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"
20 #include "xfs_bmap_util.h"
21 #include "xfs_fsops.h"
22 #include "xfs_discard.h"
23 #include "xfs_quota.h"
24 #include "xfs_export.h"
25 #include "xfs_trace.h"
26 #include "xfs_icache.h"
27 #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_da_format.h"
39 #include "xfs_da_btree.h"
41 #include <linux/mount.h>
42 #include <linux/namei.h>
45 * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
46 * a file or fs handle.
48 * XFS_IOC_PATH_TO_FSHANDLE
49 * returns fs handle for a mount point or path within that mount point
50 * XFS_IOC_FD_TO_HANDLE
51 * returns full handle for a FD opened in user space
52 * XFS_IOC_PATH_TO_HANDLE
53 * returns full handle for a path
58 xfs_fsop_handlereq_t *hreq)
68 if (cmd == XFS_IOC_FD_TO_HANDLE) {
72 inode = file_inode(f.file);
74 error = user_path_at(AT_FDCWD, hreq->path, 0, &path);
77 inode = d_inode(path.dentry);
82 * We can only generate handles for inodes residing on a XFS filesystem,
83 * and only for regular files, directories or symbolic links.
86 if (inode->i_sb->s_magic != XFS_SB_MAGIC)
90 if (!S_ISREG(inode->i_mode) &&
91 !S_ISDIR(inode->i_mode) &&
92 !S_ISLNK(inode->i_mode))
96 memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
98 if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
100 * This handle only contains an fsid, zero the rest.
102 memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
103 hsize = sizeof(xfs_fsid_t);
105 handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
106 sizeof(handle.ha_fid.fid_len);
107 handle.ha_fid.fid_pad = 0;
108 handle.ha_fid.fid_gen = inode->i_generation;
109 handle.ha_fid.fid_ino = ip->i_ino;
110 hsize = sizeof(xfs_handle_t);
114 if (copy_to_user(hreq->ohandle, &handle, hsize) ||
115 copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
121 if (cmd == XFS_IOC_FD_TO_HANDLE)
129 * No need to do permission checks on the various pathname components
130 * as the handle operations are privileged.
133 xfs_handle_acceptable(
135 struct dentry *dentry)
141 * Convert userspace handle data into a dentry.
144 xfs_handle_to_dentry(
145 struct file *parfilp,
146 void __user *uhandle,
150 struct xfs_fid64 fid;
153 * Only allow handle opens under a directory.
155 if (!S_ISDIR(file_inode(parfilp)->i_mode))
156 return ERR_PTR(-ENOTDIR);
158 if (hlen != sizeof(xfs_handle_t))
159 return ERR_PTR(-EINVAL);
160 if (copy_from_user(&handle, uhandle, hlen))
161 return ERR_PTR(-EFAULT);
162 if (handle.ha_fid.fid_len !=
163 sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
164 return ERR_PTR(-EINVAL);
166 memset(&fid, 0, sizeof(struct fid));
167 fid.ino = handle.ha_fid.fid_ino;
168 fid.gen = handle.ha_fid.fid_gen;
170 return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
171 FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
172 xfs_handle_acceptable, NULL);
175 STATIC struct dentry *
176 xfs_handlereq_to_dentry(
177 struct file *parfilp,
178 xfs_fsop_handlereq_t *hreq)
180 return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
185 struct file *parfilp,
186 xfs_fsop_handlereq_t *hreq)
188 const struct cred *cred = current_cred();
194 struct dentry *dentry;
198 if (!capable(CAP_SYS_ADMIN))
201 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
203 return PTR_ERR(dentry);
204 inode = d_inode(dentry);
206 /* Restrict xfs_open_by_handle to directories & regular files. */
207 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
212 #if BITS_PER_LONG != 32
213 hreq->oflags |= O_LARGEFILE;
216 permflag = hreq->oflags;
217 fmode = OPEN_FMODE(permflag);
218 if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
219 (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
224 if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
229 /* Can't write directories. */
230 if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
235 fd = get_unused_fd_flags(0);
241 path.mnt = parfilp->f_path.mnt;
242 path.dentry = dentry;
243 filp = dentry_open(&path, hreq->oflags, cred);
247 return PTR_ERR(filp);
250 if (S_ISREG(inode->i_mode)) {
251 filp->f_flags |= O_NOATIME;
252 filp->f_mode |= FMODE_NOCMTIME;
255 fd_install(fd, filp);
264 xfs_readlink_by_handle(
265 struct file *parfilp,
266 xfs_fsop_handlereq_t *hreq)
268 struct dentry *dentry;
272 if (!capable(CAP_SYS_ADMIN))
275 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
277 return PTR_ERR(dentry);
279 /* Restrict this handle operation to symlinks only. */
280 if (!d_is_symlink(dentry)) {
285 if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
290 error = vfs_readlink(dentry, hreq->ohandle, olen);
298 * Format an attribute and copy it out to the user's buffer.
299 * Take care to check values and protect against them changing later,
300 * we may be reading them directly out of a user buffer.
303 xfs_ioc_attr_put_listent(
304 struct xfs_attr_list_context *context,
310 struct xfs_attrlist *alist = context->buffer;
311 struct xfs_attrlist_ent *aep;
314 ASSERT(!context->seen_enough);
315 ASSERT(context->count >= 0);
316 ASSERT(context->count < (ATTR_MAX_VALUELEN/8));
317 ASSERT(context->firstu >= sizeof(*alist));
318 ASSERT(context->firstu <= context->bufsize);
321 * Only list entries in the right namespace.
323 if (context->attr_filter != (flags & XFS_ATTR_NSP_ONDISK_MASK))
326 arraytop = sizeof(*alist) +
327 context->count * sizeof(alist->al_offset[0]);
329 /* decrement by the actual bytes used by the attr */
330 context->firstu -= round_up(offsetof(struct xfs_attrlist_ent, a_name) +
331 namelen + 1, sizeof(uint32_t));
332 if (context->firstu < arraytop) {
333 trace_xfs_attr_list_full(context);
335 context->seen_enough = 1;
339 aep = context->buffer + context->firstu;
340 aep->a_valuelen = valuelen;
341 memcpy(aep->a_name, name, namelen);
342 aep->a_name[namelen] = 0;
343 alist->al_offset[context->count++] = context->firstu;
344 alist->al_count = context->count;
345 trace_xfs_attr_list_add(context);
352 if (ioc_flags & XFS_IOC_ATTR_ROOT)
353 return XFS_ATTR_ROOT;
354 if (ioc_flags & XFS_IOC_ATTR_SECURE)
355 return XFS_ATTR_SECURE;
363 if (ioc_flags & XFS_IOC_ATTR_CREATE)
365 if (ioc_flags & XFS_IOC_ATTR_REPLACE)
366 return XATTR_REPLACE;
372 struct xfs_inode *dp,
376 struct xfs_attrlist_cursor __user *ucursor)
378 struct xfs_attr_list_context context = { };
379 struct xfs_attrlist *alist;
383 if (bufsize < sizeof(struct xfs_attrlist) ||
384 bufsize > XFS_XATTR_LIST_MAX)
388 * Reject flags, only allow namespaces.
