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"
42 #include <linux/mount.h>
43 #include <linux/namei.h>
44 #include <linux/fileattr.h>
47 * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
48 * a file or fs handle.
50 * XFS_IOC_PATH_TO_FSHANDLE
51 * returns fs handle for a mount point or path within that mount point
52 * XFS_IOC_FD_TO_HANDLE
53 * returns full handle for a FD opened in user space
54 * XFS_IOC_PATH_TO_HANDLE
55 * returns full handle for a path
60 xfs_fsop_handlereq_t *hreq)
70 if (cmd == XFS_IOC_FD_TO_HANDLE) {
74 inode = file_inode(f.file);
76 error = user_path_at(AT_FDCWD, hreq->path, 0, &path);
79 inode = d_inode(path.dentry);
84 * We can only generate handles for inodes residing on a XFS filesystem,
85 * and only for regular files, directories or symbolic links.
88 if (inode->i_sb->s_magic != XFS_SB_MAGIC)
92 if (!S_ISREG(inode->i_mode) &&
93 !S_ISDIR(inode->i_mode) &&
94 !S_ISLNK(inode->i_mode))
98 memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
100 if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
102 * This handle only contains an fsid, zero the rest.
104 memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
105 hsize = sizeof(xfs_fsid_t);
107 handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
108 sizeof(handle.ha_fid.fid_len);
109 handle.ha_fid.fid_pad = 0;
110 handle.ha_fid.fid_gen = inode->i_generation;
111 handle.ha_fid.fid_ino = ip->i_ino;
112 hsize = sizeof(xfs_handle_t);
116 if (copy_to_user(hreq->ohandle, &handle, hsize) ||
117 copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
123 if (cmd == XFS_IOC_FD_TO_HANDLE)
131 * No need to do permission checks on the various pathname components
132 * as the handle operations are privileged.
135 xfs_handle_acceptable(
137 struct dentry *dentry)
143 * Convert userspace handle data into a dentry.
146 xfs_handle_to_dentry(
147 struct file *parfilp,
148 void __user *uhandle,
152 struct xfs_fid64 fid;
155 * Only allow handle opens under a directory.
157 if (!S_ISDIR(file_inode(parfilp)->i_mode))
158 return ERR_PTR(-ENOTDIR);
160 if (hlen != sizeof(xfs_handle_t))
161 return ERR_PTR(-EINVAL);
162 if (copy_from_user(&handle, uhandle, hlen))
163 return ERR_PTR(-EFAULT);
164 if (handle.ha_fid.fid_len !=
165 sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
166 return ERR_PTR(-EINVAL);
168 memset(&fid, 0, sizeof(struct fid));
169 fid.ino = handle.ha_fid.fid_ino;
170 fid.gen = handle.ha_fid.fid_gen;
172 return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
173 FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
174 xfs_handle_acceptable, NULL);
177 STATIC struct dentry *
178 xfs_handlereq_to_dentry(
179 struct file *parfilp,
180 xfs_fsop_handlereq_t *hreq)
182 return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
187 struct file *parfilp,
188 xfs_fsop_handlereq_t *hreq)
190 const struct cred *cred = current_cred();
196 struct dentry *dentry;
200 if (!capable(CAP_SYS_ADMIN))
203 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
205 return PTR_ERR(dentry);
206 inode = d_inode(dentry);
208 /* Restrict xfs_open_by_handle to directories & regular files. */
209 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
214 #if BITS_PER_LONG != 32
215 hreq->oflags |= O_LARGEFILE;
218 permflag = hreq->oflags;
219 fmode = OPEN_FMODE(permflag);
220 if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
221 (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
226 if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
231 /* Can't write directories. */
232 if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
237 fd = get_unused_fd_flags(0);
243 path.mnt = parfilp->f_path.mnt;
244 path.dentry = dentry;
245 filp = dentry_open(&path, hreq->oflags, cred);
249 return PTR_ERR(filp);
252 if (S_ISREG(inode->i_mode)) {
253 filp->f_flags |= O_NOATIME;
254 filp->f_mode |= FMODE_NOCMTIME;
257 fd_install(fd, filp);
266 xfs_readlink_by_handle(
267 struct file *parfilp,
268 xfs_fsop_handlereq_t *hreq)
270 struct dentry *dentry;
274 if (!capable(CAP_SYS_ADMIN))
277 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
279 return PTR_ERR(dentry);
281 /* Restrict this handle operation to symlinks only. */
282 if (!d_is_symlink(dentry)) {
287 if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
292 error = vfs_readlink(dentry, hreq->ohandle, olen);
300 * Format an attribute and copy it out to the user's buffer.
301 * Take care to check values and protect against them changing later,
302 * we may be reading them directly out of a user buffer.
305 xfs_ioc_attr_put_listent(
306 struct xfs_attr_list_context *context,
312 struct xfs_attrlist *alist = context->buffer;
313 struct xfs_attrlist_ent *aep;
316 ASSERT(!context->seen_enough);
317 ASSERT(context->count >= 0);
318 ASSERT(context->count < (ATTR_MAX_VALUELEN/8));
319 ASSERT(context->firstu >= sizeof(*alist));
320 ASSERT(context->firstu <= context->bufsize);
323 * Only list entries in the right namespace.
325 if (context->attr_filter != (flags & XFS_ATTR_NSP_ONDISK_MASK))
328 arraytop = sizeof(*alist) +
329 context->count * sizeof(alist->al_offset[0]);
331 /* decrement by the actual bytes used by the attr */
332 context->firstu -= round_up(offsetof(struct xfs_attrlist_ent, a_name) +
333 namelen + 1, sizeof(uint32_t));
334 if (context->firstu < arraytop) {
335 trace_xfs_attr_list_full(context);
337 context->seen_enough = 1;
341 aep = context->buffer + context->firstu;
342 aep->a_valuelen = valuelen;
343 memcpy(aep->a_name, name, namelen);
344 aep->a_name[namelen] = 0;
345 alist->al_offset[context->count++] = context->firstu;
346 alist->al_count = context->count;
347 trace_xfs_attr_list_add(context);
354 if (ioc_flags & XFS_IOC_ATTR_ROOT)
355 return XFS_ATTR_ROOT;
356 if (ioc_flags & XFS_IOC_ATTR_SECURE)
357 return XFS_ATTR_SECURE;
365 if (ioc_flags & XFS_IOC_ATTR_CREATE)
367 if (ioc_flags & XFS_IOC_ATTR_REPLACE)
368 return XATTR_REPLACE;
374 struct xfs_inode *dp,
378 struct xfs_attrlist_cursor __user *ucursor)
380 struct xfs_attr_list_context context = { };
381 struct xfs_attrlist *alist;
385 if (bufsize < sizeof(struct xfs_attrlist) ||
386 bufsize > XFS_XATTR_LIST_MAX)
390 * Reject flags, only allow namespaces.
