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_export.h"
27 #include "xfs_trace.h"
28 #include "xfs_icache.h"
29 #include "xfs_trans.h"
31 #include "xfs_btree.h"
32 #include <linux/fsmap.h>
33 #include "xfs_fsmap.h"
34 #include "scrub/xfs_scrub.h"
37 #include "xfs_health.h"
38 #include "xfs_reflink.h"
39 #include "xfs_ioctl.h"
40 #include "xfs_xattr.h"
41 #include "xfs_rtbitmap.h"
43 #include <linux/mount.h>
44 #include <linux/namei.h>
45 #include <linux/fileattr.h>
48 * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
49 * a file or fs handle.
51 * XFS_IOC_PATH_TO_FSHANDLE
52 * returns fs handle for a mount point or path within that mount point
53 * XFS_IOC_FD_TO_HANDLE
54 * returns full handle for a FD opened in user space
55 * XFS_IOC_PATH_TO_HANDLE
56 * returns full handle for a path
61 xfs_fsop_handlereq_t *hreq)
71 if (cmd == XFS_IOC_FD_TO_HANDLE) {
75 inode = file_inode(f.file);
77 error = user_path_at(AT_FDCWD, hreq->path, 0, &path);
80 inode = d_inode(path.dentry);
85 * We can only generate handles for inodes residing on a XFS filesystem,
86 * and only for regular files, directories or symbolic links.
89 if (inode->i_sb->s_magic != XFS_SB_MAGIC)
93 if (!S_ISREG(inode->i_mode) &&
94 !S_ISDIR(inode->i_mode) &&
95 !S_ISLNK(inode->i_mode))
99 memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
101 if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
103 * This handle only contains an fsid, zero the rest.
105 memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
106 hsize = sizeof(xfs_fsid_t);
108 handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
109 sizeof(handle.ha_fid.fid_len);
110 handle.ha_fid.fid_pad = 0;
111 handle.ha_fid.fid_gen = inode->i_generation;
112 handle.ha_fid.fid_ino = ip->i_ino;
113 hsize = sizeof(xfs_handle_t);
117 if (copy_to_user(hreq->ohandle, &handle, hsize) ||
118 copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
124 if (cmd == XFS_IOC_FD_TO_HANDLE)
132 * No need to do permission checks on the various pathname components
133 * as the handle operations are privileged.
136 xfs_handle_acceptable(
138 struct dentry *dentry)
144 * Convert userspace handle data into a dentry.
147 xfs_handle_to_dentry(
148 struct file *parfilp,
149 void __user *uhandle,
153 struct xfs_fid64 fid;
156 * Only allow handle opens under a directory.
158 if (!S_ISDIR(file_inode(parfilp)->i_mode))
159 return ERR_PTR(-ENOTDIR);
161 if (hlen != sizeof(xfs_handle_t))
162 return ERR_PTR(-EINVAL);
163 if (copy_from_user(&handle, uhandle, hlen))
164 return ERR_PTR(-EFAULT);
165 if (handle.ha_fid.fid_len !=
166 sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
167 return ERR_PTR(-EINVAL);
169 memset(&fid, 0, sizeof(struct fid));
170 fid.ino = handle.ha_fid.fid_ino;
171 fid.gen = handle.ha_fid.fid_gen;
173 return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
174 FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
175 xfs_handle_acceptable, NULL);
178 STATIC struct dentry *
179 xfs_handlereq_to_dentry(
180 struct file *parfilp,
181 xfs_fsop_handlereq_t *hreq)
183 return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
188 struct file *parfilp,
189 xfs_fsop_handlereq_t *hreq)
191 const struct cred *cred = current_cred();
197 struct dentry *dentry;
201 if (!capable(CAP_SYS_ADMIN))
204 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
206 return PTR_ERR(dentry);
207 inode = d_inode(dentry);
209 /* Restrict xfs_open_by_handle to directories & regular files. */
210 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
215 #if BITS_PER_LONG != 32
216 hreq->oflags |= O_LARGEFILE;
219 permflag = hreq->oflags;
220 fmode = OPEN_FMODE(permflag);
221 if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
222 (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
227 if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
232 /* Can't write directories. */
233 if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
238 fd = get_unused_fd_flags(0);
244 path.mnt = parfilp->f_path.mnt;
245 path.dentry = dentry;
246 filp = dentry_open(&path, hreq->oflags, cred);
250 return PTR_ERR(filp);
253 if (S_ISREG(inode->i_mode)) {
254 filp->f_flags |= O_NOATIME;
255 filp->f_mode |= FMODE_NOCMTIME;
258 fd_install(fd, filp);
267 xfs_readlink_by_handle(
268 struct file *parfilp,
269 xfs_fsop_handlereq_t *hreq)
271 struct dentry *dentry;
275 if (!capable(CAP_SYS_ADMIN))
278 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
280 return PTR_ERR(dentry);
282 /* Restrict this handle operation to symlinks only. */
283 if (!d_is_symlink(dentry)) {
288 if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
293 error = vfs_readlink(dentry, hreq->ohandle, olen);
301 * Format an attribute and copy it out to the user's buffer.
302 * Take care to check values and protect against them changing later,
303 * we may be reading them directly out of a user buffer.
306 xfs_ioc_attr_put_listent(
307 struct xfs_attr_list_context *context,
313 struct xfs_attrlist *alist = context->buffer;
314 struct xfs_attrlist_ent *aep;
317 ASSERT(!context->seen_enough);
318 ASSERT(context->count >= 0);
319 ASSERT(context->count < (ATTR_MAX_VALUELEN/8));
320 ASSERT(context->firstu >= sizeof(*alist));
321 ASSERT(context->firstu <= context->bufsize);
324 * Only list entries in the right namespace.