390 if (flags & ~(XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
392 if (flags == (XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
396 * Validate the cursor.
398 if (copy_from_user(&context.cursor, ucursor, sizeof(context.cursor)))
400 if (context.cursor.pad1 || context.cursor.pad2)
402 if (!context.cursor.initted &&
403 (context.cursor.hashval || context.cursor.blkno ||
404 context.cursor.offset))
407 buffer = kvzalloc(bufsize, GFP_KERNEL);
412 * Initialize the output buffer.
416 context.attr_filter = xfs_attr_filter(flags);
417 context.buffer = buffer;
418 context.bufsize = round_down(bufsize, sizeof(uint32_t));
419 context.firstu = context.bufsize;
420 context.put_listent = xfs_ioc_attr_put_listent;
422 alist = context.buffer;
425 alist->al_offset[0] = context.bufsize;
427 error = xfs_attr_list(&context);
431 if (copy_to_user(ubuf, buffer, bufsize) ||
432 copy_to_user(ucursor, &context.cursor, sizeof(context.cursor)))
440 xfs_attrlist_by_handle(
441 struct file *parfilp,
442 struct xfs_fsop_attrlist_handlereq __user *p)
444 struct xfs_fsop_attrlist_handlereq al_hreq;
445 struct dentry *dentry;
448 if (!capable(CAP_SYS_ADMIN))
450 if (copy_from_user(&al_hreq, p, sizeof(al_hreq)))
453 dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
455 return PTR_ERR(dentry);
457 error = xfs_ioc_attr_list(XFS_I(d_inode(dentry)), al_hreq.buffer,
458 al_hreq.buflen, al_hreq.flags, &p->pos);
464 xfs_attrmulti_attr_get(
467 unsigned char __user *ubuf,
471 struct xfs_da_args args = {
473 .attr_filter = xfs_attr_filter(flags),
474 .attr_flags = xfs_attr_flags(flags),
476 .namelen = strlen(name),
481 if (*len > XFS_XATTR_SIZE_MAX)
484 error = xfs_attr_get(&args);
488 *len = args.valuelen;
489 if (copy_to_user(ubuf, args.value, args.valuelen))
493 kmem_free(args.value);
498 xfs_attrmulti_attr_set(
501 const unsigned char __user *ubuf,
505 struct xfs_da_args args = {
507 .attr_filter = xfs_attr_filter(flags),
508 .attr_flags = xfs_attr_flags(flags),
510 .namelen = strlen(name),
514 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
518 if (len > XFS_XATTR_SIZE_MAX)
520 args.value = memdup_user(ubuf, len);
521 if (IS_ERR(args.value))
522 return PTR_ERR(args.value);
526 error = xfs_attr_set(&args);
527 if (!error && (flags & XFS_IOC_ATTR_ROOT))
528 xfs_forget_acl(inode, name);
534 xfs_ioc_attrmulti_one(
535 struct file *parfilp,
546 if ((flags & XFS_IOC_ATTR_ROOT) && (flags & XFS_IOC_ATTR_SECURE))
549 name = strndup_user(uname, MAXNAMELEN);
551 return PTR_ERR(name);
555 error = xfs_attrmulti_attr_get(inode, name, value, len, flags);
562 error = mnt_want_write_file(parfilp);
565 error = xfs_attrmulti_attr_set(inode, name, value, *len, flags);
566 mnt_drop_write_file(parfilp);
578 xfs_attrmulti_by_handle(
579 struct file *parfilp,
583 xfs_attr_multiop_t *ops;
584 xfs_fsop_attrmulti_handlereq_t am_hreq;
585 struct dentry *dentry;
586 unsigned int i, size;
588 if (!capable(CAP_SYS_ADMIN))
590 if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
594 if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
597 dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
599 return PTR_ERR(dentry);
602 size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
603 if (!size || size > 16 * PAGE_SIZE)
606 ops = memdup_user(am_hreq.ops, size);
608 error = PTR_ERR(ops);
613 for (i = 0; i < am_hreq.opcount; i++) {
614 ops[i].am_error = xfs_ioc_attrmulti_one(parfilp,
615 d_inode(dentry), ops[i].am_opcode,
616 ops[i].am_attrname, ops[i].am_attrvalue,
617 &ops[i].am_length, ops[i].am_flags);
620 if (copy_to_user(am_hreq.ops, ops, size))
634 struct inode *inode = file_inode(filp);
635 struct xfs_inode *ip = XFS_I(inode);
637 enum xfs_prealloc_flags flags = XFS_PREALLOC_CLEAR;
638 uint iolock = XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL;
641 if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
644 if (!(filp->f_mode & FMODE_WRITE))
647 if (!S_ISREG(inode->i_mode))
650 if (xfs_is_always_cow_inode(ip))
653 if (filp->f_flags & O_DSYNC)
654 flags |= XFS_PREALLOC_SYNC;
655 if (filp->f_mode & FMODE_NOCMTIME)
656 flags |= XFS_PREALLOC_INVISIBLE;
658 error = mnt_want_write_file(filp);
662 xfs_ilock(ip, iolock);
663 error = xfs_break_layouts(inode, &iolock, BREAK_UNMAP);
666 inode_dio_wait(inode);
668 switch (bf->l_whence) {
672 bf->l_start += filp->f_pos;
675 bf->l_start += XFS_ISIZE(ip);
682 if (bf->l_start < 0 || bf->l_start > inode->i_sb->s_maxbytes) {
687 if (bf->l_start > XFS_ISIZE(ip)) {
688 error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
689 bf->l_start - XFS_ISIZE(ip), 0);
694 iattr.ia_valid = ATTR_SIZE;
695 iattr.ia_size = bf->l_start;
696 error = xfs_vn_setattr_size(file_dentry(filp), &iattr);
700 error = xfs_update_prealloc_flags(ip, flags);
703 xfs_iunlock(ip, iolock);
704 mnt_drop_write_file(filp);
708 /* Return 0 on success or positive error */
710 xfs_fsbulkstat_one_fmt(
711 struct xfs_ibulk *breq,
712 const struct xfs_bulkstat *bstat)
714 struct xfs_bstat bs1;
716 xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
717 if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
719 return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
724 struct xfs_ibulk *breq,
725 const struct xfs_inumbers *igrp)
727 struct xfs_inogrp ig1;
729 xfs_inumbers_to_inogrp(&ig1, igrp);
730 if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
732 return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
741 struct xfs_fsop_bulkreq bulkreq;
742 struct xfs_ibulk breq = {
749 /* done = 1 if there are more stats to get and if bulkstat */
750 /* should be called again (unused here, but used in dmapi) */
752 if (!capable(CAP_SYS_ADMIN))
755 if (XFS_FORCED_SHUTDOWN(mp))