392 if (flags & ~(XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
394 if (flags == (XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
398 * Validate the cursor.
400 if (copy_from_user(&context.cursor, ucursor, sizeof(context.cursor)))
402 if (context.cursor.pad1 || context.cursor.pad2)
404 if (!context.cursor.initted &&
405 (context.cursor.hashval || context.cursor.blkno ||
406 context.cursor.offset))
409 buffer = kvzalloc(bufsize, GFP_KERNEL);
414 * Initialize the output buffer.
418 context.attr_filter = xfs_attr_filter(flags);
419 context.buffer = buffer;
420 context.bufsize = round_down(bufsize, sizeof(uint32_t));
421 context.firstu = context.bufsize;
422 context.put_listent = xfs_ioc_attr_put_listent;
424 alist = context.buffer;
427 alist->al_offset[0] = context.bufsize;
429 error = xfs_attr_list(&context);
433 if (copy_to_user(ubuf, buffer, bufsize) ||
434 copy_to_user(ucursor, &context.cursor, sizeof(context.cursor)))
442 xfs_attrlist_by_handle(
443 struct file *parfilp,
444 struct xfs_fsop_attrlist_handlereq __user *p)
446 struct xfs_fsop_attrlist_handlereq al_hreq;
447 struct dentry *dentry;
450 if (!capable(CAP_SYS_ADMIN))
452 if (copy_from_user(&al_hreq, p, sizeof(al_hreq)))
455 dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
457 return PTR_ERR(dentry);
459 error = xfs_ioc_attr_list(XFS_I(d_inode(dentry)), al_hreq.buffer,
460 al_hreq.buflen, al_hreq.flags, &p->pos);
466 xfs_attrmulti_attr_get(
469 unsigned char __user *ubuf,
473 struct xfs_da_args args = {
475 .attr_filter = xfs_attr_filter(flags),
476 .attr_flags = xfs_attr_flags(flags),
478 .namelen = strlen(name),
483 if (*len > XFS_XATTR_SIZE_MAX)
486 error = xfs_attr_get(&args);
490 *len = args.valuelen;
491 if (copy_to_user(ubuf, args.value, args.valuelen))
495 kmem_free(args.value);
500 xfs_attrmulti_attr_set(
503 const unsigned char __user *ubuf,
507 struct xfs_da_args args = {
509 .attr_filter = xfs_attr_filter(flags),
510 .attr_flags = xfs_attr_flags(flags),
512 .namelen = strlen(name),
516 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
520 if (len > XFS_XATTR_SIZE_MAX)
522 args.value = memdup_user(ubuf, len);
523 if (IS_ERR(args.value))
524 return PTR_ERR(args.value);
528 error = xfs_attr_change(&args);
529 if (!error && (flags & XFS_IOC_ATTR_ROOT))
530 xfs_forget_acl(inode, name);
536 xfs_ioc_attrmulti_one(
537 struct file *parfilp,
548 if ((flags & XFS_IOC_ATTR_ROOT) && (flags & XFS_IOC_ATTR_SECURE))
551 name = strndup_user(uname, MAXNAMELEN);
553 return PTR_ERR(name);
557 error = xfs_attrmulti_attr_get(inode, name, value, len, flags);
564 error = mnt_want_write_file(parfilp);
567 error = xfs_attrmulti_attr_set(inode, name, value, *len, flags);
568 mnt_drop_write_file(parfilp);
580 xfs_attrmulti_by_handle(
581 struct file *parfilp,
585 xfs_attr_multiop_t *ops;
586 xfs_fsop_attrmulti_handlereq_t am_hreq;
587 struct dentry *dentry;
588 unsigned int i, size;
590 if (!capable(CAP_SYS_ADMIN))
592 if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
596 if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
599 dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
601 return PTR_ERR(dentry);
604 size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
605 if (!size || size > 16 * PAGE_SIZE)
608 ops = memdup_user(am_hreq.ops, size);
610 error = PTR_ERR(ops);
615 for (i = 0; i < am_hreq.opcount; i++) {
616 ops[i].am_error = xfs_ioc_attrmulti_one(parfilp,
617 d_inode(dentry), ops[i].am_opcode,
618 ops[i].am_attrname, ops[i].am_attrvalue,
619 &ops[i].am_length, ops[i].am_flags);
622 if (copy_to_user(am_hreq.ops, ops, size))
631 /* Return 0 on success or positive error */
633 xfs_fsbulkstat_one_fmt(
634 struct xfs_ibulk *breq,
635 const struct xfs_bulkstat *bstat)
637 struct xfs_bstat bs1;
639 xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
640 if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
642 return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
647 struct xfs_ibulk *breq,
648 const struct xfs_inumbers *igrp)
650 struct xfs_inogrp ig1;
652 xfs_inumbers_to_inogrp(&ig1, igrp);
653 if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
655 return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
664 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
665 struct xfs_fsop_bulkreq bulkreq;
666 struct xfs_ibulk breq = {
668 .mnt_userns = file_mnt_user_ns(file),
674 /* done = 1 if there are more stats to get and if bulkstat */
675 /* should be called again (unused here, but used in dmapi) */
677 if (!capable(CAP_SYS_ADMIN))
680 if (xfs_is_shutdown(mp))
683 if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
686 if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
689 if (bulkreq.icount <= 0)
692 if (bulkreq.ubuffer == NULL)
695 breq.ubuffer = bulkreq.ubuffer;
696 breq.icount = bulkreq.icount;
699 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
700 * that we want to stat. However, FSINUMBERS and FSBULKSTAT expect
701 * that *lastip contains either zero or the number of the last inode to
702 * be examined by the previous call and return results starting with
703 * the next inode after that. The new bulk request back end functions
704 * take the inode to start with, so we have to compute the startino
705 * parameter from lastino to maintain correct function. lastino == 0
706 * is a special case because it has traditionally meant "first inode
709 if (cmd == XFS_IOC_FSINUMBERS) {
710 breq.startino = lastino ? lastino + 1 : 0;
711 error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
712 lastino = breq.startino - 1;
713 } else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
714 breq.startino = lastino;
716 error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
717 } else { /* XFS_IOC_FSBULKSTAT */
718 breq.startino = lastino ? lastino + 1 : 0;
719 error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
720 lastino = breq.startino - 1;
726 if (bulkreq.lastip != NULL &&
727 copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
730 if (bulkreq.ocount != NULL &&
731 copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
737 /* Return 0 on success or positive error */
740 struct xfs_ibulk *breq,
741 const struct xfs_bulkstat *bstat)
743 if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
745 return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
749 * Check the incoming bulk request @hdr from userspace and initialize the
750 * internal @breq bulk request appropriately. Returns 0 if the bulk request
751 * should proceed; -ECANCELED if there's nothing to do; or the usual
752 * negative error code.