326 if (context->attr_filter != (flags & XFS_ATTR_NSP_ONDISK_MASK))
329 arraytop = sizeof(*alist) +
330 context->count * sizeof(alist->al_offset[0]);
332 /* decrement by the actual bytes used by the attr */
333 context->firstu -= round_up(offsetof(struct xfs_attrlist_ent, a_name) +
334 namelen + 1, sizeof(uint32_t));
335 if (context->firstu < arraytop) {
336 trace_xfs_attr_list_full(context);
338 context->seen_enough = 1;
342 aep = context->buffer + context->firstu;
343 aep->a_valuelen = valuelen;
344 memcpy(aep->a_name, name, namelen);
345 aep->a_name[namelen] = 0;
346 alist->al_offset[context->count++] = context->firstu;
347 alist->al_count = context->count;
348 trace_xfs_attr_list_add(context);
355 if (ioc_flags & XFS_IOC_ATTR_ROOT)
356 return XFS_ATTR_ROOT;
357 if (ioc_flags & XFS_IOC_ATTR_SECURE)
358 return XFS_ATTR_SECURE;
366 if (ioc_flags & XFS_IOC_ATTR_CREATE)
368 if (ioc_flags & XFS_IOC_ATTR_REPLACE)
369 return XATTR_REPLACE;
375 struct xfs_inode *dp,
379 struct xfs_attrlist_cursor __user *ucursor)
381 struct xfs_attr_list_context context = { };
382 struct xfs_attrlist *alist;
386 if (bufsize < sizeof(struct xfs_attrlist) ||
387 bufsize > XFS_XATTR_LIST_MAX)
391 * Reject flags, only allow namespaces.
393 if (flags & ~(XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
395 if (flags == (XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
399 * Validate the cursor.
401 if (copy_from_user(&context.cursor, ucursor, sizeof(context.cursor)))
403 if (context.cursor.pad1 || context.cursor.pad2)
405 if (!context.cursor.initted &&
406 (context.cursor.hashval || context.cursor.blkno ||
407 context.cursor.offset))
410 buffer = kvzalloc(bufsize, GFP_KERNEL);
415 * Initialize the output buffer.
419 context.attr_filter = xfs_attr_filter(flags);
420 context.buffer = buffer;
421 context.bufsize = round_down(bufsize, sizeof(uint32_t));
422 context.firstu = context.bufsize;
423 context.put_listent = xfs_ioc_attr_put_listent;
425 alist = context.buffer;
428 alist->al_offset[0] = context.bufsize;
430 error = xfs_attr_list(&context);
434 if (copy_to_user(ubuf, buffer, bufsize) ||
435 copy_to_user(ucursor, &context.cursor, sizeof(context.cursor)))
443 xfs_attrlist_by_handle(
444 struct file *parfilp,
445 struct xfs_fsop_attrlist_handlereq __user *p)
447 struct xfs_fsop_attrlist_handlereq al_hreq;
448 struct dentry *dentry;
451 if (!capable(CAP_SYS_ADMIN))
453 if (copy_from_user(&al_hreq, p, sizeof(al_hreq)))
456 dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
458 return PTR_ERR(dentry);
460 error = xfs_ioc_attr_list(XFS_I(d_inode(dentry)), al_hreq.buffer,
461 al_hreq.buflen, al_hreq.flags, &p->pos);
467 xfs_attrmulti_attr_get(
470 unsigned char __user *ubuf,
474 struct xfs_da_args args = {
476 .attr_filter = xfs_attr_filter(flags),
477 .attr_flags = xfs_attr_flags(flags),
479 .namelen = strlen(name),
484 if (*len > XFS_XATTR_SIZE_MAX)
487 error = xfs_attr_get(&args);
491 *len = args.valuelen;
492 if (copy_to_user(ubuf, args.value, args.valuelen))
496 kmem_free(args.value);
501 xfs_attrmulti_attr_set(
504 const unsigned char __user *ubuf,
508 struct xfs_da_args args = {
510 .attr_filter = xfs_attr_filter(flags),
511 .attr_flags = xfs_attr_flags(flags),
513 .namelen = strlen(name),
517 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
521 if (len > XFS_XATTR_SIZE_MAX)
523 args.value = memdup_user(ubuf, len);
524 if (IS_ERR(args.value))
525 return PTR_ERR(args.value);
529 error = xfs_attr_change(&args);
530 if (!error && (flags & XFS_IOC_ATTR_ROOT))
531 xfs_forget_acl(inode, name);
537 xfs_ioc_attrmulti_one(
538 struct file *parfilp,
549 if ((flags & XFS_IOC_ATTR_ROOT) && (flags & XFS_IOC_ATTR_SECURE))
552 name = strndup_user(uname, MAXNAMELEN);
554 return PTR_ERR(name);
558 error = xfs_attrmulti_attr_get(inode, name, value, len, flags);
565 error = mnt_want_write_file(parfilp);
568 error = xfs_attrmulti_attr_set(inode, name, value, *len, flags);
569 mnt_drop_write_file(parfilp);
581 xfs_attrmulti_by_handle(
582 struct file *parfilp,
586 xfs_attr_multiop_t *ops;
587 xfs_fsop_attrmulti_handlereq_t am_hreq;
588 struct dentry *dentry;
589 unsigned int i, size;
591 if (!capable(CAP_SYS_ADMIN))
593 if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
597 if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
600 dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
602 return PTR_ERR(dentry);
605 size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
606 if (!size || size > 16 * PAGE_SIZE)
609 ops = memdup_user(am_hreq.ops, size);
611 error = PTR_ERR(ops);
616 for (i = 0; i < am_hreq.opcount; i++) {
617 ops[i].am_error = xfs_ioc_attrmulti_one(parfilp,
618 d_inode(dentry), ops[i].am_opcode,
619 ops[i].am_attrname, ops[i].am_attrvalue,
620 &ops[i].am_length, ops[i].am_flags);
623 if (copy_to_user(am_hreq.ops, ops, size))
632 /* Return 0 on success or positive error */
634 xfs_fsbulkstat_one_fmt(
635 struct xfs_ibulk *breq,
636 const struct xfs_bulkstat *bstat)
638 struct xfs_bstat bs1;
640 xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
641 if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
643 return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
648 struct xfs_ibulk *breq,
649 const struct xfs_inumbers *igrp)
651 struct xfs_inogrp ig1;
653 xfs_inumbers_to_inogrp(&ig1, igrp);
654 if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
656 return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
665 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
666 struct xfs_fsop_bulkreq bulkreq;
667 struct xfs_ibulk breq = {
669 .idmap = file_mnt_idmap(file),
675 /* done = 1 if there are more stats to get and if bulkstat */
676 /* should be called again (unused here, but used in dmapi) */
678 if (!capable(CAP_SYS_ADMIN))
681 if (xfs_is_shutdown(mp))
684 if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
687 if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
690 if (bulkreq.icount <= 0)
693 if (bulkreq.ubuffer == NULL)
696 breq.ubuffer = bulkreq.ubuffer;
697 breq.icount = bulkreq.icount;
700 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
701 * that we want to stat. However, FSINUMBERS and FSBULKSTAT expect
702 * that *lastip contains either zero or the number of the last inode to
703 * be examined by the previous call and return results starting with
704 * the next inode after that. The new bulk request back end functions
705 * take the inode to start with, so we have to compute the startino
706 * parameter from lastino to maintain correct function. lastino == 0
707 * is a special case because it has traditionally meant "first inode
710 if (cmd == XFS_IOC_FSINUMBERS) {
711 breq.startino = lastino ? lastino + 1 : 0;
712 error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
713 lastino = breq.startino - 1;
714 } else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
715 breq.startino = lastino;
717 error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
718 } else { /* XFS_IOC_FSBULKSTAT */
719 breq.startino = lastino ? lastino + 1 : 0;
720 error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
721 lastino = breq.startino - 1;
727 if (bulkreq.lastip != NULL &&
728 copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
731 if (bulkreq.ocount != NULL &&
732 copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
738 /* Return 0 on success or positive error */
741 struct xfs_ibulk *breq,
742 const struct xfs_bulkstat *bstat)
744 if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
746 return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
750 * Check the incoming bulk request @hdr from userspace and initialize the
751 * internal @breq bulk request appropriately. Returns 0 if the bulk request
752 * should proceed; -ECANCELED if there's nothing to do; or the usual
753 * negative error code.