758 if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
761 if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
764 if (bulkreq.icount <= 0)
767 if (bulkreq.ubuffer == NULL)
770 breq.ubuffer = bulkreq.ubuffer;
771 breq.icount = bulkreq.icount;
774 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
775 * that we want to stat. However, FSINUMBERS and FSBULKSTAT expect
776 * that *lastip contains either zero or the number of the last inode to
777 * be examined by the previous call and return results starting with
778 * the next inode after that. The new bulk request back end functions
779 * take the inode to start with, so we have to compute the startino
780 * parameter from lastino to maintain correct function. lastino == 0
781 * is a special case because it has traditionally meant "first inode
784 if (cmd == XFS_IOC_FSINUMBERS) {
785 breq.startino = lastino ? lastino + 1 : 0;
786 error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
787 lastino = breq.startino - 1;
788 } else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
789 breq.startino = lastino;
791 error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
792 } else { /* XFS_IOC_FSBULKSTAT */
793 breq.startino = lastino ? lastino + 1 : 0;
794 error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
795 lastino = breq.startino - 1;
801 if (bulkreq.lastip != NULL &&
802 copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
805 if (bulkreq.ocount != NULL &&
806 copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
812 /* Return 0 on success or positive error */
815 struct xfs_ibulk *breq,
816 const struct xfs_bulkstat *bstat)
818 if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
820 return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
824 * Check the incoming bulk request @hdr from userspace and initialize the
825 * internal @breq bulk request appropriately. Returns 0 if the bulk request
826 * should proceed; -ECANCELED if there's nothing to do; or the usual
827 * negative error code.
831 struct xfs_mount *mp,
832 struct xfs_bulk_ireq *hdr,
833 struct xfs_ibulk *breq,
834 void __user *ubuffer)
836 if (hdr->icount == 0 ||
837 (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
838 memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
841 breq->startino = hdr->ino;
842 breq->ubuffer = ubuffer;
843 breq->icount = hdr->icount;
848 * The @ino parameter is a special value, so we must look it up here.
849 * We're not allowed to have IREQ_AGNO, and we only return one inode
852 if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
853 if (hdr->flags & XFS_BULK_IREQ_AGNO)
857 case XFS_BULK_IREQ_SPECIAL_ROOT:
858 hdr->ino = mp->m_sb.sb_rootino;
867 * The IREQ_AGNO flag means that we only want results from a given AG.
868 * If @hdr->ino is zero, we start iterating in that AG. If @hdr->ino is
869 * beyond the specified AG then we return no results.
871 if (hdr->flags & XFS_BULK_IREQ_AGNO) {
872 if (hdr->agno >= mp->m_sb.sb_agcount)
875 if (breq->startino == 0)
876 breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
877 else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
880 breq->flags |= XFS_IBULK_SAME_AG;
882 /* Asking for an inode past the end of the AG? We're done! */
883 if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
885 } else if (hdr->agno)
888 /* Asking for an inode past the end of the FS? We're done! */
889 if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
896 * Update the userspace bulk request @hdr to reflect the end state of the
897 * internal bulk request @breq.
900 xfs_bulk_ireq_teardown(
901 struct xfs_bulk_ireq *hdr,
902 struct xfs_ibulk *breq)
904 hdr->ino = breq->startino;
905 hdr->ocount = breq->ocount;
908 /* Handle the v5 bulkstat ioctl. */
911 struct xfs_mount *mp,
913 struct xfs_bulkstat_req __user *arg)
915 struct xfs_bulk_ireq hdr;
916 struct xfs_ibulk breq = {
921 if (!capable(CAP_SYS_ADMIN))
924 if (XFS_FORCED_SHUTDOWN(mp))
927 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
930 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
931 if (error == -ECANCELED)
936 error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
941 xfs_bulk_ireq_teardown(&hdr, &breq);
942 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
950 struct xfs_ibulk *breq,
951 const struct xfs_inumbers *igrp)
953 if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
955 return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
958 /* Handle the v5 inumbers ioctl. */
961 struct xfs_mount *mp,
963 struct xfs_inumbers_req __user *arg)
965 struct xfs_bulk_ireq hdr;
966 struct xfs_ibulk breq = {
971 if (!capable(CAP_SYS_ADMIN))
974 if (XFS_FORCED_SHUTDOWN(mp))
977 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
980 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
981 if (error == -ECANCELED)
986 error = xfs_inumbers(&breq, xfs_inumbers_fmt);
991 xfs_bulk_ireq_teardown(&hdr, &breq);
992 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
1000 struct xfs_mount *mp,
1004 struct xfs_fsop_geom fsgeo;
1007 xfs_fs_geometry(&mp->m_sb, &fsgeo, struct_version);
1009 if (struct_version <= 3)
1010 len = sizeof(struct xfs_fsop_geom_v1);
1011 else if (struct_version == 4)
1012 len = sizeof(struct xfs_fsop_geom_v4);
1014 xfs_fsop_geom_health(mp, &fsgeo);
1015 len = sizeof(fsgeo);
1018 if (copy_to_user(arg, &fsgeo, len))
1024 xfs_ioc_ag_geometry(
1025 struct xfs_mount *mp,
1028 struct xfs_ag_geometry ageo;
1031 if (copy_from_user(&ageo, arg, sizeof(ageo)))
1035 if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
1038 error = xfs_ag_get_geometry(mp, ageo.ag_number, &ageo);
1042 if (copy_to_user(arg, &ageo, sizeof(ageo)))
1048 * Linux extended inode flags interface.
