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>
43 #include <linux/fileattr.h>
46 * xfs_find_handle maps from userspace xfs_fsop_handlereq structure to
47 * a file or fs handle.
49 * XFS_IOC_PATH_TO_FSHANDLE
50 * returns fs handle for a mount point or path within that mount point
51 * XFS_IOC_FD_TO_HANDLE
52 * returns full handle for a FD opened in user space
53 * XFS_IOC_PATH_TO_HANDLE
54 * returns full handle for a path
59 xfs_fsop_handlereq_t *hreq)
69 if (cmd == XFS_IOC_FD_TO_HANDLE) {
73 inode = file_inode(f.file);
75 error = user_path_at(AT_FDCWD, hreq->path, 0, &path);
78 inode = d_inode(path.dentry);
83 * We can only generate handles for inodes residing on a XFS filesystem,
84 * and only for regular files, directories or symbolic links.
87 if (inode->i_sb->s_magic != XFS_SB_MAGIC)
91 if (!S_ISREG(inode->i_mode) &&
92 !S_ISDIR(inode->i_mode) &&
93 !S_ISLNK(inode->i_mode))
97 memcpy(&handle.ha_fsid, ip->i_mount->m_fixedfsid, sizeof(xfs_fsid_t));
99 if (cmd == XFS_IOC_PATH_TO_FSHANDLE) {
101 * This handle only contains an fsid, zero the rest.
103 memset(&handle.ha_fid, 0, sizeof(handle.ha_fid));
104 hsize = sizeof(xfs_fsid_t);
106 handle.ha_fid.fid_len = sizeof(xfs_fid_t) -
107 sizeof(handle.ha_fid.fid_len);
108 handle.ha_fid.fid_pad = 0;
109 handle.ha_fid.fid_gen = inode->i_generation;
110 handle.ha_fid.fid_ino = ip->i_ino;
111 hsize = sizeof(xfs_handle_t);
115 if (copy_to_user(hreq->ohandle, &handle, hsize) ||
116 copy_to_user(hreq->ohandlen, &hsize, sizeof(__s32)))
122 if (cmd == XFS_IOC_FD_TO_HANDLE)
130 * No need to do permission checks on the various pathname components
131 * as the handle operations are privileged.
134 xfs_handle_acceptable(
136 struct dentry *dentry)
142 * Convert userspace handle data into a dentry.
145 xfs_handle_to_dentry(
146 struct file *parfilp,
147 void __user *uhandle,
151 struct xfs_fid64 fid;
154 * Only allow handle opens under a directory.
156 if (!S_ISDIR(file_inode(parfilp)->i_mode))
157 return ERR_PTR(-ENOTDIR);
159 if (hlen != sizeof(xfs_handle_t))
160 return ERR_PTR(-EINVAL);
161 if (copy_from_user(&handle, uhandle, hlen))
162 return ERR_PTR(-EFAULT);
163 if (handle.ha_fid.fid_len !=
164 sizeof(handle.ha_fid) - sizeof(handle.ha_fid.fid_len))
165 return ERR_PTR(-EINVAL);
167 memset(&fid, 0, sizeof(struct fid));
168 fid.ino = handle.ha_fid.fid_ino;
169 fid.gen = handle.ha_fid.fid_gen;
171 return exportfs_decode_fh(parfilp->f_path.mnt, (struct fid *)&fid, 3,
172 FILEID_INO32_GEN | XFS_FILEID_TYPE_64FLAG,
173 xfs_handle_acceptable, NULL);
176 STATIC struct dentry *
177 xfs_handlereq_to_dentry(
178 struct file *parfilp,
179 xfs_fsop_handlereq_t *hreq)
181 return xfs_handle_to_dentry(parfilp, hreq->ihandle, hreq->ihandlen);
186 struct file *parfilp,
187 xfs_fsop_handlereq_t *hreq)
189 const struct cred *cred = current_cred();
195 struct dentry *dentry;
199 if (!capable(CAP_SYS_ADMIN))
202 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
204 return PTR_ERR(dentry);
205 inode = d_inode(dentry);
207 /* Restrict xfs_open_by_handle to directories & regular files. */