757 struct xfs_mount *mp,
758 const struct xfs_bulk_ireq *hdr,
759 struct xfs_ibulk *breq,
760 void __user *ubuffer)
762 if (hdr->icount == 0 ||
763 (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
764 memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
767 breq->startino = hdr->ino;
768 breq->ubuffer = ubuffer;
769 breq->icount = hdr->icount;
774 * The @ino parameter is a special value, so we must look it up here.
775 * We're not allowed to have IREQ_AGNO, and we only return one inode
778 if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
779 if (hdr->flags & XFS_BULK_IREQ_AGNO)
783 case XFS_BULK_IREQ_SPECIAL_ROOT:
784 breq->startino = mp->m_sb.sb_rootino;
793 * The IREQ_AGNO flag means that we only want results from a given AG.
794 * If @hdr->ino is zero, we start iterating in that AG. If @hdr->ino is
795 * beyond the specified AG then we return no results.
797 if (hdr->flags & XFS_BULK_IREQ_AGNO) {
798 if (hdr->agno >= mp->m_sb.sb_agcount)
801 if (breq->startino == 0)
802 breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
803 else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
806 breq->flags |= XFS_IBULK_SAME_AG;
808 /* Asking for an inode past the end of the AG? We're done! */
809 if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
811 } else if (hdr->agno)
814 /* Asking for an inode past the end of the FS? We're done! */
815 if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
818 if (hdr->flags & XFS_BULK_IREQ_NREXT64)
819 breq->flags |= XFS_IBULK_NREXT64;
825 * Update the userspace bulk request @hdr to reflect the end state of the
826 * internal bulk request @breq.
829 xfs_bulk_ireq_teardown(
830 struct xfs_bulk_ireq *hdr,
831 struct xfs_ibulk *breq)
833 hdr->ino = breq->startino;
834 hdr->ocount = breq->ocount;
837 /* Handle the v5 bulkstat ioctl. */
842 struct xfs_bulkstat_req __user *arg)
844 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
845 struct xfs_bulk_ireq hdr;
846 struct xfs_ibulk breq = {
848 .idmap = file_mnt_idmap(file),
852 if (!capable(CAP_SYS_ADMIN))
855 if (xfs_is_shutdown(mp))
858 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
861 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
862 if (error == -ECANCELED)
867 error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
872 xfs_bulk_ireq_teardown(&hdr, &breq);
873 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
881 struct xfs_ibulk *breq,
882 const struct xfs_inumbers *igrp)
884 if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
886 return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
889 /* Handle the v5 inumbers ioctl. */
892 struct xfs_mount *mp,
894 struct xfs_inumbers_req __user *arg)
896 struct xfs_bulk_ireq hdr;
897 struct xfs_ibulk breq = {
902 if (!capable(CAP_SYS_ADMIN))
905 if (xfs_is_shutdown(mp))
908 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
911 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
912 if (error == -ECANCELED)
917 error = xfs_inumbers(&breq, xfs_inumbers_fmt);
922 xfs_bulk_ireq_teardown(&hdr, &breq);
923 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
931 struct xfs_mount *mp,
935 struct xfs_fsop_geom fsgeo;
938 xfs_fs_geometry(mp, &fsgeo, struct_version);
940 if (struct_version <= 3)
941 len = sizeof(struct xfs_fsop_geom_v1);
942 else if (struct_version == 4)
943 len = sizeof(struct xfs_fsop_geom_v4);
945 xfs_fsop_geom_health(mp, &fsgeo);
949 if (copy_to_user(arg, &fsgeo, len))
956 struct xfs_mount *mp,
959 struct xfs_perag *pag;
960 struct xfs_ag_geometry ageo;
963 if (copy_from_user(&ageo, arg, sizeof(ageo)))
967 if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
970 pag = xfs_perag_get(mp, ageo.ag_number);
974 error = xfs_ag_get_geometry(pag, &ageo);
979 if (copy_to_user(arg, &ageo, sizeof(ageo)))
985 * Linux extended inode flags interface.