756 struct xfs_mount *mp,
757 struct xfs_bulk_ireq *hdr,
758 struct xfs_ibulk *breq,
759 void __user *ubuffer)
761 if (hdr->icount == 0 ||
762 (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
763 memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
766 breq->startino = hdr->ino;
767 breq->ubuffer = ubuffer;
768 breq->icount = hdr->icount;
773 * The @ino parameter is a special value, so we must look it up here.
774 * We're not allowed to have IREQ_AGNO, and we only return one inode
777 if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
778 if (hdr->flags & XFS_BULK_IREQ_AGNO)
782 case XFS_BULK_IREQ_SPECIAL_ROOT:
783 hdr->ino = mp->m_sb.sb_rootino;
792 * The IREQ_AGNO flag means that we only want results from a given AG.
793 * If @hdr->ino is zero, we start iterating in that AG. If @hdr->ino is
794 * beyond the specified AG then we return no results.
796 if (hdr->flags & XFS_BULK_IREQ_AGNO) {
797 if (hdr->agno >= mp->m_sb.sb_agcount)
800 if (breq->startino == 0)
801 breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
802 else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
805 breq->flags |= XFS_IBULK_SAME_AG;
807 /* Asking for an inode past the end of the AG? We're done! */
808 if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
810 } else if (hdr->agno)
813 /* Asking for an inode past the end of the FS? We're done! */
814 if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
817 if (hdr->flags & XFS_BULK_IREQ_NREXT64)
818 breq->flags |= XFS_IBULK_NREXT64;
824 * Update the userspace bulk request @hdr to reflect the end state of the
825 * internal bulk request @breq.
828 xfs_bulk_ireq_teardown(
829 struct xfs_bulk_ireq *hdr,
830 struct xfs_ibulk *breq)
832 hdr->ino = breq->startino;
833 hdr->ocount = breq->ocount;
836 /* Handle the v5 bulkstat ioctl. */
841 struct xfs_bulkstat_req __user *arg)
843 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
844 struct xfs_bulk_ireq hdr;
845 struct xfs_ibulk breq = {
847 .mnt_userns = file_mnt_user_ns(file),
851 if (!capable(CAP_SYS_ADMIN))
854 if (xfs_is_shutdown(mp))
857 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
860 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
861 if (error == -ECANCELED)
866 error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
871 xfs_bulk_ireq_teardown(&hdr, &breq);
872 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
880 struct xfs_ibulk *breq,
881 const struct xfs_inumbers *igrp)
883 if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
885 return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
888 /* Handle the v5 inumbers ioctl. */
891 struct xfs_mount *mp,
893 struct xfs_inumbers_req __user *arg)
895 struct xfs_bulk_ireq hdr;
896 struct xfs_ibulk breq = {
901 if (!capable(CAP_SYS_ADMIN))
904 if (xfs_is_shutdown(mp))
907 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
910 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
911 if (error == -ECANCELED)
916 error = xfs_inumbers(&breq, xfs_inumbers_fmt);
921 xfs_bulk_ireq_teardown(&hdr, &breq);
922 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
930 struct xfs_mount *mp,
934 struct xfs_fsop_geom fsgeo;
937 xfs_fs_geometry(mp, &fsgeo, struct_version);
939 if (struct_version <= 3)
940 len = sizeof(struct xfs_fsop_geom_v1);
941 else if (struct_version == 4)
942 len = sizeof(struct xfs_fsop_geom_v4);
944 xfs_fsop_geom_health(mp, &fsgeo);
948 if (copy_to_user(arg, &fsgeo, len))
955 struct xfs_mount *mp,
958 struct xfs_perag *pag;
959 struct xfs_ag_geometry ageo;
962 if (copy_from_user(&ageo, arg, sizeof(ageo)))
966 if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
969 pag = xfs_perag_get(mp, ageo.ag_number);
973 error = xfs_ag_get_geometry(pag, &ageo);
978 if (copy_to_user(arg, &ageo, sizeof(ageo)))
984 * Linux extended inode flags interface.