1052 xfs_merge_ioc_xflags(
1056 unsigned int xflags = start;
1058 if (flags & FS_IMMUTABLE_FL)
1059 xflags |= FS_XFLAG_IMMUTABLE;
1061 xflags &= ~FS_XFLAG_IMMUTABLE;
1062 if (flags & FS_APPEND_FL)
1063 xflags |= FS_XFLAG_APPEND;
1065 xflags &= ~FS_XFLAG_APPEND;
1066 if (flags & FS_SYNC_FL)
1067 xflags |= FS_XFLAG_SYNC;
1069 xflags &= ~FS_XFLAG_SYNC;
1070 if (flags & FS_NOATIME_FL)
1071 xflags |= FS_XFLAG_NOATIME;
1073 xflags &= ~FS_XFLAG_NOATIME;
1074 if (flags & FS_NODUMP_FL)
1075 xflags |= FS_XFLAG_NODUMP;
1077 xflags &= ~FS_XFLAG_NODUMP;
1078 if (flags & FS_DAX_FL)
1079 xflags |= FS_XFLAG_DAX;
1081 xflags &= ~FS_XFLAG_DAX;
1091 unsigned int flags = 0;
1093 if (di_flags & XFS_DIFLAG_IMMUTABLE)
1094 flags |= FS_IMMUTABLE_FL;
1095 if (di_flags & XFS_DIFLAG_APPEND)
1096 flags |= FS_APPEND_FL;
1097 if (di_flags & XFS_DIFLAG_SYNC)
1098 flags |= FS_SYNC_FL;
1099 if (di_flags & XFS_DIFLAG_NOATIME)
1100 flags |= FS_NOATIME_FL;
1101 if (di_flags & XFS_DIFLAG_NODUMP)
1102 flags |= FS_NODUMP_FL;
1103 if (di_flags2 & XFS_DIFLAG2_DAX) {
1111 struct xfs_inode *ip,
1115 struct xfs_ifork *ifp = attr ? ip->i_afp : &ip->i_df;
1117 simple_fill_fsxattr(fa, xfs_ip2xflags(ip));
1118 fa->fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
1119 fa->fsx_cowextsize = ip->i_d.di_cowextsize <<
1120 ip->i_mount->m_sb.sb_blocklog;
1121 fa->fsx_projid = ip->i_d.di_projid;
1122 if (ifp && (ifp->if_flags & XFS_IFEXTENTS))
1123 fa->fsx_nextents = xfs_iext_count(ifp);
1125 fa->fsx_nextents = xfs_ifork_nextents(ifp);
1136 xfs_ilock(ip, XFS_ILOCK_SHARED);
1137 xfs_fill_fsxattr(ip, attr, &fa);
1138 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1140 if (copy_to_user(arg, &fa, sizeof(fa)))
1147 struct xfs_inode *ip,
1148 unsigned int xflags)
1150 /* can't set PREALLOC this way, just preserve it */
1152 (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
1154 if (xflags & FS_XFLAG_IMMUTABLE)
1155 di_flags |= XFS_DIFLAG_IMMUTABLE;
1156 if (xflags & FS_XFLAG_APPEND)
1157 di_flags |= XFS_DIFLAG_APPEND;
1158 if (xflags & FS_XFLAG_SYNC)
1159 di_flags |= XFS_DIFLAG_SYNC;
1160 if (xflags & FS_XFLAG_NOATIME)
1161 di_flags |= XFS_DIFLAG_NOATIME;
1162 if (xflags & FS_XFLAG_NODUMP)
1163 di_flags |= XFS_DIFLAG_NODUMP;
1164 if (xflags & FS_XFLAG_NODEFRAG)
1165 di_flags |= XFS_DIFLAG_NODEFRAG;
1166 if (xflags & FS_XFLAG_FILESTREAM)
1167 di_flags |= XFS_DIFLAG_FILESTREAM;
1168 if (S_ISDIR(VFS_I(ip)->i_mode)) {
1169 if (xflags & FS_XFLAG_RTINHERIT)
1170 di_flags |= XFS_DIFLAG_RTINHERIT;
1171 if (xflags & FS_XFLAG_NOSYMLINKS)
1172 di_flags |= XFS_DIFLAG_NOSYMLINKS;
1173 if (xflags & FS_XFLAG_EXTSZINHERIT)
1174 di_flags |= XFS_DIFLAG_EXTSZINHERIT;
1175 if (xflags & FS_XFLAG_PROJINHERIT)
1176 di_flags |= XFS_DIFLAG_PROJINHERIT;
1177 } else if (S_ISREG(VFS_I(ip)->i_mode)) {
1178 if (xflags & FS_XFLAG_REALTIME)
1179 di_flags |= XFS_DIFLAG_REALTIME;
1180 if (xflags & FS_XFLAG_EXTSIZE)
1181 di_flags |= XFS_DIFLAG_EXTSIZE;
1189 struct xfs_inode *ip,
1190 unsigned int xflags)
1192 uint64_t di_flags2 =
1193 (ip->i_d.di_flags2 & (XFS_DIFLAG2_REFLINK |
1194 XFS_DIFLAG2_BIGTIME));
1196 if (xflags & FS_XFLAG_DAX)
1197 di_flags2 |= XFS_DIFLAG2_DAX;
1198 if (xflags & FS_XFLAG_COWEXTSIZE)
1199 di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
1205 xfs_ioctl_setattr_xflags(
1206 struct xfs_trans *tp,
1207 struct xfs_inode *ip,
1210 struct xfs_mount *mp = ip->i_mount;
1213 /* Can't change realtime flag if any extents are allocated. */
1214 if ((ip->i_df.if_nextents || ip->i_delayed_blks) &&
1215 XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1218 /* If realtime flag is set then must have realtime device */
1219 if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1220 if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1221 (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1225 /* Clear reflink if we are actually able to set the rt flag. */
1226 if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1227 ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
1229 /* Don't allow us to set DAX mode for a reflinked file for now. */
1230 if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1233 /* diflags2 only valid for v3 inodes. */
1234 di_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1235 if (di_flags2 && !xfs_sb_version_has_v3inode(&mp->m_sb))
1238 ip->i_d.di_flags = xfs_flags2diflags(ip, fa->fsx_xflags);
1239 ip->i_d.di_flags2 = di_flags2;
1241 xfs_diflags_to_iflags(ip, false);
1242 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1243 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1244 XFS_STATS_INC(mp, xs_ig_attrchg);
1249 xfs_ioctl_setattr_prepare_dax(
1250 struct xfs_inode *ip,
1253 struct xfs_mount *mp = ip->i_mount;
1254 struct inode *inode = VFS_I(ip);
1256 if (S_ISDIR(inode->i_mode))
1259 if ((mp->m_flags & XFS_MOUNT_DAX_ALWAYS) ||
1260 (mp->m_flags & XFS_MOUNT_DAX_NEVER))
1263 if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
1264 !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)) ||
1265 (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
1266 (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)))
1267 d_mark_dontcache(inode);
1271 * Set up the transaction structure for the setattr operation, checking that we
1272 * have permission to do so. On success, return a clean transaction and the
1273 * inode locked exclusively ready for further operation specific checks. On
1274 * failure, return an error without modifying or locking the inode.