990 struct xfs_inode *ip,
994 struct xfs_mount *mp = ip->i_mount;
995 struct xfs_ifork *ifp = xfs_ifork_ptr(ip, whichfork);
997 fileattr_fill_xflags(fa, xfs_ip2xflags(ip));
999 if (ip->i_diflags & XFS_DIFLAG_EXTSIZE) {
1000 fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1001 } else if (ip->i_diflags & XFS_DIFLAG_EXTSZINHERIT) {
1003 * Don't let a misaligned extent size hint on a directory
1004 * escape to userspace if it won't pass the setattr checks
1007 if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
1008 xfs_extlen_to_rtxmod(mp, ip->i_extsize) > 0) {
1009 fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE |
1010 FS_XFLAG_EXTSZINHERIT);
1011 fa->fsx_extsize = 0;
1013 fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1017 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1018 fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
1019 fa->fsx_projid = ip->i_projid;
1020 if (ifp && !xfs_need_iread_extents(ifp))
1021 fa->fsx_nextents = xfs_iext_count(ifp);
1023 fa->fsx_nextents = xfs_ifork_nextents(ifp);
1027 xfs_ioc_fsgetxattra(
1033 xfs_ilock(ip, XFS_ILOCK_SHARED);
1034 xfs_fill_fsxattr(ip, XFS_ATTR_FORK, &fa);
1035 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1037 return copy_fsxattr_to_user(&fa, arg);
1042 struct dentry *dentry,
1043 struct fileattr *fa)
1045 struct xfs_inode *ip = XFS_I(d_inode(dentry));
1047 if (d_is_special(dentry))
1050 xfs_ilock(ip, XFS_ILOCK_SHARED);
1051 xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
1052 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1059 struct xfs_inode *ip,
1060 unsigned int xflags)
1062 /* can't set PREALLOC this way, just preserve it */
1064 (ip->i_diflags & XFS_DIFLAG_PREALLOC);
1066 if (xflags & FS_XFLAG_IMMUTABLE)
1067 di_flags |= XFS_DIFLAG_IMMUTABLE;
1068 if (xflags & FS_XFLAG_APPEND)
1069 di_flags |= XFS_DIFLAG_APPEND;
1070 if (xflags & FS_XFLAG_SYNC)
1071 di_flags |= XFS_DIFLAG_SYNC;
1072 if (xflags & FS_XFLAG_NOATIME)
1073 di_flags |= XFS_DIFLAG_NOATIME;
1074 if (xflags & FS_XFLAG_NODUMP)
1075 di_flags |= XFS_DIFLAG_NODUMP;
1076 if (xflags & FS_XFLAG_NODEFRAG)
1077 di_flags |= XFS_DIFLAG_NODEFRAG;
1078 if (xflags & FS_XFLAG_FILESTREAM)
1079 di_flags |= XFS_DIFLAG_FILESTREAM;
1080 if (S_ISDIR(VFS_I(ip)->i_mode)) {
1081 if (xflags & FS_XFLAG_RTINHERIT)
1082 di_flags |= XFS_DIFLAG_RTINHERIT;
1083 if (xflags & FS_XFLAG_NOSYMLINKS)
1084 di_flags |= XFS_DIFLAG_NOSYMLINKS;
1085 if (xflags & FS_XFLAG_EXTSZINHERIT)
1086 di_flags |= XFS_DIFLAG_EXTSZINHERIT;
1087 if (xflags & FS_XFLAG_PROJINHERIT)
1088 di_flags |= XFS_DIFLAG_PROJINHERIT;
1089 } else if (S_ISREG(VFS_I(ip)->i_mode)) {
1090 if (xflags & FS_XFLAG_REALTIME)
1091 di_flags |= XFS_DIFLAG_REALTIME;
1092 if (xflags & FS_XFLAG_EXTSIZE)
1093 di_flags |= XFS_DIFLAG_EXTSIZE;
1101 struct xfs_inode *ip,
1102 unsigned int xflags)
1104 uint64_t di_flags2 =
1105 (ip->i_diflags2 & (XFS_DIFLAG2_REFLINK |
1106 XFS_DIFLAG2_BIGTIME |
1107 XFS_DIFLAG2_NREXT64));
1109 if (xflags & FS_XFLAG_DAX)
1110 di_flags2 |= XFS_DIFLAG2_DAX;
1111 if (xflags & FS_XFLAG_COWEXTSIZE)
1112 di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
1118 xfs_ioctl_setattr_xflags(
1119 struct xfs_trans *tp,
1120 struct xfs_inode *ip,
1121 struct fileattr *fa)
1123 struct xfs_mount *mp = ip->i_mount;
1124 bool rtflag = (fa->fsx_xflags & FS_XFLAG_REALTIME);
1127 if (rtflag != XFS_IS_REALTIME_INODE(ip)) {
1128 /* Can't change realtime flag if any extents are allocated. */
1129 if (ip->i_df.if_nextents || ip->i_delayed_blks)
1134 /* If realtime flag is set then must have realtime device */
1135 if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1136 xfs_extlen_to_rtxmod(mp, ip->i_extsize))
1139 /* Clear reflink if we are actually able to set the rt flag. */
1140 if (xfs_is_reflink_inode(ip))
1141 ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
1144 /* diflags2 only valid for v3 inodes. */
1145 i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1146 if (i_flags2 && !xfs_has_v3inodes(mp))
1149 ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1150 ip->i_diflags2 = i_flags2;
1152 xfs_diflags_to_iflags(ip, false);
1155 * Make the stable writes flag match that of the device the inode
1156 * resides on when flipping the RT flag.
1158 if (rtflag != XFS_IS_REALTIME_INODE(ip) && S_ISREG(VFS_I(ip)->i_mode))
1159 xfs_update_stable_writes(ip);
1161 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1162 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1163 XFS_STATS_INC(mp, xs_ig_attrchg);
1168 xfs_ioctl_setattr_prepare_dax(
1169 struct xfs_inode *ip,
1170 struct fileattr *fa)
1172 struct xfs_mount *mp = ip->i_mount;
1173 struct inode *inode = VFS_I(ip);
1175 if (S_ISDIR(inode->i_mode))
1178 if (xfs_has_dax_always(mp) || xfs_has_dax_never(mp))
1181 if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
1182 !(ip->i_diflags2 & XFS_DIFLAG2_DAX)) ||
1183 (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
1184 (ip->i_diflags2 & XFS_DIFLAG2_DAX)))
1185 d_mark_dontcache(inode);
1189 * Set up the transaction structure for the setattr operation, checking that we
1190 * have permission to do so. On success, return a clean transaction and the
1191 * inode locked exclusively ready for further operation specific checks. On
1192 * failure, return an error without modifying or locking the inode.