989 struct xfs_inode *ip,
993 struct xfs_mount *mp = ip->i_mount;
994 struct xfs_ifork *ifp = xfs_ifork_ptr(ip, whichfork);
996 fileattr_fill_xflags(fa, xfs_ip2xflags(ip));
998 if (ip->i_diflags & XFS_DIFLAG_EXTSIZE) {
999 fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1000 } else if (ip->i_diflags & XFS_DIFLAG_EXTSZINHERIT) {
1002 * Don't let a misaligned extent size hint on a directory
1003 * escape to userspace if it won't pass the setattr checks
1006 if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
1007 ip->i_extsize % mp->m_sb.sb_rextsize > 0) {
1008 fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE |
1009 FS_XFLAG_EXTSZINHERIT);
1010 fa->fsx_extsize = 0;
1012 fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1016 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1017 fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
1018 fa->fsx_projid = ip->i_projid;
1019 if (ifp && !xfs_need_iread_extents(ifp))
1020 fa->fsx_nextents = xfs_iext_count(ifp);
1022 fa->fsx_nextents = xfs_ifork_nextents(ifp);
1026 xfs_ioc_fsgetxattra(
1032 xfs_ilock(ip, XFS_ILOCK_SHARED);
1033 xfs_fill_fsxattr(ip, XFS_ATTR_FORK, &fa);
1034 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1036 return copy_fsxattr_to_user(&fa, arg);
1041 struct dentry *dentry,
1042 struct fileattr *fa)
1044 struct xfs_inode *ip = XFS_I(d_inode(dentry));
1046 if (d_is_special(dentry))
1049 xfs_ilock(ip, XFS_ILOCK_SHARED);
1050 xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
1051 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1058 struct xfs_inode *ip,
1059 unsigned int xflags)
1061 /* can't set PREALLOC this way, just preserve it */
1063 (ip->i_diflags & XFS_DIFLAG_PREALLOC);
1065 if (xflags & FS_XFLAG_IMMUTABLE)
1066 di_flags |= XFS_DIFLAG_IMMUTABLE;
1067 if (xflags & FS_XFLAG_APPEND)
1068 di_flags |= XFS_DIFLAG_APPEND;
1069 if (xflags & FS_XFLAG_SYNC)
1070 di_flags |= XFS_DIFLAG_SYNC;
1071 if (xflags & FS_XFLAG_NOATIME)
1072 di_flags |= XFS_DIFLAG_NOATIME;
1073 if (xflags & FS_XFLAG_NODUMP)
1074 di_flags |= XFS_DIFLAG_NODUMP;
1075 if (xflags & FS_XFLAG_NODEFRAG)
1076 di_flags |= XFS_DIFLAG_NODEFRAG;
1077 if (xflags & FS_XFLAG_FILESTREAM)
1078 di_flags |= XFS_DIFLAG_FILESTREAM;
1079 if (S_ISDIR(VFS_I(ip)->i_mode)) {
1080 if (xflags & FS_XFLAG_RTINHERIT)
1081 di_flags |= XFS_DIFLAG_RTINHERIT;
1082 if (xflags & FS_XFLAG_NOSYMLINKS)
1083 di_flags |= XFS_DIFLAG_NOSYMLINKS;
1084 if (xflags & FS_XFLAG_EXTSZINHERIT)
1085 di_flags |= XFS_DIFLAG_EXTSZINHERIT;
1086 if (xflags & FS_XFLAG_PROJINHERIT)
1087 di_flags |= XFS_DIFLAG_PROJINHERIT;
1088 } else if (S_ISREG(VFS_I(ip)->i_mode)) {
1089 if (xflags & FS_XFLAG_REALTIME)
1090 di_flags |= XFS_DIFLAG_REALTIME;
1091 if (xflags & FS_XFLAG_EXTSIZE)
1092 di_flags |= XFS_DIFLAG_EXTSIZE;
1100 struct xfs_inode *ip,
1101 unsigned int xflags)
1103 uint64_t di_flags2 =
1104 (ip->i_diflags2 & (XFS_DIFLAG2_REFLINK |
1105 XFS_DIFLAG2_BIGTIME |
1106 XFS_DIFLAG2_NREXT64));
1108 if (xflags & FS_XFLAG_DAX)
1109 di_flags2 |= XFS_DIFLAG2_DAX;
1110 if (xflags & FS_XFLAG_COWEXTSIZE)
1111 di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
1117 xfs_ioctl_setattr_xflags(
1118 struct xfs_trans *tp,
1119 struct xfs_inode *ip,
1120 struct fileattr *fa)
1122 struct xfs_mount *mp = ip->i_mount;
1125 /* Can't change realtime flag if any extents are allocated. */
1126 if ((ip->i_df.if_nextents || ip->i_delayed_blks) &&
1127 XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1130 /* If realtime flag is set then must have realtime device */
1131 if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1132 if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1133 (ip->i_extsize % mp->m_sb.sb_rextsize))
1137 /* Clear reflink if we are actually able to set the rt flag. */
1138 if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1139 ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
1141 /* Don't allow us to set DAX mode for a reflinked file for now. */
1142 if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1145 /* diflags2 only valid for v3 inodes. */
1146 i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1147 if (i_flags2 && !xfs_has_v3inodes(mp))
1150 ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1151 ip->i_diflags2 = i_flags2;
1153 xfs_diflags_to_iflags(ip, false);
1154 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1155 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1156 XFS_STATS_INC(mp, xs_ig_attrchg);
1161 xfs_ioctl_setattr_prepare_dax(
1162 struct xfs_inode *ip,
1163 struct fileattr *fa)
1165 struct xfs_mount *mp = ip->i_mount;
1166 struct inode *inode = VFS_I(ip);
1168 if (S_ISDIR(inode->i_mode))
1171 if (xfs_has_dax_always(mp) || xfs_has_dax_never(mp))
1174 if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
1175 !(ip->i_diflags2 & XFS_DIFLAG2_DAX)) ||
1176 (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
1177 (ip->i_diflags2 & XFS_DIFLAG2_DAX)))
1178 d_mark_dontcache(inode);
1182 * Set up the transaction structure for the setattr operation, checking that we
1183 * have permission to do so. On success, return a clean transaction and the
1184 * inode locked exclusively ready for further operation specific checks. On
1185 * failure, return an error without modifying or locking the inode.
1187 static struct xfs_trans *
1188 xfs_ioctl_setattr_get_trans(
1189 struct xfs_inode *ip,
1190 struct xfs_dquot *pdqp)
1192 struct xfs_mount *mp = ip->i_mount;
1193 struct xfs_trans *tp;
1196 if (xfs_is_readonly(mp))
1199 if (xfs_is_shutdown(mp))
1202 error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
1203 has_capability_noaudit(current, CAP_FOWNER), &tp);
1207 if (xfs_has_wsync(mp))
1208 xfs_trans_set_sync(tp);
1213 return ERR_PTR(error);
1217 * Validate a proposed extent size hint. For regular files, the hint can only
1218 * be changed if no extents are allocated.