208 if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode))) {
213 #if BITS_PER_LONG != 32
214 hreq->oflags |= O_LARGEFILE;
217 permflag = hreq->oflags;
218 fmode = OPEN_FMODE(permflag);
219 if ((!(permflag & O_APPEND) || (permflag & O_TRUNC)) &&
220 (fmode & FMODE_WRITE) && IS_APPEND(inode)) {
225 if ((fmode & FMODE_WRITE) && IS_IMMUTABLE(inode)) {
230 /* Can't write directories. */
231 if (S_ISDIR(inode->i_mode) && (fmode & FMODE_WRITE)) {
236 fd = get_unused_fd_flags(0);
242 path.mnt = parfilp->f_path.mnt;
243 path.dentry = dentry;
244 filp = dentry_open(&path, hreq->oflags, cred);
248 return PTR_ERR(filp);
251 if (S_ISREG(inode->i_mode)) {
252 filp->f_flags |= O_NOATIME;
253 filp->f_mode |= FMODE_NOCMTIME;
256 fd_install(fd, filp);
265 xfs_readlink_by_handle(
266 struct file *parfilp,
267 xfs_fsop_handlereq_t *hreq)
269 struct dentry *dentry;
273 if (!capable(CAP_SYS_ADMIN))
276 dentry = xfs_handlereq_to_dentry(parfilp, hreq);
278 return PTR_ERR(dentry);
280 /* Restrict this handle operation to symlinks only. */
281 if (!d_is_symlink(dentry)) {
286 if (copy_from_user(&olen, hreq->ohandlen, sizeof(__u32))) {
291 error = vfs_readlink(dentry, hreq->ohandle, olen);
299 * Format an attribute and copy it out to the user's buffer.
300 * Take care to check values and protect against them changing later,
301 * we may be reading them directly out of a user buffer.
304 xfs_ioc_attr_put_listent(
305 struct xfs_attr_list_context *context,
311 struct xfs_attrlist *alist = context->buffer;
312 struct xfs_attrlist_ent *aep;
315 ASSERT(!context->seen_enough);
316 ASSERT(context->count >= 0);
317 ASSERT(context->count < (ATTR_MAX_VALUELEN/8));
318 ASSERT(context->firstu >= sizeof(*alist));
319 ASSERT(context->firstu <= context->bufsize);
322 * Only list entries in the right namespace.
324 if (context->attr_filter != (flags & XFS_ATTR_NSP_ONDISK_MASK))
327 arraytop = sizeof(*alist) +
328 context->count * sizeof(alist->al_offset[0]);
330 /* decrement by the actual bytes used by the attr */
331 context->firstu -= round_up(offsetof(struct xfs_attrlist_ent, a_name) +
332 namelen + 1, sizeof(uint32_t));
333 if (context->firstu < arraytop) {
334 trace_xfs_attr_list_full(context);
336 context->seen_enough = 1;
340 aep = context->buffer + context->firstu;
341 aep->a_valuelen = valuelen;
342 memcpy(aep->a_name, name, namelen);
343 aep->a_name[namelen] = 0;
344 alist->al_offset[context->count++] = context->firstu;
345 alist->al_count = context->count;
346 trace_xfs_attr_list_add(context);
353 if (ioc_flags & XFS_IOC_ATTR_ROOT)
354 return XFS_ATTR_ROOT;
355 if (ioc_flags & XFS_IOC_ATTR_SECURE)
356 return XFS_ATTR_SECURE;
364 if (ioc_flags & XFS_IOC_ATTR_CREATE)
366 if (ioc_flags & XFS_IOC_ATTR_REPLACE)
367 return XATTR_REPLACE;
373 struct xfs_inode *dp,
377 struct xfs_attrlist_cursor __user *ucursor)
379 struct xfs_attr_list_context context = { };
380 struct xfs_attrlist *alist;
384 if (bufsize < sizeof(struct xfs_attrlist) ||
385 bufsize > XFS_XATTR_LIST_MAX)
389 * Reject flags, only allow namespaces.