1194 static struct xfs_trans *
1195 xfs_ioctl_setattr_get_trans(
1196 struct xfs_inode *ip,
1197 struct xfs_dquot *pdqp)
1199 struct xfs_mount *mp = ip->i_mount;
1200 struct xfs_trans *tp;
1203 if (xfs_is_readonly(mp))
1206 if (xfs_is_shutdown(mp))
1209 error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
1210 has_capability_noaudit(current, CAP_FOWNER), &tp);
1214 if (xfs_has_wsync(mp))
1215 xfs_trans_set_sync(tp);
1220 return ERR_PTR(error);
1224 * Validate a proposed extent size hint. For regular files, the hint can only
1225 * be changed if no extents are allocated.
1228 xfs_ioctl_setattr_check_extsize(
1229 struct xfs_inode *ip,
1230 struct fileattr *fa)
1232 struct xfs_mount *mp = ip->i_mount;
1233 xfs_failaddr_t failaddr;
1234 uint16_t new_diflags;
1239 if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
1240 XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
1243 if (fa->fsx_extsize & mp->m_blockmask)
1246 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1249 * Inode verifiers do not check that the extent size hint is an integer
1250 * multiple of the rt extent size on a directory with both rtinherit
1251 * and extszinherit flags set. Don't let sysadmins misconfigure
1254 if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
1255 (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
1256 unsigned int rtextsize_bytes;
1258 rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
1259 if (fa->fsx_extsize % rtextsize_bytes)
1263 failaddr = xfs_inode_validate_extsize(ip->i_mount,
1264 XFS_B_TO_FSB(mp, fa->fsx_extsize),
1265 VFS_I(ip)->i_mode, new_diflags);
1266 return failaddr != NULL ? -EINVAL : 0;
1270 xfs_ioctl_setattr_check_cowextsize(
1271 struct xfs_inode *ip,
1272 struct fileattr *fa)
1274 struct xfs_mount *mp = ip->i_mount;
1275 xfs_failaddr_t failaddr;
1276 uint64_t new_diflags2;
1277 uint16_t new_diflags;
1282 if (fa->fsx_cowextsize & mp->m_blockmask)
1285 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1286 new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1288 failaddr = xfs_inode_validate_cowextsize(ip->i_mount,
1289 XFS_B_TO_FSB(mp, fa->fsx_cowextsize),
1290 VFS_I(ip)->i_mode, new_diflags, new_diflags2);
1291 return failaddr != NULL ? -EINVAL : 0;
1295 xfs_ioctl_setattr_check_projid(
1296 struct xfs_inode *ip,
1297 struct fileattr *fa)
1302 /* Disallow 32bit project ids if 32bit IDs are not enabled. */
1303 if (fa->fsx_projid > (uint16_t)-1 &&
1304 !xfs_has_projid32(ip->i_mount))
1311 struct mnt_idmap *idmap,
1312 struct dentry *dentry,
1313 struct fileattr *fa)
1315 struct xfs_inode *ip = XFS_I(d_inode(dentry));
1316 struct xfs_mount *mp = ip->i_mount;
1317 struct xfs_trans *tp;
1318 struct xfs_dquot *pdqp = NULL;
1319 struct xfs_dquot *olddquot = NULL;
1322 trace_xfs_ioctl_setattr(ip);
1324 if (d_is_special(dentry))
1327 if (!fa->fsx_valid) {
1328 if (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL |
1329 FS_NOATIME_FL | FS_NODUMP_FL |
1330 FS_SYNC_FL | FS_DAX_FL | FS_PROJINHERIT_FL))
1334 error = xfs_ioctl_setattr_check_projid(ip, fa);
1339 * If disk quotas is on, we make sure that the dquots do exist on disk,
1340 * before we start any other transactions. Trying to do this later
1341 * is messy. We don't care to take a readlock to look at the ids
1342 * in inode here, because we can't hold it across the trans_reserve.
1343 * If the IDs do change before we take the ilock, we're covered
1344 * because the i_*dquot fields will get updated anyway.
1346 if (fa->fsx_valid && XFS_IS_QUOTA_ON(mp)) {
1347 error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
1348 VFS_I(ip)->i_gid, fa->fsx_projid,
1349 XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
1354 xfs_ioctl_setattr_prepare_dax(ip, fa);
1356 tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
1358 error = PTR_ERR(tp);
1359 goto error_free_dquots;
1362 error = xfs_ioctl_setattr_check_extsize(ip, fa);
1364 goto error_trans_cancel;
1366 error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1368 goto error_trans_cancel;
1370 error = xfs_ioctl_setattr_xflags(tp, ip, fa);
1372 goto error_trans_cancel;
1377 * Change file ownership. Must be the owner or privileged. CAP_FSETID
1378 * overrides the following restrictions:
1380 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1381 * successful return from chown()
1384 if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1385 !capable_wrt_inode_uidgid(idmap, VFS_I(ip), CAP_FSETID))
1386 VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1388 /* Change the ownerships and register project quota modifications */
1389 if (ip->i_projid != fa->fsx_projid) {
1390 if (XFS_IS_PQUOTA_ON(mp)) {
1391 olddquot = xfs_qm_vop_chown(tp, ip,
1392 &ip->i_pdquot, pdqp);
1394 ip->i_projid = fa->fsx_projid;
1398 * Only set the extent size hint if we've already determined that the
1399 * extent size hint should be set on the inode. If no extent size flags
1400 * are set on the inode then unconditionally clear the extent size hint.
1402 if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1403 ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1407 if (xfs_has_v3inodes(mp)) {
1408 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1409 ip->i_cowextsize = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1411 ip->i_cowextsize = 0;
1415 error = xfs_trans_commit(tp);
1418 * Release any dquot(s) the inode had kept before chown.