1221 xfs_ioctl_setattr_check_extsize(
1222 struct xfs_inode *ip,
1223 struct fileattr *fa)
1225 struct xfs_mount *mp = ip->i_mount;
1226 xfs_failaddr_t failaddr;
1227 uint16_t new_diflags;
1232 if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
1233 XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
1236 if (fa->fsx_extsize & mp->m_blockmask)
1239 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1242 * Inode verifiers do not check that the extent size hint is an integer
1243 * multiple of the rt extent size on a directory with both rtinherit
1244 * and extszinherit flags set. Don't let sysadmins misconfigure
1247 if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
1248 (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
1249 unsigned int rtextsize_bytes;
1251 rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
1252 if (fa->fsx_extsize % rtextsize_bytes)
1256 failaddr = xfs_inode_validate_extsize(ip->i_mount,
1257 XFS_B_TO_FSB(mp, fa->fsx_extsize),
1258 VFS_I(ip)->i_mode, new_diflags);
1259 return failaddr != NULL ? -EINVAL : 0;
1263 xfs_ioctl_setattr_check_cowextsize(
1264 struct xfs_inode *ip,
1265 struct fileattr *fa)
1267 struct xfs_mount *mp = ip->i_mount;
1268 xfs_failaddr_t failaddr;
1269 uint64_t new_diflags2;
1270 uint16_t new_diflags;
1275 if (fa->fsx_cowextsize & mp->m_blockmask)
1278 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1279 new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1281 failaddr = xfs_inode_validate_cowextsize(ip->i_mount,
1282 XFS_B_TO_FSB(mp, fa->fsx_cowextsize),
1283 VFS_I(ip)->i_mode, new_diflags, new_diflags2);
1284 return failaddr != NULL ? -EINVAL : 0;
1288 xfs_ioctl_setattr_check_projid(
1289 struct xfs_inode *ip,
1290 struct fileattr *fa)
1295 /* Disallow 32bit project ids if 32bit IDs are not enabled. */
1296 if (fa->fsx_projid > (uint16_t)-1 &&
1297 !xfs_has_projid32(ip->i_mount))
1304 struct user_namespace *mnt_userns,
1305 struct dentry *dentry,
1306 struct fileattr *fa)
1308 struct xfs_inode *ip = XFS_I(d_inode(dentry));
1309 struct xfs_mount *mp = ip->i_mount;
1310 struct xfs_trans *tp;
1311 struct xfs_dquot *pdqp = NULL;
1312 struct xfs_dquot *olddquot = NULL;
1315 trace_xfs_ioctl_setattr(ip);
1317 if (d_is_special(dentry))
1320 if (!fa->fsx_valid) {
1321 if (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL |
1322 FS_NOATIME_FL | FS_NODUMP_FL |
1323 FS_SYNC_FL | FS_DAX_FL | FS_PROJINHERIT_FL))
1327 error = xfs_ioctl_setattr_check_projid(ip, fa);
1332 * If disk quotas is on, we make sure that the dquots do exist on disk,
1333 * before we start any other transactions. Trying to do this later
1334 * is messy. We don't care to take a readlock to look at the ids
1335 * in inode here, because we can't hold it across the trans_reserve.
1336 * If the IDs do change before we take the ilock, we're covered
1337 * because the i_*dquot fields will get updated anyway.
1339 if (fa->fsx_valid && XFS_IS_QUOTA_ON(mp)) {
1340 error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
1341 VFS_I(ip)->i_gid, fa->fsx_projid,
1342 XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
1347 xfs_ioctl_setattr_prepare_dax(ip, fa);
1349 tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
1351 error = PTR_ERR(tp);
1352 goto error_free_dquots;
1355 error = xfs_ioctl_setattr_check_extsize(ip, fa);
1357 goto error_trans_cancel;
1359 error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1361 goto error_trans_cancel;
1363 error = xfs_ioctl_setattr_xflags(tp, ip, fa);
1365 goto error_trans_cancel;
1370 * Change file ownership. Must be the owner or privileged. CAP_FSETID
1371 * overrides the following restrictions:
1373 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1374 * successful return from chown()
1377 if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1378 !capable_wrt_inode_uidgid(mnt_userns, VFS_I(ip), CAP_FSETID))
1379 VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1381 /* Change the ownerships and register project quota modifications */
1382 if (ip->i_projid != fa->fsx_projid) {
1383 if (XFS_IS_PQUOTA_ON(mp)) {
1384 olddquot = xfs_qm_vop_chown(tp, ip,
1385 &ip->i_pdquot, pdqp);
1387 ip->i_projid = fa->fsx_projid;
1391 * Only set the extent size hint if we've already determined that the
1392 * extent size hint should be set on the inode. If no extent size flags
1393 * are set on the inode then unconditionally clear the extent size hint.
1395 if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1396 ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1400 if (xfs_has_v3inodes(mp)) {
1401 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1402 ip->i_cowextsize = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1404 ip->i_cowextsize = 0;
1408 error = xfs_trans_commit(tp);
1411 * Release any dquot(s) the inode had kept before chown.