391 if (flags & ~(XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
393 if (flags == (XFS_IOC_ATTR_ROOT | XFS_IOC_ATTR_SECURE))
397 * Validate the cursor.
399 if (copy_from_user(&context.cursor, ucursor, sizeof(context.cursor)))
401 if (context.cursor.pad1 || context.cursor.pad2)
403 if (!context.cursor.initted &&
404 (context.cursor.hashval || context.cursor.blkno ||
405 context.cursor.offset))
408 buffer = kvzalloc(bufsize, GFP_KERNEL);
413 * Initialize the output buffer.
417 context.attr_filter = xfs_attr_filter(flags);
418 context.buffer = buffer;
419 context.bufsize = round_down(bufsize, sizeof(uint32_t));
420 context.firstu = context.bufsize;
421 context.put_listent = xfs_ioc_attr_put_listent;
423 alist = context.buffer;
426 alist->al_offset[0] = context.bufsize;
428 error = xfs_attr_list(&context);
432 if (copy_to_user(ubuf, buffer, bufsize) ||
433 copy_to_user(ucursor, &context.cursor, sizeof(context.cursor)))
441 xfs_attrlist_by_handle(
442 struct file *parfilp,
443 struct xfs_fsop_attrlist_handlereq __user *p)
445 struct xfs_fsop_attrlist_handlereq al_hreq;
446 struct dentry *dentry;
449 if (!capable(CAP_SYS_ADMIN))
451 if (copy_from_user(&al_hreq, p, sizeof(al_hreq)))
454 dentry = xfs_handlereq_to_dentry(parfilp, &al_hreq.hreq);
456 return PTR_ERR(dentry);
458 error = xfs_ioc_attr_list(XFS_I(d_inode(dentry)), al_hreq.buffer,
459 al_hreq.buflen, al_hreq.flags, &p->pos);
465 xfs_attrmulti_attr_get(
468 unsigned char __user *ubuf,
472 struct xfs_da_args args = {
474 .attr_filter = xfs_attr_filter(flags),
475 .attr_flags = xfs_attr_flags(flags),
477 .namelen = strlen(name),
482 if (*len > XFS_XATTR_SIZE_MAX)
485 error = xfs_attr_get(&args);
489 *len = args.valuelen;
490 if (copy_to_user(ubuf, args.value, args.valuelen))
494 kmem_free(args.value);
499 xfs_attrmulti_attr_set(
502 const unsigned char __user *ubuf,
506 struct xfs_da_args args = {
508 .attr_filter = xfs_attr_filter(flags),
509 .attr_flags = xfs_attr_flags(flags),
511 .namelen = strlen(name),
515 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
519 if (len > XFS_XATTR_SIZE_MAX)
521 args.value = memdup_user(ubuf, len);
522 if (IS_ERR(args.value))
523 return PTR_ERR(args.value);
527 error = xfs_attr_set(&args);
528 if (!error && (flags & XFS_IOC_ATTR_ROOT))
529 xfs_forget_acl(inode, name);
535 xfs_ioc_attrmulti_one(
536 struct file *parfilp,
547 if ((flags & XFS_IOC_ATTR_ROOT) && (flags & XFS_IOC_ATTR_SECURE))
550 name = strndup_user(uname, MAXNAMELEN);
552 return PTR_ERR(name);
556 error = xfs_attrmulti_attr_get(inode, name, value, len, flags);
563 error = mnt_want_write_file(parfilp);
566 error = xfs_attrmulti_attr_set(inode, name, value, *len, flags);
567 mnt_drop_write_file(parfilp);
579 xfs_attrmulti_by_handle(
580 struct file *parfilp,
584 xfs_attr_multiop_t *ops;
585 xfs_fsop_attrmulti_handlereq_t am_hreq;
586 struct dentry *dentry;
587 unsigned int i, size;
589 if (!capable(CAP_SYS_ADMIN))
591 if (copy_from_user(&am_hreq, arg, sizeof(xfs_fsop_attrmulti_handlereq_t)))
595 if (am_hreq.opcount >= INT_MAX / sizeof(xfs_attr_multiop_t))
598 dentry = xfs_handlereq_to_dentry(parfilp, &am_hreq.hreq);