1420 xfs_qm_dqrele(olddquot);
1421 xfs_qm_dqrele(pdqp);
1426 xfs_trans_cancel(tp);
1428 xfs_qm_dqrele(pdqp);
1435 struct getbmapx __user *u,
1438 if (put_user(p->bmv_offset, &u->bmv_offset) ||
1439 put_user(p->bmv_block, &u->bmv_block) ||
1440 put_user(p->bmv_length, &u->bmv_length) ||
1441 put_user(0, &u->bmv_count) ||
1442 put_user(0, &u->bmv_entries))
1444 if (recsize < sizeof(struct getbmapx))
1446 if (put_user(0, &u->bmv_iflags) ||
1447 put_user(p->bmv_oflags, &u->bmv_oflags) ||
1448 put_user(0, &u->bmv_unused1) ||
1449 put_user(0, &u->bmv_unused2))
1460 struct getbmapx bmx = { 0 };
1461 struct kgetbmap *buf;
1466 case XFS_IOC_GETBMAPA:
1467 bmx.bmv_iflags = BMV_IF_ATTRFORK;
1469 case XFS_IOC_GETBMAP:
1470 /* struct getbmap is a strict subset of struct getbmapx. */
1471 recsize = sizeof(struct getbmap);
1473 case XFS_IOC_GETBMAPX:
1474 recsize = sizeof(struct getbmapx);
1480 if (copy_from_user(&bmx, arg, recsize))
1483 if (bmx.bmv_count < 2)
1485 if (bmx.bmv_count >= INT_MAX / recsize)
1488 buf = kvcalloc(bmx.bmv_count, sizeof(*buf), GFP_KERNEL);
1492 error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1497 if (copy_to_user(arg, &bmx, recsize))
1501 for (i = 0; i < bmx.bmv_entries; i++) {
1502 if (!xfs_getbmap_format(buf + i, arg, recsize))
1515 struct xfs_inode *ip,
1516 struct fsmap_head __user *arg)
1518 struct xfs_fsmap_head xhead = {0};
1519 struct fsmap_head head;
1522 __u32 last_flags = 0;
1526 if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1528 if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1529 memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1530 sizeof(head.fmh_keys[0].fmr_reserved)) ||
1531 memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1532 sizeof(head.fmh_keys[1].fmr_reserved)))
1536 * Use an internal memory buffer so that we don't have to copy fsmap
1537 * data to userspace while holding locks. Start by trying to allocate
1538 * up to 128k for the buffer, but fall back to a single page if needed.
1540 count = min_t(unsigned int, head.fmh_count,
1541 131072 / sizeof(struct fsmap));
1542 recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
1544 count = min_t(unsigned int, head.fmh_count,
1545 PAGE_SIZE / sizeof(struct fsmap));
1546 recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
1551 xhead.fmh_iflags = head.fmh_iflags;
1552 xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1553 xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1555 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1556 trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1558 head.fmh_entries = 0;
1560 struct fsmap __user *user_recs;
1561 struct fsmap *last_rec;
1563 user_recs = &arg->fmh_recs[head.fmh_entries];
1564 xhead.fmh_entries = 0;
1565 xhead.fmh_count = min_t(unsigned int, count,
1566 head.fmh_count - head.fmh_entries);
1568 /* Run query, record how many entries we got. */
1569 error = xfs_getfsmap(ip->i_mount, &xhead, recs);
1573 * There are no more records in the result set. Copy
1574 * whatever we got to userspace and break out.
1580 * The internal memory buffer is full. Copy whatever
1581 * records we got to userspace and go again if we have
1582 * not yet filled the userspace buffer.
1589 head.fmh_entries += xhead.fmh_entries;
1590 head.fmh_oflags = xhead.fmh_oflags;
1593 * If the caller wanted a record count or there aren't any
1594 * new records to return, we're done.
1596 if (head.fmh_count == 0 || xhead.fmh_entries == 0)
1599 /* Copy all the records we got out to userspace. */
1600 if (copy_to_user(user_recs, recs,
1601 xhead.fmh_entries * sizeof(struct fsmap))) {
1606 /* Remember the last record flags we copied to userspace. */
1607 last_rec = &recs[xhead.fmh_entries - 1];
1608 last_flags = last_rec->fmr_flags;
1610 /* Set up the low key for the next iteration. */
1611 xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
1612 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1613 } while (!done && head.fmh_entries < head.fmh_count);
1616 * If there are no more records in the query result set and we're not
1617 * in counting mode, mark the last record returned with the LAST flag.
1619 if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
1620 struct fsmap __user *user_rec;
1622 last_flags |= FMR_OF_LAST;
1623 user_rec = &arg->fmh_recs[head.fmh_entries - 1];
1625 if (copy_to_user(&user_rec->fmr_flags, &last_flags,
1626 sizeof(last_flags))) {
1632 /* copy back header */
1633 if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1644 xfs_ioc_scrub_metadata(
1648 struct xfs_scrub_metadata scrub;
1651 if (!capable(CAP_SYS_ADMIN))
1654 if (copy_from_user(&scrub, arg, sizeof(scrub)))
1657 error = xfs_scrub_metadata(file, &scrub);
1661 if (copy_to_user(arg, &scrub, sizeof(scrub)))
1671 xfs_inode_t *ip, *tip;
1675 /* Pull information for the target fd */
1676 f = fdget((int)sxp->sx_fdtarget);
1682 if (!(f.file->f_mode & FMODE_WRITE) ||
1683 !(f.file->f_mode & FMODE_READ) ||
1684 (f.file->f_flags & O_APPEND)) {
1689 tmp = fdget((int)sxp->sx_fdtmp);
1695 if (!(tmp.file->f_mode & FMODE_WRITE) ||
1696 !(tmp.file->f_mode & FMODE_READ) ||
1697 (tmp.file->f_flags & O_APPEND)) {
1699 goto out_put_tmp_file;
1702 if (IS_SWAPFILE(file_inode(f.file)) ||
1703 IS_SWAPFILE(file_inode(tmp.file))) {
1705 goto out_put_tmp_file;
1709 * We need to ensure that the fds passed in point to XFS inodes
1710 * before we cast and access them as XFS structures as we have no
1711 * control over what the user passes us here.