1276 static struct xfs_trans *
1277 xfs_ioctl_setattr_get_trans(
1278 struct xfs_inode *ip)
1280 struct xfs_mount *mp = ip->i_mount;
1281 struct xfs_trans *tp;
1284 if (mp->m_flags & XFS_MOUNT_RDONLY)
1287 if (XFS_FORCED_SHUTDOWN(mp))
1290 error = xfs_trans_alloc(mp, &M_RES(mp)->tr_ichange, 0, 0, 0, &tp);
1294 xfs_ilock(ip, XFS_ILOCK_EXCL);
1295 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL);
1298 * CAP_FOWNER overrides the following restrictions:
1300 * The user ID of the calling process must be equal to the file owner
1301 * ID, except in cases where the CAP_FSETID capability is applicable.
1303 if (!inode_owner_or_capable(VFS_I(ip))) {
1308 if (mp->m_flags & XFS_MOUNT_WSYNC)
1309 xfs_trans_set_sync(tp);
1314 xfs_trans_cancel(tp);
1316 return ERR_PTR(error);
1320 * extent size hint validation is somewhat cumbersome. Rules are:
1322 * 1. extent size hint is only valid for directories and regular files
1323 * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1324 * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1325 * 4. can only be changed on regular files if no extents are allocated
1326 * 5. can be changed on directories at any time
1327 * 6. extsize hint of 0 turns off hints, clears inode flags.
1328 * 7. Extent size must be a multiple of the appropriate block size.
1329 * 8. for non-realtime files, the extent size hint must be limited
1330 * to half the AG size to avoid alignment extending the extent beyond the
1333 * Please keep this function in sync with xfs_scrub_inode_extsize.
1336 xfs_ioctl_setattr_check_extsize(
1337 struct xfs_inode *ip,
1340 struct xfs_mount *mp = ip->i_mount;
1342 xfs_fsblock_t extsize_fsb;
1344 if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
1345 ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1348 if (fa->fsx_extsize == 0)
1351 extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1352 if (extsize_fsb > MAXEXTLEN)
1355 if (XFS_IS_REALTIME_INODE(ip) ||
1356 (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1357 size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1359 size = mp->m_sb.sb_blocksize;
1360 if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1364 if (fa->fsx_extsize % size)
1371 * CoW extent size hint validation rules are:
1373 * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
1374 * The inode does not have to have any shared blocks, but it must be a v3.
1375 * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
1376 * for a directory, the hint is propagated to new files.
1377 * 3. Can be changed on files & directories at any time.
1378 * 4. CoW extsize hint of 0 turns off hints, clears inode flags.
1379 * 5. Extent size must be a multiple of the appropriate block size.
1380 * 6. The extent size hint must be limited to half the AG size to avoid
1381 * alignment extending the extent beyond the limits of the AG.
1383 * Please keep this function in sync with xfs_scrub_inode_cowextsize.
1386 xfs_ioctl_setattr_check_cowextsize(
1387 struct xfs_inode *ip,
1390 struct xfs_mount *mp = ip->i_mount;
1392 xfs_fsblock_t cowextsize_fsb;
1394 if (!(fa->fsx_xflags & FS_XFLAG_COWEXTSIZE))
1397 if (!xfs_sb_version_hasreflink(&ip->i_mount->m_sb))
1400 if (fa->fsx_cowextsize == 0)
1403 cowextsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1404 if (cowextsize_fsb > MAXEXTLEN)
1407 size = mp->m_sb.sb_blocksize;
1408 if (cowextsize_fsb > mp->m_sb.sb_agblocks / 2)
1411 if (fa->fsx_cowextsize % size)
1418 xfs_ioctl_setattr_check_projid(
1419 struct xfs_inode *ip,
1422 /* Disallow 32bit project ids if projid32bit feature is not enabled. */
1423 if (fa->fsx_projid > (uint16_t)-1 &&
1424 !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1434 struct fsxattr old_fa;
1435 struct xfs_mount *mp = ip->i_mount;
1436 struct xfs_trans *tp;
1437 struct xfs_dquot *pdqp = NULL;
1438 struct xfs_dquot *olddquot = NULL;
1441 trace_xfs_ioctl_setattr(ip);
1443 code = xfs_ioctl_setattr_check_projid(ip, fa);
1448 * If disk quotas is on, we make sure that the dquots do exist on disk,
1449 * before we start any other transactions. Trying to do this later
1450 * is messy. We don't care to take a readlock to look at the ids
1451 * in inode here, because we can't hold it across the trans_reserve.
1452 * If the IDs do change before we take the ilock, we're covered
1453 * because the i_*dquot fields will get updated anyway.
1455 if (XFS_IS_QUOTA_ON(mp)) {
1456 code = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
1457 VFS_I(ip)->i_gid, fa->fsx_projid,
1458 XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
1463 xfs_ioctl_setattr_prepare_dax(ip, fa);
1465 tp = xfs_ioctl_setattr_get_trans(ip);
1468 goto error_free_dquots;
1471 if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp) &&
1472 ip->i_d.di_projid != fa->fsx_projid) {
1473 code = xfs_qm_vop_chown_reserve(tp, ip, NULL, NULL, pdqp,
1474 capable(CAP_FOWNER) ? XFS_QMOPT_FORCE_RES : 0);
1475 if (code) /* out of quota */
1476 goto error_trans_cancel;
1479 xfs_fill_fsxattr(ip, false, &old_fa);
1480 code = vfs_ioc_fssetxattr_check(VFS_I(ip), &old_fa, fa);
1482 goto error_trans_cancel;
1484 code = xfs_ioctl_setattr_check_extsize(ip, fa);
1486 goto error_trans_cancel;
1488 code = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1490 goto error_trans_cancel;
1492 code = xfs_ioctl_setattr_xflags(tp, ip, fa);
1494 goto error_trans_cancel;
1497 * Change file ownership. Must be the owner or privileged. CAP_FSETID
1498 * overrides the following restrictions:
1500 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1501 * successful return from chown()
1504 if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1505 !capable_wrt_inode_uidgid(VFS_I(ip), CAP_FSETID))
1506 VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1508 /* Change the ownerships and register project quota modifications */
1509 if (ip->i_d.di_projid != fa->fsx_projid) {
1510 if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1511 olddquot = xfs_qm_vop_chown(tp, ip,
1512 &ip->i_pdquot, pdqp);
1514 ip->i_d.di_projid = fa->fsx_projid;
1518 * Only set the extent size hint if we've already determined that the
1519 * extent size hint should be set on the inode. If no extent size flags
1520 * are set on the inode then unconditionally clear the extent size hint.
1522 if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1523 ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1525 ip->i_d.di_extsize = 0;
1526 if (xfs_sb_version_has_v3inode(&mp->m_sb) &&
1527 (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
1528 ip->i_d.di_cowextsize = fa->fsx_cowextsize >>
1529 mp->m_sb.sb_blocklog;
1531 ip->i_d.di_cowextsize = 0;
1533 code = xfs_trans_commit(tp);
1536 * Release any dquot(s) the inode had kept before chown.
