsize(
626 struct xfs_inode *ip,
629 struct xfs_mount *mp = ip->i_mount;
630 xfs_failaddr_t failaddr;
631 uint64_t new_diflags2;
632 uint16_t new_diflags;
637 if (fa->fsx_cowextsize & mp->m_blockmask)
640 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
641 new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
643 failaddr = xfs_inode_validate_cowextsize(ip->i_mount,
644 XFS_B_TO_FSB(mp, fa->fsx_cowextsize),
645 VFS_I(ip)->i_mode, new_diflags, new_diflags2);
646 return failaddr != NULL ? -EINVAL : 0;
650 xfs_ioctl_setattr_check_projid(
651 struct xfs_inode *ip,
657 /* Disallow 32bit project ids if 32bit IDs are not enabled. */
658 if (fa->fsx_projid > (uint16_t)-1 &&
659 !xfs_has_projid32(ip->i_mount))
666 struct mnt_idmap *idmap,
667 struct dentry *dentry,
670 struct xfs_inode *ip = XFS_I(d_inode(dentry));
671 struct xfs_mount *mp = ip->i_mount;
672 struct xfs_trans *tp;
673 struct xfs_dquot *pdqp = NULL;
674 struct xfs_dquot *olddquot = NULL;
677 trace_xfs_ioctl_setattr(ip);
679 if (d_is_special(dentry))
682 if (!fa->fsx_valid) {
683 if (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL |
684 FS_NOATIME_FL | FS_NODUMP_FL |
685 FS_SYNC_FL | FS_DAX_FL | FS_PROJINHERIT_FL))
689 error = xfs_ioctl_setattr_check_projid(ip, fa);
694 * If disk quotas is on, we make sure that the dquots do exist on disk,
695 * before we start any other transactions. Trying to do this later
696 * is messy. We don't care to take a readlock to look at the ids
697 * in inode here, because we can't hold it across the trans_reserve.
698 * If the IDs do change before we take the ilock, we're covered
699 * because the i_*dquot fields will get updated anyway.
701 if (fa->fsx_valid && XFS_IS_QUOTA_ON(mp)) {
702 error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
703 VFS_I(ip)->i_gid, fa->fsx_projid,
704 XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
709 xfs_ioctl_setattr_prepare_dax(ip, fa);
711 tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
714 goto error_free_dquots;
717 error = xfs_ioctl_setattr_check_extsize(ip, fa);
719 goto error_trans_cancel;
721 error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
723 goto error_trans_cancel;
725 error = xfs_ioctl_setattr_xflags(tp, ip, fa);
727 goto error_trans_cancel;
732 * Change file ownership. Must be the owner or privileged. CAP_FSETID
733 * overrides the following restrictions:
735 * The set-user-ID and set-group-ID bits of a file will be cleared upon
736 * successful return from chown()
739 if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
740 !capable_wrt_inode_uidgid(idmap, VFS_I(ip), CAP_FSETID))
741 VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
743 /* Change the ownerships and register project quota modifications */
744 if (ip->i_projid != fa->fsx_projid) {
745 if (XFS_IS_PQUOTA_ON(mp)) {
746 olddquot = xfs_qm_vop_chown(tp, ip,
747 &ip->i_pdquot, pdqp);
749 ip->i_projid = fa->fsx_projid;
753 * Only set the extent size hint if we've already determined that the
754 * extent size hint should be set on the inode. If no extent size flags
755 * are set on the inode then unconditionally clear the extent size hint.
757 if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
758 ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
762 if (xfs_has_v3inodes(mp)) {
763 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
764 ip->i_cowextsize = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
766 ip->i_cowextsize = 0;
770 error = xfs_trans_commit(tp);
773 * Release any dquot(s) the inode had kept before chown.