1713 if (f.file->f_op != &xfs_file_operations ||
1714 tmp.file->f_op != &xfs_file_operations) {
1716 goto out_put_tmp_file;
1719 ip = XFS_I(file_inode(f.file));
1720 tip = XFS_I(file_inode(tmp.file));
1722 if (ip->i_mount != tip->i_mount) {
1724 goto out_put_tmp_file;
1727 if (ip->i_ino == tip->i_ino) {
1729 goto out_put_tmp_file;
1732 if (xfs_is_shutdown(ip->i_mount)) {
1734 goto out_put_tmp_file;
1737 error = xfs_swap_extents(ip, tip, sxp);
1749 struct xfs_mount *mp,
1750 char __user *user_label)
1752 struct xfs_sb *sbp = &mp->m_sb;
1753 char label[XFSLABEL_MAX + 1];
1756 BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1758 /* 1 larger than sb_fname, so this ensures a trailing NUL char */
1759 memset(label, 0, sizeof(label));
1760 spin_lock(&mp->m_sb_lock);
1761 strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1762 spin_unlock(&mp->m_sb_lock);
1764 if (copy_to_user(user_label, label, sizeof(label)))
1772 struct xfs_mount *mp,
1773 char __user *newlabel)
1775 struct xfs_sb *sbp = &mp->m_sb;
1776 char label[XFSLABEL_MAX + 1];
1780 if (!capable(CAP_SYS_ADMIN))
1783 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1784 * smaller, at 12 bytes. We copy one more to be sure we find the
1785 * (required) NULL character to test the incoming label length.
1786 * NB: The on disk label doesn't need to be null terminated.
1788 if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1790 len = strnlen(label, XFSLABEL_MAX + 1);
1791 if (len > sizeof(sbp->sb_fname))
1794 error = mnt_want_write_file(filp);
1798 spin_lock(&mp->m_sb_lock);
1799 memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1800 memcpy(sbp->sb_fname, label, len);
1801 spin_unlock(&mp->m_sb_lock);
1804 * Now we do several things to satisfy userspace.
1805 * In addition to normal logging of the primary superblock, we also
1806 * immediately write these changes to sector zero for the primary, then
1807 * update all backup supers (as xfs_db does for a label change), then
1808 * invalidate the block device page cache. This is so that any prior
1809 * buffered reads from userspace (i.e. from blkid) are invalidated,
1810 * and userspace will see the newly-written label.
1812 error = xfs_sync_sb_buf(mp);
1816 * growfs also updates backup supers so lock against that.
1818 mutex_lock(&mp->m_growlock);
1819 error = xfs_update_secondary_sbs(mp);
1820 mutex_unlock(&mp->m_growlock);
1822 invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1825 mnt_drop_write_file(filp);
1830 xfs_fs_eofblocks_from_user(
1831 struct xfs_fs_eofblocks *src,
1832 struct xfs_icwalk *dst)
1834 if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1837 if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1840 if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1841 memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1845 if (src->eof_flags & XFS_EOF_FLAGS_SYNC)
1846 dst->icw_flags |= XFS_ICWALK_FLAG_SYNC;
1847 if (src->eof_flags & XFS_EOF_FLAGS_UID)
1848 dst->icw_flags |= XFS_ICWALK_FLAG_UID;
1849 if (src->eof_flags & XFS_EOF_FLAGS_GID)
1850 dst->icw_flags |= XFS_ICWALK_FLAG_GID;
1851 if (src->eof_flags & XFS_EOF_FLAGS_PRID)
1852 dst->icw_flags |= XFS_ICWALK_FLAG_PRID;
1853 if (src->eof_flags & XFS_EOF_FLAGS_MINFILESIZE)
1854 dst->icw_flags |= XFS_ICWALK_FLAG_MINFILESIZE;
1856 dst->icw_prid = src->eof_prid;
1857 dst->icw_min_file_size = src->eof_min_file_size;
1859 dst->icw_uid = INVALID_UID;
1860 if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1861 dst->icw_uid = make_kuid(current_user_ns(), src->eof_uid);
1862 if (!uid_valid(dst->icw_uid))
1866 dst->icw_gid = INVALID_GID;
1867 if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1868 dst->icw_gid = make_kgid(current_user_ns(), src->eof_gid);
1869 if (!gid_valid(dst->icw_gid))
1876 * These long-unused ioctls were removed from the official ioctl API in 5.17,
1877 * but retain these definitions so that we can log warnings about them.
1879 #define XFS_IOC_ALLOCSP _IOW ('X', 10, struct xfs_flock64)
1880 #define XFS_IOC_FREESP _IOW ('X', 11, struct xfs_flock64)
1881 #define XFS_IOC_ALLOCSP64 _IOW ('X', 36, struct xfs_flock64)
1882 #define XFS_IOC_FREESP64 _IOW ('X', 37, struct xfs_flock64)
1885 * Note: some of the ioctl's return positive numbers as a
1886 * byte count indicating success, such as readlink_by_handle.
1887 * So we don't "sign flip" like most other routines. This means
1888 * true errors need to be returned as a negative value.