1413 xfs_qm_dqrele(olddquot);
1414 xfs_qm_dqrele(pdqp);
1419 xfs_trans_cancel(tp);
1421 xfs_qm_dqrele(pdqp);
1428 struct getbmapx __user *u,
1431 if (put_user(p->bmv_offset, &u->bmv_offset) ||
1432 put_user(p->bmv_block, &u->bmv_block) ||
1433 put_user(p->bmv_length, &u->bmv_length) ||
1434 put_user(0, &u->bmv_count) ||
1435 put_user(0, &u->bmv_entries))
1437 if (recsize < sizeof(struct getbmapx))
1439 if (put_user(0, &u->bmv_iflags) ||
1440 put_user(p->bmv_oflags, &u->bmv_oflags) ||
1441 put_user(0, &u->bmv_unused1) ||
1442 put_user(0, &u->bmv_unused2))
1453 struct getbmapx bmx = { 0 };
1454 struct kgetbmap *buf;
1459 case XFS_IOC_GETBMAPA:
1460 bmx.bmv_iflags = BMV_IF_ATTRFORK;
1462 case XFS_IOC_GETBMAP:
1463 /* struct getbmap is a strict subset of struct getbmapx. */
1464 recsize = sizeof(struct getbmap);
1466 case XFS_IOC_GETBMAPX:
1467 recsize = sizeof(struct getbmapx);
1473 if (copy_from_user(&bmx, arg, recsize))
1476 if (bmx.bmv_count < 2)
1478 if (bmx.bmv_count >= INT_MAX / recsize)
1481 buf = kvcalloc(bmx.bmv_count, sizeof(*buf), GFP_KERNEL);
1485 error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1490 if (copy_to_user(arg, &bmx, recsize))
1494 for (i = 0; i < bmx.bmv_entries; i++) {
1495 if (!xfs_getbmap_format(buf + i, arg, recsize))
1508 struct xfs_inode *ip,
1509 struct fsmap_head __user *arg)
1511 struct xfs_fsmap_head xhead = {0};
1512 struct fsmap_head head;
1515 __u32 last_flags = 0;
1519 if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1521 if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1522 memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1523 sizeof(head.fmh_keys[0].fmr_reserved)) ||
1524 memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1525 sizeof(head.fmh_keys[1].fmr_reserved)))
1529 * Use an internal memory buffer so that we don't have to copy fsmap
1530 * data to userspace while holding locks. Start by trying to allocate
1531 * up to 128k for the buffer, but fall back to a single page if needed.
1533 count = min_t(unsigned int, head.fmh_count,
1534 131072 / sizeof(struct fsmap));
1535 recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
1537 count = min_t(unsigned int, head.fmh_count,
1538 PAGE_SIZE / sizeof(struct fsmap));
1539 recs = kvcalloc(count, sizeof(struct fsmap), GFP_KERNEL);
1544 xhead.fmh_iflags = head.fmh_iflags;
1545 xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1546 xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1548 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1549 trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1551 head.fmh_entries = 0;
1553 struct fsmap __user *user_recs;
1554 struct fsmap *last_rec;
1556 user_recs = &arg->fmh_recs[head.fmh_entries];
1557 xhead.fmh_entries = 0;
1558 xhead.fmh_count = min_t(unsigned int, count,
1559 head.fmh_count - head.fmh_entries);
1561 /* Run query, record how many entries we got. */
1562 error = xfs_getfsmap(ip->i_mount, &xhead, recs);
1566 * There are no more records in the result set. Copy
1567 * whatever we got to userspace and break out.
1573 * The internal memory buffer is full. Copy whatever
1574 * records we got to userspace and go again if we have
1575 * not yet filled the userspace buffer.
1582 head.fmh_entries += xhead.fmh_entries;
1583 head.fmh_oflags = xhead.fmh_oflags;
1586 * If the caller wanted a record count or there aren't any
1587 * new records to return, we're done.
1589 if (head.fmh_count == 0 || xhead.fmh_entries == 0)
1592 /* Copy all the records we got out to userspace. */
1593 if (copy_to_user(user_recs, recs,
1594 xhead.fmh_entries * sizeof(struct fsmap))) {
1599 /* Remember the last record flags we copied to userspace. */
1600 last_rec = &recs[xhead.fmh_entries - 1];
1601 last_flags = last_rec->fmr_flags;
1603 /* Set up the low key for the next iteration. */
1604 xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
1605 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1606 } while (!done && head.fmh_entries < head.fmh_count);
1609 * If there are no more records in the query result set and we're not
1610 * in counting mode, mark the last record returned with the LAST flag.
1612 if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
1613 struct fsmap __user *user_rec;
1615 last_flags |= FMR_OF_LAST;
1616 user_rec = &arg->fmh_recs[head.fmh_entries - 1];
1618 if (copy_to_user(&user_rec->fmr_flags, &last_flags,
1619 sizeof(last_flags))) {
1625 /* copy back header */
1626 if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1637 xfs_ioc_scrub_metadata(
1641 struct xfs_scrub_metadata scrub;
1644 if (!capable(CAP_SYS_ADMIN))
1647 if (copy_from_user(&scrub, arg, sizeof(scrub)))
1650 error = xfs_scrub_metadata(file, &scrub);
1654 if (copy_to_user(arg, &scrub, sizeof(scrub)))
1664 xfs_inode_t *ip, *tip;
1668 /* Pull information for the target fd */
1669 f = fdget((int)sxp->sx_fdtarget);
1675 if (!(f.file->f_mode & FMODE_WRITE) ||
1676 !(f.file->f_mode & FMODE_READ) ||
1677 (f.file->f_flags & O_APPEND)) {
1682 tmp = fdget((int)sxp->sx_fdtmp);
1688 if (!(tmp.file->f_mode & FMODE_WRITE) ||
1689 !(tmp.file->f_mode & FMODE_READ) ||
1690 (tmp.file->f_flags & O_APPEND)) {
1692 goto out_put_tmp_file;
1695 if (IS_SWAPFILE(file_inode(f.file)) ||
1696 IS_SWAPFILE(file_inode(tmp.file))) {
1698 goto out_put_tmp_file;
1702 * We need to ensure that the fds passed in point to XFS inodes
1703 * before we cast and access them as XFS structures as we have no
1704 * control over what the user passes us here.