600 return PTR_ERR(dentry);
603 size = am_hreq.opcount * sizeof(xfs_attr_multiop_t);
604 if (!size || size > 16 * PAGE_SIZE)
607 ops = memdup_user(am_hreq.ops, size);
609 error = PTR_ERR(ops);
614 for (i = 0; i < am_hreq.opcount; i++) {
615 ops[i].am_error = xfs_ioc_attrmulti_one(parfilp,
616 d_inode(dentry), ops[i].am_opcode,
617 ops[i].am_attrname, ops[i].am_attrvalue,
618 &ops[i].am_length, ops[i].am_flags);
621 if (copy_to_user(am_hreq.ops, ops, size))
635 struct inode *inode = file_inode(filp);
636 struct xfs_inode *ip = XFS_I(inode);
638 enum xfs_prealloc_flags flags = XFS_PREALLOC_CLEAR;
639 uint iolock = XFS_IOLOCK_EXCL | XFS_MMAPLOCK_EXCL;
642 if (inode->i_flags & (S_IMMUTABLE|S_APPEND))
645 if (!(filp->f_mode & FMODE_WRITE))
648 if (!S_ISREG(inode->i_mode))
651 if (xfs_is_always_cow_inode(ip))
654 if (filp->f_flags & O_DSYNC)
655 flags |= XFS_PREALLOC_SYNC;
656 if (filp->f_mode & FMODE_NOCMTIME)
657 flags |= XFS_PREALLOC_INVISIBLE;
659 error = mnt_want_write_file(filp);
663 xfs_ilock(ip, iolock);
664 error = xfs_break_layouts(inode, &iolock, BREAK_UNMAP);
667 inode_dio_wait(inode);
669 switch (bf->l_whence) {
673 bf->l_start += filp->f_pos;
676 bf->l_start += XFS_ISIZE(ip);
683 if (bf->l_start < 0 || bf->l_start > inode->i_sb->s_maxbytes) {
688 if (bf->l_start > XFS_ISIZE(ip)) {
689 error = xfs_alloc_file_space(ip, XFS_ISIZE(ip),
690 bf->l_start - XFS_ISIZE(ip), 0);
695 iattr.ia_valid = ATTR_SIZE;
696 iattr.ia_size = bf->l_start;
697 error = xfs_vn_setattr_size(file_mnt_user_ns(filp), file_dentry(filp),
702 error = xfs_update_prealloc_flags(ip, flags);
705 xfs_iunlock(ip, iolock);
706 mnt_drop_write_file(filp);
710 /* Return 0 on success or positive error */
712 xfs_fsbulkstat_one_fmt(
713 struct xfs_ibulk *breq,
714 const struct xfs_bulkstat *bstat)
716 struct xfs_bstat bs1;
718 xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
719 if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
721 return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
726 struct xfs_ibulk *breq,
727 const struct xfs_inumbers *igrp)
729 struct xfs_inogrp ig1;
731 xfs_inumbers_to_inogrp(&ig1, igrp);
732 if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
734 return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
743 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
744 struct xfs_fsop_bulkreq bulkreq;
745 struct xfs_ibulk breq = {
747 .mnt_userns = file_mnt_user_ns(file),
753 /* done = 1 if there are more stats to get and if bulkstat */
754 /* should be called again (unused here, but used in dmapi) */
756 if (!capable(CAP_SYS_ADMIN))
759 if (xfs_is_shutdown(mp))
762 if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
765 if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
768 if (bulkreq.icount <= 0)
771 if (bulkreq.ubuffer == NULL)
774 breq.ubuffer = bulkreq.ubuffer;
775 breq.icount = bulkreq.icount;
778 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
779 * that we want to stat. However, FSINUMBERS and FSBULKSTAT expect
780 * that *lastip contains either zero or the number of the last inode to