775 xfs_qm_dqrele(olddquot);
781 xfs_trans_cancel(tp);
790 struct getbmapx __user *u,
793 if (put_user(p->bmv_offset, &u->bmv_offset) ||
794 put_user(p->bmv_block, &u->bmv_block) ||
795 put_user(p->bmv_length, &u->bmv_length) ||
796 put_user(0, &u->bmv_count) ||
797 put_user(0, &u->bmv_entries))
799 if (recsize < sizeof(struct getbmapx))
801 if (put_user(0, &u->bmv_iflags) ||
802 put_user(p->bmv_oflags, &u->bmv_oflags) ||
803 put_user(0, &u->bmv_unused1) ||
804 put_user(0, &u->bmv_unused2))
815 struct getbmapx bmx = { 0 };
816 struct kgetbmap *buf;
821 case XFS_IOC_GETBMAPA:
822 bmx.bmv_iflags = BMV_IF_ATTRFORK;
824 case XFS_IOC_GETBMAP:
825 /* struct getbmap is a strict subset of struct getbmapx. */
826 recsize = sizeof(struct getbmap);
828 case XFS_IOC_GETBMAPX:
829 recsize = sizeof(struct getbmapx);
835 if (copy_from_user(&bmx, arg, recsize))
838 if (bmx.bmv_count < 2)
840 if (bmx.bmv_count >= INT_MAX / recsize)
843 buf = kvcalloc(bmx.bmv_count, sizeof(*buf), GFP_KERNEL);
847 error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
852 if (copy_to_user(arg, &bmx, recsize))
856 for (i = 0; i < bmx.bmv_entries; i++) {
857 if (!xfs_getbmap_format(buf + i, arg, recsize))
870 struct xfs_inode *ip,
871 struct fsmap_head __user *arg)
873 struct xfs_fsmap_head xhead = {0};
874 struct fsmap_head head;
877 __u32 last_flags = 0;
881 if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
883 if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
884 memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
885 sizeof(head.fmh_keys[0].fmr_reserved)) ||
886 memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
887 sizeof(head.fmh_keys[1].fmr_reserved)))
891 * Use an internal memory buffer so that we don't have to copy fsmap
892 * data to userspace while holding locks. Start by trying to allocate
893 * up to 128k for the buffer, but fall back to a single page if needed.
895 count = min_t(unsigned int, head.fmh_count,
896 131072 / sizeof(struct fsmap));
897 recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
899 count = min_t(unsigned int, head.fmh_count,
900 PAGE_SIZE / sizeof(struct fsmap));
901 recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
906 xhead.fmh_iflags = head.fmh_iflags;
907 xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
908 xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
910 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
911 trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
913 head.fmh_entries = 0;
915 struct fsmap __user *user_recs;
916 struct fsmap *last_rec;
918 user_recs = &arg->fmh_recs[head.fmh_entries];
919 xhead.fmh_entries = 0;
920 xhead.fmh_count = min_t(unsigned int, count,
921 head.fmh_count - head.fmh_entries);
923 /* Run query, record how many entries we got. */
924 error = xfs_getfsmap(ip->i_mount, &xhead, recs);
928 * There are no more records in the result set. Copy
929 * whatever we got to userspace and break out.
935 * The internal memory buffer is full. Copy whatever
936 * records we got to userspace and go again if we have
937 * not yet filled the userspace buffer.
944 head.fmh_entries += xhead.fmh_entries;
945 head.fmh_oflags = xhead.fmh_oflags;
948 * If the caller wanted a record count or there aren't any
949 * new records to return, we're done.
951 if (head.fmh_count == 0 || xhead.fmh_entries == 0)
954 /* Copy all the records we got out to userspace. */
955 if (copy_to_user(user_recs, recs,
956 xhead.fmh_entries * sizeof(struct fsmap))) {
961 /* Remember the last record flags we copied to userspace. */
962 last_rec = &recs[xhead.fmh_entries - 1];
963 last_flags = last_rec->fmr_flags;
965 /* Set up the low key for the next iteration. */
966 xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
967 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
968 } while (!done && head.fmh_entries < head.fmh_count);
971 * If there are no more records in the query result set and we're not
972 * in counting mode, mark the last record returned with the LAST flag.