1896 struct inode *inode = file_inode(filp);
1897 struct xfs_inode *ip = XFS_I(inode);
1898 struct xfs_mount *mp = ip->i_mount;
1899 void __user *arg = (void __user *)p;
1902 trace_xfs_file_ioctl(ip);
1906 return xfs_ioc_trim(mp, arg);
1907 case FS_IOC_GETFSLABEL:
1908 return xfs_ioc_getlabel(mp, arg);
1909 case FS_IOC_SETFSLABEL:
1910 return xfs_ioc_setlabel(filp, mp, arg);
1911 case XFS_IOC_ALLOCSP:
1912 case XFS_IOC_FREESP:
1913 case XFS_IOC_ALLOCSP64:
1914 case XFS_IOC_FREESP64:
1916 "%s should use fallocate; XFS_IOC_{ALLOC,FREE}SP ioctl unsupported",
1919 case XFS_IOC_DIOINFO: {
1920 struct xfs_buftarg *target = xfs_inode_buftarg(ip);
1923 da.d_mem = da.d_miniosz = target->bt_logical_sectorsize;
1924 da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1926 if (copy_to_user(arg, &da, sizeof(da)))
1931 case XFS_IOC_FSBULKSTAT_SINGLE:
1932 case XFS_IOC_FSBULKSTAT:
1933 case XFS_IOC_FSINUMBERS:
1934 return xfs_ioc_fsbulkstat(filp, cmd, arg);
1936 case XFS_IOC_BULKSTAT:
1937 return xfs_ioc_bulkstat(filp, cmd, arg);
1938 case XFS_IOC_INUMBERS:
1939 return xfs_ioc_inumbers(mp, cmd, arg);
1941 case XFS_IOC_FSGEOMETRY_V1:
1942 return xfs_ioc_fsgeometry(mp, arg, 3);
1943 case XFS_IOC_FSGEOMETRY_V4:
1944 return xfs_ioc_fsgeometry(mp, arg, 4);
1945 case XFS_IOC_FSGEOMETRY:
1946 return xfs_ioc_fsgeometry(mp, arg, 5);
1948 case XFS_IOC_AG_GEOMETRY:
1949 return xfs_ioc_ag_geometry(mp, arg);
1951 case XFS_IOC_GETVERSION:
1952 return put_user(inode->i_generation, (int __user *)arg);
1954 case XFS_IOC_FSGETXATTRA:
1955 return xfs_ioc_fsgetxattra(ip, arg);
1957 case XFS_IOC_GETBMAP:
1958 case XFS_IOC_GETBMAPA:
1959 case XFS_IOC_GETBMAPX:
1960 return xfs_ioc_getbmap(filp, cmd, arg);
1962 case FS_IOC_GETFSMAP:
1963 return xfs_ioc_getfsmap(ip, arg);
1965 case XFS_IOC_SCRUB_METADATA:
1966 return xfs_ioc_scrub_metadata(filp, arg);
1968 case XFS_IOC_FD_TO_HANDLE:
1969 case XFS_IOC_PATH_TO_HANDLE:
1970 case XFS_IOC_PATH_TO_FSHANDLE: {
1971 xfs_fsop_handlereq_t hreq;
1973 if (copy_from_user(&hreq, arg, sizeof(hreq)))
1975 return xfs_find_handle(cmd, &hreq);
1977 case XFS_IOC_OPEN_BY_HANDLE: {
1978 xfs_fsop_handlereq_t hreq;
1980 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1982 return xfs_open_by_handle(filp, &hreq);
1985 case XFS_IOC_READLINK_BY_HANDLE: {
1986 xfs_fsop_handlereq_t hreq;
1988 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1990 return xfs_readlink_by_handle(filp, &hreq);
1992 case XFS_IOC_ATTRLIST_BY_HANDLE:
1993 return xfs_attrlist_by_handle(filp, arg);
1995 case XFS_IOC_ATTRMULTI_BY_HANDLE:
1996 return xfs_attrmulti_by_handle(filp, arg);
1998 case XFS_IOC_SWAPEXT: {
1999 struct xfs_swapext sxp;
2001 if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
2003 error = mnt_want_write_file(filp);
2006 error = xfs_ioc_swapext(&sxp);
2007 mnt_drop_write_file(filp);
2011 case XFS_IOC_FSCOUNTS: {
2012 xfs_fsop_counts_t out;
2014 xfs_fs_counts(mp, &out);
2016 if (copy_to_user(arg, &out, sizeof(out)))
2021 case XFS_IOC_SET_RESBLKS: {
2022 xfs_fsop_resblks_t inout;
2025 if (!capable(CAP_SYS_ADMIN))
2028 if (xfs_is_readonly(mp))
2031 if (copy_from_user(&inout, arg, sizeof(inout)))
2034 error = mnt_want_write_file(filp);
2038 /* input parameter is passed in resblks field of structure */
2040 error = xfs_reserve_blocks(mp, &in, &inout);
2041 mnt_drop_write_file(filp);
2045 if (copy_to_user(arg, &inout, sizeof(inout)))
2050 case XFS_IOC_GET_RESBLKS: {
2051 xfs_fsop_resblks_t out;
2053 if (!capable(CAP_SYS_ADMIN))
2056 error = xfs_reserve_blocks(mp, NULL, &out);
2060 if (copy_to_user(arg, &out, sizeof(out)))
2066 case XFS_IOC_FSGROWFSDATA: {
2067 struct xfs_growfs_data in;
2069 if (copy_from_user(&in, arg, sizeof(in)))
2072 error = mnt_want_write_file(filp);
2075 error = xfs_growfs_data(mp, &in);
2076 mnt_drop_write_file(filp);
2080 case XFS_IOC_FSGROWFSLOG: {
2081 struct xfs_growfs_log in;
2083 if (copy_from_user(&in, arg, sizeof(in)))
2086 error = mnt_want_write_file(filp);
2089 error = xfs_growfs_log(mp, &in);
2090 mnt_drop_write_file(filp);
2094 case XFS_IOC_FSGROWFSRT: {
2097 if (copy_from_user(&in, arg, sizeof(in)))
2100 error = mnt_want_write_file(filp);
2103 error = xfs_growfs_rt(mp, &in);
2104 mnt_drop_write_file(filp);
2108 case XFS_IOC_GOINGDOWN: {
2111 if (!capable(CAP_SYS_ADMIN))
2114 if (get_user(in, (uint32_t __user *)arg))
2117 return xfs_fs_goingdown(mp, in);
2120 case XFS_IOC_ERROR_INJECTION: {
2121 xfs_error_injection_t in;
2123 if (!capable(CAP_SYS_ADMIN))
2126 if (copy_from_user(&in, arg, sizeof(in)))
2129 return xfs_errortag_add(mp, in.errtag);
2132 case XFS_IOC_ERROR_CLEARALL:
2133 if (!capable(CAP_SYS_ADMIN))
2136 return xfs_errortag_clearall(mp);
2138 case XFS_IOC_FREE_EOFBLOCKS: {
2139 struct xfs_fs_eofblocks eofb;
2140 struct xfs_icwalk icw;
2142 if (!capable(CAP_SYS_ADMIN))
2145 if (xfs_is_readonly(mp))
2148 if (copy_from_user(&eofb, arg, sizeof(eofb)))
2151 error = xfs_fs_eofblocks_from_user(&eofb, &icw);
2155 trace_xfs_ioc_free_eofblocks(mp, &icw, _RET_IP_);
2157 sb_start_write(mp->m_super);
2158 error = xfs_blockgc_free_space(mp, &icw);
2159 sb_end_write(mp->m_super);