781 * be examined by the previous call and return results starting with
782 * the next inode after that. The new bulk request back end functions
783 * take the inode to start with, so we have to compute the startino
784 * parameter from lastino to maintain correct function. lastino == 0
785 * is a special case because it has traditionally meant "first inode
788 if (cmd == XFS_IOC_FSINUMBERS) {
789 breq.startino = lastino ? lastino + 1 : 0;
790 error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
791 lastino = breq.startino - 1;
792 } else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
793 breq.startino = lastino;
795 error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
796 } else { /* XFS_IOC_FSBULKSTAT */
797 breq.startino = lastino ? lastino + 1 : 0;
798 error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
799 lastino = breq.startino - 1;
805 if (bulkreq.lastip != NULL &&
806 copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
809 if (bulkreq.ocount != NULL &&
810 copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
816 /* Return 0 on success or positive error */
819 struct xfs_ibulk *breq,
820 const struct xfs_bulkstat *bstat)
822 if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
824 return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
828 * Check the incoming bulk request @hdr from userspace and initialize the
829 * internal @breq bulk request appropriately. Returns 0 if the bulk request
830 * should proceed; -ECANCELED if there's nothing to do; or the usual
831 * negative error code.
835 struct xfs_mount *mp,
836 struct xfs_bulk_ireq *hdr,
837 struct xfs_ibulk *breq,
838 void __user *ubuffer)
840 if (hdr->icount == 0 ||
841 (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
842 memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
845 breq->startino = hdr->ino;
846 breq->ubuffer = ubuffer;
847 breq->icount = hdr->icount;
852 * The @ino parameter is a special value, so we must look it up here.
853 * We're not allowed to have IREQ_AGNO, and we only return one inode
856 if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
857 if (hdr->flags & XFS_BULK_IREQ_AGNO)
861 case XFS_BULK_IREQ_SPECIAL_ROOT:
862 hdr->ino = mp->m_sb.sb_rootino;
871 * The IREQ_AGNO flag means that we only want results from a given AG.
872 * If @hdr->ino is zero, we start iterating in that AG. If @hdr->ino is
873 * beyond the specified AG then we return no results.
875 if (hdr->flags & XFS_BULK_IREQ_AGNO) {
876 if (hdr->agno >= mp->m_sb.sb_agcount)
879 if (breq->startino == 0)
880 breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
881 else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
884 breq->flags |= XFS_IBULK_SAME_AG;
886 /* Asking for an inode past the end of the AG? We're done! */
887 if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
889 } else if (hdr->agno)
892 /* Asking for an inode past the end of the FS? We're done! */
893 if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
900 * Update the userspace bulk request @hdr to reflect the end state of the
901 * internal bulk request @breq.