1538 xfs_qm_dqrele(olddquot);
1539 xfs_qm_dqrele(pdqp);
1544 xfs_trans_cancel(tp);
1546 xfs_qm_dqrele(pdqp);
1559 if (copy_from_user(&fa, arg, sizeof(fa)))
1562 error = mnt_want_write_file(filp);
1565 error = xfs_ioctl_setattr(ip, &fa);
1566 mnt_drop_write_file(filp);
1577 flags = xfs_di2lxflags(ip->i_d.di_flags, ip->i_d.di_flags2);
1578 if (copy_to_user(arg, &flags, sizeof(flags)))
1585 struct xfs_inode *ip,
1589 struct xfs_trans *tp;
1591 struct fsxattr old_fa;
1595 if (copy_from_user(&flags, arg, sizeof(flags)))
1598 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1599 FS_NOATIME_FL | FS_NODUMP_FL | \
1600 FS_SYNC_FL | FS_DAX_FL))
1603 fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1605 error = mnt_want_write_file(filp);
1609 xfs_ioctl_setattr_prepare_dax(ip, &fa);
1611 tp = xfs_ioctl_setattr_get_trans(ip);
1613 error = PTR_ERR(tp);
1614 goto out_drop_write;
1617 xfs_fill_fsxattr(ip, false, &old_fa);
1618 error = vfs_ioc_fssetxattr_check(VFS_I(ip), &old_fa, &fa);
1620 xfs_trans_cancel(tp);
1621 goto out_drop_write;
1624 error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1626 xfs_trans_cancel(tp);
1627 goto out_drop_write;
1630 error = xfs_trans_commit(tp);
1632 mnt_drop_write_file(filp);
1639 struct getbmapx __user *u,
1642 if (put_user(p->bmv_offset, &u->bmv_offset) ||
1643 put_user(p->bmv_block, &u->bmv_block) ||
1644 put_user(p->bmv_length, &u->bmv_length) ||
1645 put_user(0, &u->bmv_count) ||
1646 put_user(0, &u->bmv_entries))
1648 if (recsize < sizeof(struct getbmapx))
1650 if (put_user(0, &u->bmv_iflags) ||
1651 put_user(p->bmv_oflags, &u->bmv_oflags) ||
1652 put_user(0, &u->bmv_unused1) ||
1653 put_user(0, &u->bmv_unused2))
1664 struct getbmapx bmx = { 0 };
1665 struct kgetbmap *buf;
1670 case XFS_IOC_GETBMAPA:
1671 bmx.bmv_iflags = BMV_IF_ATTRFORK;
1673 case XFS_IOC_GETBMAP:
1674 if (file->f_mode & FMODE_NOCMTIME)
1675 bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1676 /* struct getbmap is a strict subset of struct getbmapx. */
1677 recsize = sizeof(struct getbmap);
1679 case XFS_IOC_GETBMAPX:
1680 recsize = sizeof(struct getbmapx);
1686 if (copy_from_user(&bmx, arg, recsize))
1689 if (bmx.bmv_count < 2)
1691 if (bmx.bmv_count > ULONG_MAX / recsize)
1694 buf = kvzalloc(bmx.bmv_count * sizeof(*buf), GFP_KERNEL);
1698 error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1703 if (copy_to_user(arg, &bmx, recsize))
1707 for (i = 0; i < bmx.bmv_entries; i++) {
1708 if (!xfs_getbmap_format(buf + i, arg, recsize))
1721 struct xfs_inode *ip,
1722 struct fsmap_head __user *arg)
1724 struct xfs_fsmap_head xhead = {0};
1725 struct fsmap_head head;
1728 __u32 last_flags = 0;
1732 if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1734 if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1735 memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1736 sizeof(head.fmh_keys[0].fmr_reserved)) ||
1737 memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1738 sizeof(head.fmh_keys[1].fmr_reserved)))
1742 * Use an internal memory buffer so that we don't have to copy fsmap
1743 * data to userspace while holding locks. Start by trying to allocate
1744 * up to 128k for the buffer, but fall back to a single page if needed.
1746 count = min_t(unsigned int, head.fmh_count,
1747 131072 / sizeof(struct fsmap));
1748 recs = kvzalloc(count * sizeof(struct fsmap), GFP_KERNEL);
1750 count = min_t(unsigned int, head.fmh_count,
1751 PAGE_SIZE / sizeof(struct fsmap));
1752 recs = kvzalloc(count * sizeof(struct fsmap), GFP_KERNEL);
1757 xhead.fmh_iflags = head.fmh_iflags;
1758 xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1759 xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1761 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1762 trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1764 head.fmh_entries = 0;
1766 struct fsmap __user *user_recs;
1767 struct fsmap *last_rec;
1769 user_recs = &arg->fmh_recs[head.fmh_entries];
1770 xhead.fmh_entries = 0;
1771 xhead.fmh_count = min_t(unsigned int, count,
1772 head.fmh_count - head.fmh_entries);
1774 /* Run query, record how many entries we got. */
1775 error = xfs_getfsmap(ip->i_mount, &xhead, recs);
1779 * There are no more records in the result set. Copy
1780 * whatever we got to userspace and break out.
1786 * The internal memory buffer is full. Copy whatever
1787 * records we got to userspace and go again if we have
1788 * not yet filled the userspace buffer.
1795 head.fmh_entries += xhead.fmh_entries;
1796 head.fmh_oflags = xhead.fmh_oflags;
1799 * If the caller wanted a record count or there aren't any
1800 * new records to return, we're done.
1802 if (head.fmh_count == 0 || xhead.fmh_entries == 0)
1805 /* Copy all the records we got out to userspace. */
1806 if (copy_to_user(user_recs, recs,
1807 xhead.fmh_entries * sizeof(struct fsmap))) {
1812 /* Remember the last record flags we copied to userspace. */
1813 last_rec = &recs[xhead.fmh_entries - 1];
1814 last_flags = last_rec->fmr_flags;
1816 /* Set up the low key for the next iteration. */
1817 xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
1818 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1819 } while (!done && head.fmh_entries < head.fmh_count);
1822 * If there are no more records in the query result set and we're not
1823 * in counting mode, mark the last record returned with the LAST flag.