974 if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
975 struct fsmap __user *user_rec;
977 last_flags |= FMR_OF_LAST;
978 user_rec = &arg->fmh_recs[head.fmh_entries - 1];
980 if (copy_to_user(&user_rec->fmr_flags, &last_flags,
981 sizeof(last_flags))) {
987 /* copy back header */
988 if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1002 xfs_inode_t *ip, *tip;
1006 /* Pull information for the target fd */
1007 f = fdget((int)sxp->sx_fdtarget);
1013 if (!(f.file->f_mode & FMODE_WRITE) ||
1014 !(f.file->f_mode & FMODE_READ) ||
1015 (f.file->f_flags & O_APPEND)) {
1020 tmp = fdget((int)sxp->sx_fdtmp);
1026 if (!(tmp.file->f_mode & FMODE_WRITE) ||
1027 !(tmp.file->f_mode & FMODE_READ) ||
1028 (tmp.file->f_flags & O_APPEND)) {
1030 goto out_put_tmp_file;
1033 if (IS_SWAPFILE(file_inode(f.file)) ||
1034 IS_SWAPFILE(file_inode(tmp.file))) {
1036 goto out_put_tmp_file;
1040 * We need to ensure that the fds passed in point to XFS inodes
1041 * before we cast and access them as XFS structures as we have no
1042 * control over what the user passes us here.
1044 if (f.file->f_op != &xfs_file_operations ||
1045 tmp.file->f_op != &xfs_file_operations) {
1047 goto out_put_tmp_file;
1050 ip = XFS_I(file_inode(f.file));
1051 tip = XFS_I(file_inode(tmp.file));
1053 if (ip->i_mount != tip->i_mount) {
1055 goto out_put_tmp_file;
1058 if (ip->i_ino == tip->i_ino) {
1060 goto out_put_tmp_file;
1063 if (xfs_is_shutdown(ip->i_mount)) {
1065 goto out_put_tmp_file;
1068 error = xfs_swap_extents(ip, tip, sxp);
1080 struct xfs_mount *mp,
1081 char __user *user_label)
1083 struct xfs_sb *sbp = &mp->m_sb;
1084 char label[XFSLABEL_MAX + 1];
1087 BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1089 /* 1 larger than sb_fname, so this ensures a trailing NUL char */
1090 memset(label, 0, sizeof(label));
1091 spin_lock(&mp->m_sb_lock);
1092 strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1093 spin_unlock(&mp->m_sb_lock);
1095 if (copy_to_user(user_label, label, sizeof(label)))
1103 struct xfs_mount *mp,
1104 char __user *newlabel)
1106 struct xfs_sb *sbp = &mp->m_sb;
1107 char label[XFSLABEL_MAX + 1];
1111 if (!capable(CAP_SYS_ADMIN))
1114 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1115 * smaller, at 12 bytes. We copy one more to be sure we find the
1116 * (required) NULL character to test the incoming label length.
1117 * NB: The on disk label doesn't need to be null terminated.
1119 if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1121 len = strnlen(label, XFSLABEL_MAX + 1);
1122 if (len > sizeof(sbp->sb_fname))
1125 error = mnt_want_write_file(filp);
1129 spin_lock(&mp->m_sb_lock);
1130 memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1131 memcpy(sbp->sb_fname, label, len);
1132 spin_unlock(&mp->m_sb_lock);
1135 * Now we do several things to satisfy userspace.
1136 * In addition to normal logging of the primary superblock, we also
1137 * immediately write these changes to sector zero for the primary, then
1138 * update all backup supers (as xfs_db does for a label change), then
1139 * invalidate the block device page cache. This is so that any prior
1140 * buffered reads from userspace (i.e. from blkid) are invalidated,
1141 * and userspace will see the newly-written label.