904 xfs_bulk_ireq_teardown(
905 struct xfs_bulk_ireq *hdr,
906 struct xfs_ibulk *breq)
908 hdr->ino = breq->startino;
909 hdr->ocount = breq->ocount;
912 /* Handle the v5 bulkstat ioctl. */
917 struct xfs_bulkstat_req __user *arg)
919 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
920 struct xfs_bulk_ireq hdr;
921 struct xfs_ibulk breq = {
923 .mnt_userns = file_mnt_user_ns(file),
927 if (!capable(CAP_SYS_ADMIN))
930 if (xfs_is_shutdown(mp))
933 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
936 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
937 if (error == -ECANCELED)
942 error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
947 xfs_bulk_ireq_teardown(&hdr, &breq);
948 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
956 struct xfs_ibulk *breq,
957 const struct xfs_inumbers *igrp)
959 if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
961 return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
964 /* Handle the v5 inumbers ioctl. */
967 struct xfs_mount *mp,
969 struct xfs_inumbers_req __user *arg)
971 struct xfs_bulk_ireq hdr;
972 struct xfs_ibulk breq = {
977 if (!capable(CAP_SYS_ADMIN))
980 if (xfs_is_shutdown(mp))
983 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
986 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
987 if (error == -ECANCELED)
992 error = xfs_inumbers(&breq, xfs_inumbers_fmt);
997 xfs_bulk_ireq_teardown(&hdr, &breq);
998 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
1006 struct xfs_mount *mp,
1010 struct xfs_fsop_geom fsgeo;
1013 xfs_fs_geometry(mp, &fsgeo, struct_version);
1015 if (struct_version <= 3)
1016 len = sizeof(struct xfs_fsop_geom_v1);
1017 else if (struct_version == 4)
1018 len = sizeof(struct xfs_fsop_geom_v4);
1020 xfs_fsop_geom_health(mp, &fsgeo);
1021 len = sizeof(fsgeo);
1024 if (copy_to_user(arg, &fsgeo, len))
1030 xfs_ioc_ag_geometry(
1031 struct xfs_mount *mp,
1034 struct xfs_ag_geometry ageo;
1037 if (copy_from_user(&ageo, arg, sizeof(ageo)))
1041 if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
1044 error = xfs_ag_get_geometry(mp, ageo.ag_number, &ageo);
1048 if (copy_to_user(arg, &ageo, sizeof(ageo)))
1054 * Linux extended inode flags interface.
1059 struct xfs_inode *ip,
1061 struct fileattr *fa)
1063 struct xfs_mount *mp = ip->i_mount;
1064 struct xfs_ifork *ifp = XFS_IFORK_PTR(ip, whichfork);
1066 fileattr_fill_xflags(fa, xfs_ip2xflags(ip));
1068 if (ip->i_diflags & XFS_DIFLAG_EXTSIZE) {
1069 fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1070 } else if (ip->i_diflags & XFS_DIFLAG_EXTSZINHERIT) {
1072 * Don't let a misaligned extent size hint on a directory
1073 * escape to userspace if it won't pass the setattr checks
1076 if ((ip->i_diflags & XFS_DIFLAG_RTINHERIT) &&
1077 ip->i_extsize % mp->m_sb.sb_rextsize > 0) {
1078 fa->fsx_xflags &= ~(FS_XFLAG_EXTSIZE |
1079 FS_XFLAG_EXTSZINHERIT);
1080 fa->fsx_extsize = 0;
1082 fa->fsx_extsize = XFS_FSB_TO_B(mp, ip->i_extsize);
1086 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1087 fa->fsx_cowextsize = XFS_FSB_TO_B(mp, ip->i_cowextsize);
1088 fa->fsx_projid = ip->i_projid;
1089 if (ifp && !xfs_need_iread_extents(ifp))
1090 fa->fsx_nextents = xfs_iext_count(ifp);
1092 fa->fsx_nextents = xfs_ifork_nextents(ifp);
1096 xfs_ioc_fsgetxattra(
1102 xfs_ilock(ip, XFS_ILOCK_SHARED);
1103 xfs_fill_fsxattr(ip, XFS_ATTR_FORK, &fa);
1104 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1106 return copy_fsxattr_to_user(&fa, arg);