1825 if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
1826 struct fsmap __user *user_rec;
1828 last_flags |= FMR_OF_LAST;
1829 user_rec = &arg->fmh_recs[head.fmh_entries - 1];
1831 if (copy_to_user(&user_rec->fmr_flags, &last_flags,
1832 sizeof(last_flags))) {
1838 /* copy back header */
1839 if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1850 xfs_ioc_scrub_metadata(
1851 struct xfs_inode *ip,
1854 struct xfs_scrub_metadata scrub;
1857 if (!capable(CAP_SYS_ADMIN))
1860 if (copy_from_user(&scrub, arg, sizeof(scrub)))
1863 error = xfs_scrub_metadata(ip, &scrub);
1867 if (copy_to_user(arg, &scrub, sizeof(scrub)))
1877 xfs_inode_t *ip, *tip;
1881 /* Pull information for the target fd */
1882 f = fdget((int)sxp->sx_fdtarget);
1888 if (!(f.file->f_mode & FMODE_WRITE) ||
1889 !(f.file->f_mode & FMODE_READ) ||
1890 (f.file->f_flags & O_APPEND)) {
1895 tmp = fdget((int)sxp->sx_fdtmp);
1901 if (!(tmp.file->f_mode & FMODE_WRITE) ||
1902 !(tmp.file->f_mode & FMODE_READ) ||
1903 (tmp.file->f_flags & O_APPEND)) {
1905 goto out_put_tmp_file;
1908 if (IS_SWAPFILE(file_inode(f.file)) ||
1909 IS_SWAPFILE(file_inode(tmp.file))) {
1911 goto out_put_tmp_file;
1915 * We need to ensure that the fds passed in point to XFS inodes
1916 * before we cast and access them as XFS structures as we have no
1917 * control over what the user passes us here.
1919 if (f.file->f_op != &xfs_file_operations ||
1920 tmp.file->f_op != &xfs_file_operations) {
1922 goto out_put_tmp_file;
1925 ip = XFS_I(file_inode(f.file));
1926 tip = XFS_I(file_inode(tmp.file));
1928 if (ip->i_mount != tip->i_mount) {
1930 goto out_put_tmp_file;
1933 if (ip->i_ino == tip->i_ino) {
1935 goto out_put_tmp_file;
1938 if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1940 goto out_put_tmp_file;
1943 error = xfs_swap_extents(ip, tip, sxp);
1955 struct xfs_mount *mp,
1956 char __user *user_label)
1958 struct xfs_sb *sbp = &mp->m_sb;
1959 char label[XFSLABEL_MAX + 1];
1962 BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1964 /* 1 larger than sb_fname, so this ensures a trailing NUL char */
1965 memset(label, 0, sizeof(label));
1966 spin_lock(&mp->m_sb_lock);
1967 strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1968 spin_unlock(&mp->m_sb_lock);
1970 if (copy_to_user(user_label, label, sizeof(label)))
1978 struct xfs_mount *mp,
1979 char __user *newlabel)
1981 struct xfs_sb *sbp = &mp->m_sb;
1982 char label[XFSLABEL_MAX + 1];
1986 if (!capable(CAP_SYS_ADMIN))
1989 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1990 * smaller, at 12 bytes. We copy one more to be sure we find the
1991 * (required) NULL character to test the incoming label length.
1992 * NB: The on disk label doesn't need to be null terminated.
1994 if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1996 len = strnlen(label, XFSLABEL_MAX + 1);
1997 if (len > sizeof(sbp->sb_fname))
2000 error = mnt_want_write_file(filp);
2004 spin_lock(&mp->m_sb_lock);
2005 memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
2006 memcpy(sbp->sb_fname, label, len);
2007 spin_unlock(&mp->m_sb_lock);
2010 * Now we do several things to satisfy userspace.
2011 * In addition to normal logging of the primary superblock, we also
2012 * immediately write these changes to sector zero for the primary, then
2013 * update all backup supers (as xfs_db does for a label change), then
2014 * invalidate the block device page cache. This is so that any prior
2015 * buffered reads from userspace (i.e. from blkid) are invalidated,
2016 * and userspace will see the newly-written label.
2018 error = xfs_sync_sb_buf(mp);
2022 * growfs also updates backup supers so lock against that.
2024 mutex_lock(&mp->m_growlock);
2025 error = xfs_update_secondary_sbs(mp);
2026 mutex_unlock(&mp->m_growlock);
2028 invalidate_bdev(mp->m_ddev_targp->bt_bdev);
2031 mnt_drop_write_file(filp);
2036 xfs_fs_eofblocks_from_user(
2037 struct xfs_fs_eofblocks *src,
2038 struct xfs_eofblocks *dst)
2040 if (src->eof_version != XFS_EOFBLOCKS_VERSION)
2043 if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
2046 if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
2047 memchr_inv(src->pad64, 0, sizeof(src->pad64)))
2050 dst->eof_flags = src->eof_flags;
2051 dst->eof_prid = src->eof_prid;
2052 dst->eof_min_file_size = src->eof_min_file_size;
2054 dst->eof_uid = INVALID_UID;
2055 if (src->eof_flags & XFS_EOF_FLAGS_UID) {
2056 dst->eof_uid = make_kuid(current_user_ns(), src->eof_uid);
2057 if (!uid_valid(dst->eof_uid))
2061 dst->eof_gid = INVALID_GID;
2062 if (src->eof_flags & XFS_EOF_FLAGS_GID) {
2063 dst->eof_gid = make_kgid(current_user_ns(), src->eof_gid);
2064 if (!gid_valid(dst->eof_gid))
2071 * Note: some of the ioctl's return positive numbers as a
2072 * byte count indicating success, such as readlink_by_handle.
2073 * So we don't "sign flip" like most other routines. This means
2074 * true errors need to be returned as a negative value.