1143 error = xfs_sync_sb_buf(mp);
1147 * growfs also updates backup supers so lock against that.
1149 mutex_lock(&mp->m_growlock);
1150 error = xfs_update_secondary_sbs(mp);
1151 mutex_unlock(&mp->m_growlock);
1153 invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1156 mnt_drop_write_file(filp);
1161 xfs_fs_eofblocks_from_user(
1162 struct xfs_fs_eofblocks *src,
1163 struct xfs_icwalk *dst)
1165 if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1168 if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1171 if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1172 memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1176 if (src->eof_flags & XFS_EOF_FLAGS_SYNC)
1177 dst->icw_flags |= XFS_ICWALK_FLAG_SYNC;
1178 if (src->eof_flags & XFS_EOF_FLAGS_UID)
1179 dst->icw_flags |= XFS_ICWALK_FLAG_UID;
1180 if (src->eof_flags & XFS_EOF_FLAGS_GID)
1181 dst->icw_flags |= XFS_ICWALK_FLAG_GID;
1182 if (src->eof_flags & XFS_EOF_FLAGS_PRID)
1183 dst->icw_flags |= XFS_ICWALK_FLAG_PRID;
1184 if (src->eof_flags & XFS_EOF_FLAGS_MINFILESIZE)
1185 dst->icw_flags |= XFS_ICWALK_FLAG_MINFILESIZE;
1187 dst->icw_prid = src->eof_prid;
1188 dst->icw_min_file_size = src->eof_min_file_size;
1190 dst->icw_uid = INVALID_UID;
1191 if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1192 dst->icw_uid = make_kuid(current_user_ns(), src->eof_uid);
1193 if (!uid_valid(dst->icw_uid))
1197 dst->icw_gid = INVALID_GID;
1198 if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1199 dst->icw_gid = make_kgid(current_user_ns(), src->eof_gid);
1200 if (!gid_valid(dst->icw_gid))
1207 xfs_ioctl_getset_resblocks(
1212 struct xfs_mount *mp = XFS_I(file_inode(filp))->i_mount;
1213 struct xfs_fsop_resblks fsop = { };
1216 if (!capable(CAP_SYS_ADMIN))
1219 if (cmd == XFS_IOC_SET_RESBLKS) {
1220 if (xfs_is_readonly(mp))
1223 if (copy_from_user(&fsop, arg, sizeof(fsop)))
1226 error = mnt_want_write_file(filp);
1229 error = xfs_reserve_blocks(mp, fsop.resblks);
1230 mnt_drop_write_file(filp);
1235 spin_lock(&mp->m_sb_lock);
1236 fsop.resblks = mp->m_resblks;
1237 fsop.resblks_avail = mp->m_resblks_avail;
1238 spin_unlock(&mp->m_sb_lock);
1240 if (copy_to_user(arg, &fsop, sizeof(fsop)))
1246 xfs_ioctl_fs_counts(
1247 struct xfs_mount *mp,
1248 struct xfs_fsop_counts __user *uarg)
1250 struct xfs_fsop_counts out = {
1251 .allocino = percpu_counter_read_positive(&mp->m_icount),
1252 .freeino = percpu_counter_read_positive(&mp->m_ifree),
1253 .freedata = percpu_counter_read_positive(&mp->m_fdblocks) -
1254 xfs_fdblocks_unavailable(mp),
1255 .freertx = percpu_counter_read_positive(&mp->m_frextents),
1258 if (copy_to_user(uarg, &out, sizeof(out)))
1264 * These long-unused ioctls were removed from the official ioctl API in 5.17,
1265 * but retain these definitions so that we can log warnings about them.