1706 if (f.file->f_op != &xfs_file_operations ||
1707 tmp.file->f_op != &xfs_file_operations) {
1709 goto out_put_tmp_file;
1712 ip = XFS_I(file_inode(f.file));
1713 tip = XFS_I(file_inode(tmp.file));
1715 if (ip->i_mount != tip->i_mount) {
1717 goto out_put_tmp_file;
1720 if (ip->i_ino == tip->i_ino) {
1722 goto out_put_tmp_file;
1725 if (xfs_is_shutdown(ip->i_mount)) {
1727 goto out_put_tmp_file;
1730 error = xfs_swap_extents(ip, tip, sxp);
1742 struct xfs_mount *mp,
1743 char __user *user_label)
1745 struct xfs_sb *sbp = &mp->m_sb;
1746 char label[XFSLABEL_MAX + 1];
1749 BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1751 /* 1 larger than sb_fname, so this ensures a trailing NUL char */
1752 memset(label, 0, sizeof(label));
1753 spin_lock(&mp->m_sb_lock);
1754 strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1755 spin_unlock(&mp->m_sb_lock);
1757 if (copy_to_user(user_label, label, sizeof(label)))
1765 struct xfs_mount *mp,
1766 char __user *newlabel)
1768 struct xfs_sb *sbp = &mp->m_sb;
1769 char label[XFSLABEL_MAX + 1];
1773 if (!capable(CAP_SYS_ADMIN))
1776 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1777 * smaller, at 12 bytes. We copy one more to be sure we find the
1778 * (required) NULL character to test the incoming label length.
1779 * NB: The on disk label doesn't need to be null terminated.
1781 if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1783 len = strnlen(label, XFSLABEL_MAX + 1);
1784 if (len > sizeof(sbp->sb_fname))
1787 error = mnt_want_write_file(filp);
1791 spin_lock(&mp->m_sb_lock);
1792 memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1793 memcpy(sbp->sb_fname, label, len);
1794 spin_unlock(&mp->m_sb_lock);
1797 * Now we do several things to satisfy userspace.
1798 * In addition to normal logging of the primary superblock, we also
1799 * immediately write these changes to sector zero for the primary, then
1800 * update all backup supers (as xfs_db does for a label change), then
1801 * invalidate the block device page cache. This is so that any prior
1802 * buffered reads from userspace (i.e. from blkid) are invalidated,
1803 * and userspace will see the newly-written label.
1805 error = xfs_sync_sb_buf(mp);
1809 * growfs also updates backup supers so lock against that.
1811 mutex_lock(&mp->m_growlock);
1812 error = xfs_update_secondary_sbs(mp);
1813 mutex_unlock(&mp->m_growlock);
1815 invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1818 mnt_drop_write_file(filp);
1823 xfs_fs_eofblocks_from_user(
1824 struct xfs_fs_eofblocks *src,
1825 struct xfs_icwalk *dst)
1827 if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1830 if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1833 if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1834 memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1838 if (src->eof_flags & XFS_EOF_FLAGS_SYNC)
1839 dst->icw_flags |= XFS_ICWALK_FLAG_SYNC;
1840 if (src->eof_flags & XFS_EOF_FLAGS_UID)
1841 dst->icw_flags |= XFS_ICWALK_FLAG_UID;
1842 if (src->eof_flags & XFS_EOF_FLAGS_GID)
1843 dst->icw_flags |= XFS_ICWALK_FLAG_GID;
1844 if (src->eof_flags & XFS_EOF_FLAGS_PRID)
1845 dst->icw_flags |= XFS_ICWALK_FLAG_PRID;
1846 if (src->eof_flags & XFS_EOF_FLAGS_MINFILESIZE)
1847 dst->icw_flags |= XFS_ICWALK_FLAG_MINFILESIZE;
1849 dst->icw_prid = src->eof_prid;
1850 dst->icw_min_file_size = src->eof_min_file_size;
1852 dst->icw_uid = INVALID_UID;
1853 if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1854 dst->icw_uid = make_kuid(current_user_ns(), src->eof_uid);
1855 if (!uid_valid(dst->icw_uid))
1859 dst->icw_gid = INVALID_GID;
1860 if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1861 dst->icw_gid = make_kgid(current_user_ns(), src->eof_gid);
1862 if (!gid_valid(dst->icw_gid))
1869 * These long-unused ioctls were removed from the official ioctl API in 5.17,
1870 * but retain these definitions so that we can log warnings about them.
1872 #define XFS_IOC_ALLOCSP _IOW ('X', 10, struct xfs_flock64)
1873 #define XFS_IOC_FREESP _IOW ('X', 11, struct xfs_flock64)
1874 #define XFS_IOC_ALLOCSP64 _IOW ('X', 36, struct xfs_flock64)
1875 #define XFS_IOC_FREESP64 _IOW ('X', 37, struct xfs_flock64)
1878 * Note: some of the ioctl's return positive numbers as a
1879 * byte count indicating success, such as readlink_by_handle.
1880 * So we don't "sign flip" like most other routines. This means
1881 * true errors need to be returned as a negative value.