2082 struct inode *inode = file_inode(filp);
2083 struct xfs_inode *ip = XFS_I(inode);
2084 struct xfs_mount *mp = ip->i_mount;
2085 void __user *arg = (void __user *)p;
2088 trace_xfs_file_ioctl(ip);
2092 return xfs_ioc_trim(mp, arg);
2093 case FS_IOC_GETFSLABEL:
2094 return xfs_ioc_getlabel(mp, arg);
2095 case FS_IOC_SETFSLABEL:
2096 return xfs_ioc_setlabel(filp, mp, arg);
2097 case XFS_IOC_ALLOCSP:
2098 case XFS_IOC_FREESP:
2099 case XFS_IOC_ALLOCSP64:
2100 case XFS_IOC_FREESP64: {
2103 if (copy_from_user(&bf, arg, sizeof(bf)))
2105 return xfs_ioc_space(filp, &bf);
2107 case XFS_IOC_DIOINFO: {
2108 struct xfs_buftarg *target = xfs_inode_buftarg(ip);
2111 da.d_mem = da.d_miniosz = target->bt_logical_sectorsize;
2112 da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
2114 if (copy_to_user(arg, &da, sizeof(da)))
2119 case XFS_IOC_FSBULKSTAT_SINGLE:
2120 case XFS_IOC_FSBULKSTAT:
2121 case XFS_IOC_FSINUMBERS:
2122 return xfs_ioc_fsbulkstat(mp, cmd, arg);
2124 case XFS_IOC_BULKSTAT:
2125 return xfs_ioc_bulkstat(mp, cmd, arg);
2126 case XFS_IOC_INUMBERS:
2127 return xfs_ioc_inumbers(mp, cmd, arg);
2129 case XFS_IOC_FSGEOMETRY_V1:
2130 return xfs_ioc_fsgeometry(mp, arg, 3);
2131 case XFS_IOC_FSGEOMETRY_V4:
2132 return xfs_ioc_fsgeometry(mp, arg, 4);
2133 case XFS_IOC_FSGEOMETRY:
2134 return xfs_ioc_fsgeometry(mp, arg, 5);
2136 case XFS_IOC_AG_GEOMETRY:
2137 return xfs_ioc_ag_geometry(mp, arg);
2139 case XFS_IOC_GETVERSION:
2140 return put_user(inode->i_generation, (int __user *)arg);
2142 case XFS_IOC_FSGETXATTR:
2143 return xfs_ioc_fsgetxattr(ip, 0, arg);
2144 case XFS_IOC_FSGETXATTRA:
2145 return xfs_ioc_fsgetxattr(ip, 1, arg);
2146 case XFS_IOC_FSSETXATTR:
2147 return xfs_ioc_fssetxattr(ip, filp, arg);
2148 case XFS_IOC_GETXFLAGS:
2149 return xfs_ioc_getxflags(ip, arg);
2150 case XFS_IOC_SETXFLAGS:
2151 return xfs_ioc_setxflags(ip, filp, arg);
2153 case XFS_IOC_GETBMAP:
2154 case XFS_IOC_GETBMAPA:
2155 case XFS_IOC_GETBMAPX:
2156 return xfs_ioc_getbmap(filp, cmd, arg);
2158 case FS_IOC_GETFSMAP:
2159 return xfs_ioc_getfsmap(ip, arg);
2161 case XFS_IOC_SCRUB_METADATA:
2162 return xfs_ioc_scrub_metadata(ip, arg);
2164 case XFS_IOC_FD_TO_HANDLE:
2165 case XFS_IOC_PATH_TO_HANDLE:
2166 case XFS_IOC_PATH_TO_FSHANDLE: {
2167 xfs_fsop_handlereq_t hreq;
2169 if (copy_from_user(&hreq, arg, sizeof(hreq)))
2171 return xfs_find_handle(cmd, &hreq);
2173 case XFS_IOC_OPEN_BY_HANDLE: {
2174 xfs_fsop_handlereq_t hreq;
2176 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2178 return xfs_open_by_handle(filp, &hreq);
2181 case XFS_IOC_READLINK_BY_HANDLE: {
2182 xfs_fsop_handlereq_t hreq;
2184 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2186 return xfs_readlink_by_handle(filp, &hreq);
2188 case XFS_IOC_ATTRLIST_BY_HANDLE:
2189 return xfs_attrlist_by_handle(filp, arg);
2191 case XFS_IOC_ATTRMULTI_BY_HANDLE:
2192 return xfs_attrmulti_by_handle(filp, arg);
2194 case XFS_IOC_SWAPEXT: {
2195 struct xfs_swapext sxp;
2197 if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
2199 error = mnt_want_write_file(filp);
2202 error = xfs_ioc_swapext(&sxp);
2203 mnt_drop_write_file(filp);
2207 case XFS_IOC_FSCOUNTS: {
2208 xfs_fsop_counts_t out;
2210 xfs_fs_counts(mp, &out);
2212 if (copy_to_user(arg, &out, sizeof(out)))
2217 case XFS_IOC_SET_RESBLKS: {
2218 xfs_fsop_resblks_t inout;
2221 if (!capable(CAP_SYS_ADMIN))
2224 if (mp->m_flags & XFS_MOUNT_RDONLY)
2227 if (copy_from_user(&inout, arg, sizeof(inout)))
2230 error = mnt_want_write_file(filp);
2234 /* input parameter is passed in resblks field of structure */
2236 error = xfs_reserve_blocks(mp, &in, &inout);
2237 mnt_drop_write_file(filp);
2241 if (copy_to_user(arg, &inout, sizeof(inout)))
2246 case XFS_IOC_GET_RESBLKS: {
2247 xfs_fsop_resblks_t out;
2249 if (!capable(CAP_SYS_ADMIN))
2252 error = xfs_reserve_blocks(mp, NULL, &out);
2256 if (copy_to_user(arg, &out, sizeof(out)))
2262 case XFS_IOC_FSGROWFSDATA: {
2263 xfs_growfs_data_t in;
2265 if (copy_from_user(&in, arg, sizeof(in)))
2268 error = mnt_want_write_file(filp);
2271 error = xfs_growfs_data(mp, &in);
2272 mnt_drop_write_file(filp);
2276 case XFS_IOC_FSGROWFSLOG: {
2277 xfs_growfs_log_t in;
2279 if (copy_from_user(&in, arg, sizeof(in)))
2282 error = mnt_want_write_file(filp);
2285 error = xfs_growfs_log(mp, &in);
2286 mnt_drop_write_file(filp);
2290 case XFS_IOC_FSGROWFSRT: {
2293 if (copy_from_user(&in, arg, sizeof(in)))
2296 error = mnt_want_write_file(filp);
2299 error = xfs_growfs_rt(mp, &in);
2300 mnt_drop_write_file(filp);
2304 case XFS_IOC_GOINGDOWN: {
2307 if (!capable(CAP_SYS_ADMIN))
2310 if (get_user(in, (uint32_t __user *)arg))
2313 return xfs_fs_goingdown(mp, in);
2316 case XFS_IOC_ERROR_INJECTION: {
2317 xfs_error_injection_t in;
2319 if (!capable(CAP_SYS_ADMIN))
2322 if (copy_from_user(&in, arg, sizeof(in)))
2325 return xfs_errortag_add(mp, in.errtag);
2328 case XFS_IOC_ERROR_CLEARALL:
2329 if (!capable(CAP_SYS_ADMIN))
2332 return xfs_errortag_clearall(mp);
2334 case XFS_IOC_FREE_EOFBLOCKS: {
2335 struct xfs_fs_eofblocks eofb;
2336 struct xfs_eofblocks keofb;
2338 if (!capable(CAP_SYS_ADMIN))
2341 if (mp->m_flags & XFS_MOUNT_RDONLY)
2344 if (copy_from_user(&eofb, arg, sizeof(eofb)))
2347 error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2351 sb_start_write(mp->m_super);
2352 error = xfs_icache_free_eofblocks(mp, &keofb);
2353 sb_end_write(mp->m_super);