1267 #define XFS_IOC_ALLOCSP _IOW ('X', 10, struct xfs_flock64)
1268 #define XFS_IOC_FREESP _IOW ('X', 11, struct xfs_flock64)
1269 #define XFS_IOC_ALLOCSP64 _IOW ('X', 36, struct xfs_flock64)
1270 #define XFS_IOC_FREESP64 _IOW ('X', 37, struct xfs_flock64)
1273 * Note: some of the ioctl's return positive numbers as a
1274 * byte count indicating success, such as readlink_by_handle.
1275 * So we don't "sign flip" like most other routines. This means
1276 * true errors need to be returned as a negative value.
1284 struct inode *inode = file_inode(filp);
1285 struct xfs_inode *ip = XFS_I(inode);
1286 struct xfs_mount *mp = ip->i_mount;
1287 void __user *arg = (void __user *)p;
1290 trace_xfs_file_ioctl(ip);
1294 return xfs_ioc_trim(mp, arg);
1295 case FS_IOC_GETFSLABEL:
1296 return xfs_ioc_getlabel(mp, arg);
1297 case FS_IOC_SETFSLABEL:
1298 return xfs_ioc_setlabel(filp, mp, arg);
1299 case XFS_IOC_ALLOCSP:
1300 case XFS_IOC_FREESP:
1301 case XFS_IOC_ALLOCSP64:
1302 case XFS_IOC_FREESP64:
1304 "%s should use fallocate; XFS_IOC_{ALLOC,FREE}SP ioctl unsupported",
1307 case XFS_IOC_DIOINFO: {
1308 struct xfs_buftarg *target = xfs_inode_buftarg(ip);
1311 da.d_mem = da.d_miniosz = target->bt_logical_sectorsize;
1312 da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1314 if (copy_to_user(arg, &da, sizeof(da)))
1319 case XFS_IOC_FSBULKSTAT_SINGLE:
1320 case XFS_IOC_FSBULKSTAT:
1321 case XFS_IOC_FSINUMBERS:
1322 return xfs_ioc_fsbulkstat(filp, cmd, arg);
1324 case XFS_IOC_BULKSTAT:
1325 return xfs_ioc_bulkstat(filp, cmd, arg);
1326 case XFS_IOC_INUMBERS:
1327 return xfs_ioc_inumbers(mp, cmd, arg);
1329 case XFS_IOC_FSGEOMETRY_V1:
1330 return xfs_ioc_fsgeometry(mp, arg, 3);
1331 case XFS_IOC_FSGEOMETRY_V4:
1332 return xfs_ioc_fsgeometry(mp, arg, 4);
1333 case XFS_IOC_FSGEOMETRY:
1334 return xfs_ioc_fsgeometry(mp, arg, 5);
1336 case XFS_IOC_AG_GEOMETRY:
1337 return xfs_ioc_ag_geometry(mp, arg);
1339 case XFS_IOC_GETVERSION:
1340 return put_user(inode->i_generation, (int __user *)arg);
1342 case XFS_IOC_FSGETXATTRA:
1343 return xfs_ioc_fsgetxattra(ip, arg);
1344 case XFS_IOC_GETPARENTS:
1345 return xfs_ioc_getparents(filp, arg);
1346 case XFS_IOC_GETPARENTS_BY_HANDLE:
1347 return xfs_ioc_getparents_by_handle(filp, arg);
1348 case XFS_IOC_GETBMAP:
1349 case XFS_IOC_GETBMAPA:
1350 case XFS_IOC_GETBMAPX:
1351 return xfs_ioc_getbmap(filp, cmd, arg);
1353 case FS_IOC_GETFSMAP:
1354 return xfs_ioc_getfsmap(ip, arg);
1356 case XFS_IOC_SCRUBV_METADATA:
1357 return xfs_ioc_scrubv_metadata(filp, arg);
1358 case XFS_IOC_SCRUB_METADATA:
1359 return xfs_ioc_scrub_metadata(filp, arg);
1361 case XFS_IOC_FD_TO_HANDLE:
1362 case XFS_IOC_PATH_TO_HANDLE:
1363 case XFS_IOC_PATH_TO_FSHANDLE: {
1364 xfs_fsop_handlereq_t hreq;
1366 if (copy_from_user(&hreq, arg, sizeof(hreq)))
1368 return xfs_find_handle(cmd, &hreq);
1370 case XFS_IOC_OPEN_BY_HANDLE: {
1371 xfs_fsop_handlereq_t hreq;
1373 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1375 return xfs_open_by_handle(filp, &hreq);
1378 case XFS_IOC_READLINK_BY_HANDLE: {
1379 xfs_fsop_handlereq_t hreq;
1381 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1383 return xfs_readlink_by_handle(filp, &hreq);
1385 case XFS_IOC_ATTRLIST_BY_HANDLE:
1386 return xfs_attrlist_by_handle(filp, arg);
1388 case XFS_IOC_ATTRMULTI_BY_HANDLE:
1389 return xfs_attrmulti_by_handle(filp, arg);
1391 case XFS_IOC_SWAPEXT: {
1392 struct xfs_swapext sxp;
1394 if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1396 error = mnt_want_write_file(filp);
1399 error = xfs_ioc_swapext(&sxp);
1400 mnt_drop_write_file(filp);
1404 case XFS_IOC_FSCOUNTS:
1405 return xfs_ioctl_fs_counts(mp, arg);
1407 case XFS_IOC_SET_RESBLKS:
1408 case XFS_IOC_GET_RESBLKS:
1409 return xfs_ioctl_getset_resblocks(filp, cmd, arg);
1411 case XFS_IOC_FSGROWFSDATA: {
1412 struct xfs_growfs_data in;
1414 if (copy_from_user(&in, arg, sizeof(in)))
1417 error = mnt_want_write_file(filp);
1420 error = xfs_growfs_data(mp, &in);
1421 mnt_drop_write_file(filp);
1425 case XFS_IOC_FSGROWFSLOG: {
1426 struct xfs_growfs_log in;
1428 if (copy_from_user(&in, arg, sizeof(in)))
1431 error = mnt_want_write_file(filp);
1434 error = xfs_growfs_log(mp, &in);
1435 mnt_drop_write_file(filp);
1439 case XFS_IOC_FSGROWFSRT: {
1442 if (copy_from_user(&in, arg, sizeof(in)))
1445 error = mnt_want_write_file(filp);
1448 error = xfs_growfs_rt(mp, &in);
1449 mnt_drop_write_file(filp);
1453 case XFS_IOC_GOINGDOWN: {
1456 if (!capable(CAP_SYS_ADMIN))
1459 if (get_user(in, (uint32_t __user *)arg))
1462 return xfs_fs_goingdown(mp, in);
1465 case XFS_IOC_ERROR_INJECTION: {
1466 xfs_error_injection_t in;
1468 if (!capable(CAP_SYS_ADMIN))
1471 if (copy_from_user(&in, arg, sizeof(in)))
1474 return xfs_errortag_add(mp, in.errtag);
1477 case XFS_IOC_ERROR_CLEARALL:
1478 if (!capable(CAP_SYS_ADMIN))
1481 return xfs_errortag_clearall(mp);
1483 case XFS_IOC_FREE_EOFBLOCKS: {
1484 struct xfs_fs_eofblocks eofb;
1485 struct xfs_icwalk icw;
1487 if (!capable(CAP_SYS_ADMIN))
1490 if (xfs_is_readonly(mp))
1493 if (copy_from_user(&eofb, arg, sizeof(eofb)))
1496 error = xfs_fs_eofblocks_from_user(&eofb, &icw);
1500 trace_xfs_ioc_free_eofblocks(mp, &icw, _RET_IP_);
1502 sb_start_write(mp->m_super);
1503 error = xfs_blockgc_free_space(mp, &icw);
1504 sb_end_write(mp->m_super);
1508 case XFS_IOC_EXCHANGE_RANGE:
1509 return xfs_ioc_exchange_range(filp, arg);