1889 struct inode *inode = file_inode(filp);
1890 struct xfs_inode *ip = XFS_I(inode);
1891 struct xfs_mount *mp = ip->i_mount;
1892 void __user *arg = (void __user *)p;
1895 trace_xfs_file_ioctl(ip);
1899 return xfs_ioc_trim(mp, arg);
1900 case FS_IOC_GETFSLABEL:
1901 return xfs_ioc_getlabel(mp, arg);
1902 case FS_IOC_SETFSLABEL:
1903 return xfs_ioc_setlabel(filp, mp, arg);
1904 case XFS_IOC_ALLOCSP:
1905 case XFS_IOC_FREESP:
1906 case XFS_IOC_ALLOCSP64:
1907 case XFS_IOC_FREESP64:
1909 "%s should use fallocate; XFS_IOC_{ALLOC,FREE}SP ioctl unsupported",
1912 case XFS_IOC_DIOINFO: {
1913 struct xfs_buftarg *target = xfs_inode_buftarg(ip);
1916 da.d_mem = da.d_miniosz = target->bt_logical_sectorsize;
1917 da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1919 if (copy_to_user(arg, &da, sizeof(da)))
1924 case XFS_IOC_FSBULKSTAT_SINGLE:
1925 case XFS_IOC_FSBULKSTAT:
1926 case XFS_IOC_FSINUMBERS:
1927 return xfs_ioc_fsbulkstat(filp, cmd, arg);
1929 case XFS_IOC_BULKSTAT:
1930 return xfs_ioc_bulkstat(filp, cmd, arg);
1931 case XFS_IOC_INUMBERS:
1932 return xfs_ioc_inumbers(mp, cmd, arg);
1934 case XFS_IOC_FSGEOMETRY_V1:
1935 return xfs_ioc_fsgeometry(mp, arg, 3);
1936 case XFS_IOC_FSGEOMETRY_V4:
1937 return xfs_ioc_fsgeometry(mp, arg, 4);
1938 case XFS_IOC_FSGEOMETRY:
1939 return xfs_ioc_fsgeometry(mp, arg, 5);
1941 case XFS_IOC_AG_GEOMETRY:
1942 return xfs_ioc_ag_geometry(mp, arg);
1944 case XFS_IOC_GETVERSION:
1945 return put_user(inode->i_generation, (int __user *)arg);
1947 case XFS_IOC_FSGETXATTRA:
1948 return xfs_ioc_fsgetxattra(ip, arg);
1950 case XFS_IOC_GETBMAP:
1951 case XFS_IOC_GETBMAPA:
1952 case XFS_IOC_GETBMAPX:
1953 return xfs_ioc_getbmap(filp, cmd, arg);
1955 case FS_IOC_GETFSMAP:
1956 return xfs_ioc_getfsmap(ip, arg);
1958 case XFS_IOC_SCRUB_METADATA:
1959 return xfs_ioc_scrub_metadata(filp, arg);
1961 case XFS_IOC_FD_TO_HANDLE:
1962 case XFS_IOC_PATH_TO_HANDLE:
1963 case XFS_IOC_PATH_TO_FSHANDLE: {
1964 xfs_fsop_handlereq_t hreq;
1966 if (copy_from_user(&hreq, arg, sizeof(hreq)))
1968 return xfs_find_handle(cmd, &hreq);
1970 case XFS_IOC_OPEN_BY_HANDLE: {
1971 xfs_fsop_handlereq_t hreq;
1973 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1975 return xfs_open_by_handle(filp, &hreq);
1978 case XFS_IOC_READLINK_BY_HANDLE: {
1979 xfs_fsop_handlereq_t hreq;
1981 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
1983 return xfs_readlink_by_handle(filp, &hreq);
1985 case XFS_IOC_ATTRLIST_BY_HANDLE:
1986 return xfs_attrlist_by_handle(filp, arg);
1988 case XFS_IOC_ATTRMULTI_BY_HANDLE:
1989 return xfs_attrmulti_by_handle(filp, arg);
1991 case XFS_IOC_SWAPEXT: {
1992 struct xfs_swapext sxp;
1994 if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
1996 error = mnt_want_write_file(filp);
1999 error = xfs_ioc_swapext(&sxp);
2000 mnt_drop_write_file(filp);
2004 case XFS_IOC_FSCOUNTS: {
2005 xfs_fsop_counts_t out;
2007 xfs_fs_counts(mp, &out);
2009 if (copy_to_user(arg, &out, sizeof(out)))
2014 case XFS_IOC_SET_RESBLKS: {
2015 xfs_fsop_resblks_t inout;
2018 if (!capable(CAP_SYS_ADMIN))
2021 if (xfs_is_readonly(mp))
2024 if (copy_from_user(&inout, arg, sizeof(inout)))
2027 error = mnt_want_write_file(filp);
2031 /* input parameter is passed in resblks field of structure */
2033 error = xfs_reserve_blocks(mp, &in, &inout);
2034 mnt_drop_write_file(filp);
2038 if (copy_to_user(arg, &inout, sizeof(inout)))
2043 case XFS_IOC_GET_RESBLKS: {
2044 xfs_fsop_resblks_t out;
2046 if (!capable(CAP_SYS_ADMIN))
2049 error = xfs_reserve_blocks(mp, NULL, &out);
2053 if (copy_to_user(arg, &out, sizeof(out)))
2059 case XFS_IOC_FSGROWFSDATA: {
2060 struct xfs_growfs_data in;
2062 if (copy_from_user(&in, arg, sizeof(in)))
2065 error = mnt_want_write_file(filp);
2068 error = xfs_growfs_data(mp, &in);
2069 mnt_drop_write_file(filp);
2073 case XFS_IOC_FSGROWFSLOG: {
2074 struct xfs_growfs_log in;
2076 if (copy_from_user(&in, arg, sizeof(in)))
2079 error = mnt_want_write_file(filp);
2082 error = xfs_growfs_log(mp, &in);
2083 mnt_drop_write_file(filp);
2087 case XFS_IOC_FSGROWFSRT: {
2090 if (copy_from_user(&in, arg, sizeof(in)))
2093 error = mnt_want_write_file(filp);
2096 error = xfs_growfs_rt(mp, &in);
2097 mnt_drop_write_file(filp);
2101 case XFS_IOC_GOINGDOWN: {
2104 if (!capable(CAP_SYS_ADMIN))
2107 if (get_user(in, (uint32_t __user *)arg))
2110 return xfs_fs_goingdown(mp, in);
2113 case XFS_IOC_ERROR_INJECTION: {
2114 xfs_error_injection_t in;
2116 if (!capable(CAP_SYS_ADMIN))
2119 if (copy_from_user(&in, arg, sizeof(in)))
2122 return xfs_errortag_add(mp, in.errtag);
2125 case XFS_IOC_ERROR_CLEARALL:
2126 if (!capable(CAP_SYS_ADMIN))
2129 return xfs_errortag_clearall(mp);
2131 case XFS_IOC_FREE_EOFBLOCKS: {
2132 struct xfs_fs_eofblocks eofb;
2133 struct xfs_icwalk icw;
2135 if (!capable(CAP_SYS_ADMIN))
2138 if (xfs_is_readonly(mp))
2141 if (copy_from_user(&eofb, arg, sizeof(eofb)))
2144 error = xfs_fs_eofblocks_from_user(&eofb, &icw);
2148 trace_xfs_ioc_free_eofblocks(mp, &icw, _RET_IP_);
2150 sb_start_write(mp->m_super);
2151 error = xfs_blockgc_free_space(mp, &icw);
2152 sb_end_write(mp->m_super);