1111 struct dentry *dentry,
1112 struct fileattr *fa)
1114 struct xfs_inode *ip = XFS_I(d_inode(dentry));
1116 if (d_is_special(dentry))
1119 xfs_ilock(ip, XFS_ILOCK_SHARED);
1120 xfs_fill_fsxattr(ip, XFS_DATA_FORK, fa);
1121 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1128 struct xfs_inode *ip,
1129 unsigned int xflags)
1131 /* can't set PREALLOC this way, just preserve it */
1133 (ip->i_diflags & XFS_DIFLAG_PREALLOC);
1135 if (xflags & FS_XFLAG_IMMUTABLE)
1136 di_flags |= XFS_DIFLAG_IMMUTABLE;
1137 if (xflags & FS_XFLAG_APPEND)
1138 di_flags |= XFS_DIFLAG_APPEND;
1139 if (xflags & FS_XFLAG_SYNC)
1140 di_flags |= XFS_DIFLAG_SYNC;
1141 if (xflags & FS_XFLAG_NOATIME)
1142 di_flags |= XFS_DIFLAG_NOATIME;
1143 if (xflags & FS_XFLAG_NODUMP)
1144 di_flags |= XFS_DIFLAG_NODUMP;
1145 if (xflags & FS_XFLAG_NODEFRAG)
1146 di_flags |= XFS_DIFLAG_NODEFRAG;
1147 if (xflags & FS_XFLAG_FILESTREAM)
1148 di_flags |= XFS_DIFLAG_FILESTREAM;
1149 if (S_ISDIR(VFS_I(ip)->i_mode)) {
1150 if (xflags & FS_XFLAG_RTINHERIT)
1151 di_flags |= XFS_DIFLAG_RTINHERIT;
1152 if (xflags & FS_XFLAG_NOSYMLINKS)
1153 di_flags |= XFS_DIFLAG_NOSYMLINKS;
1154 if (xflags & FS_XFLAG_EXTSZINHERIT)
1155 di_flags |= XFS_DIFLAG_EXTSZINHERIT;
1156 if (xflags & FS_XFLAG_PROJINHERIT)
1157 di_flags |= XFS_DIFLAG_PROJINHERIT;
1158 } else if (S_ISREG(VFS_I(ip)->i_mode)) {
1159 if (xflags & FS_XFLAG_REALTIME)
1160 di_flags |= XFS_DIFLAG_REALTIME;
1161 if (xflags & FS_XFLAG_EXTSIZE)
1162 di_flags |= XFS_DIFLAG_EXTSIZE;
1170 struct xfs_inode *ip,
1171 unsigned int xflags)
1173 uint64_t di_flags2 =
1174 (ip->i_diflags2 & (XFS_DIFLAG2_REFLINK |
1175 XFS_DIFLAG2_BIGTIME));
1177 if (xflags & FS_XFLAG_DAX)
1178 di_flags2 |= XFS_DIFLAG2_DAX;
1179 if (xflags & FS_XFLAG_COWEXTSIZE)
1180 di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
1186 xfs_ioctl_setattr_xflags(
1187 struct xfs_trans *tp,
1188 struct xfs_inode *ip,
1189 struct fileattr *fa)
1191 struct xfs_mount *mp = ip->i_mount;
1194 /* Can't change realtime flag if any extents are allocated. */
1195 if ((ip->i_df.if_nextents || ip->i_delayed_blks) &&
1196 XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1199 /* If realtime flag is set then must have realtime device */
1200 if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1201 if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1202 (ip->i_extsize % mp->m_sb.sb_rextsize))
1206 /* Clear reflink if we are actually able to set the rt flag. */
1207 if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1208 ip->i_diflags2 &= ~XFS_DIFLAG2_REFLINK;
1210 /* Don't allow us to set DAX mode for a reflinked file for now. */
1211 if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1214 /* diflags2 only valid for v3 inodes. */
1215 i_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1216 if (i_flags2 && !xfs_has_v3inodes(mp))
1219 ip->i_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1220 ip->i_diflags2 = i_flags2;
1222 xfs_diflags_to_iflags(ip, false);
1223 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1224 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1225 XFS_STATS_INC(mp, xs_ig_attrchg);
1230 xfs_ioctl_setattr_prepare_dax(
1231 struct xfs_inode *ip,
1232 struct fileattr *fa)
1234 struct xfs_mount *mp = ip->i_mount;
1235 struct inode *inode = VFS_I(ip);
1237 if (S_ISDIR(inode->i_mode))
1240 if (xfs_has_dax_always(mp) || xfs_has_dax_never(mp))
1243 if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
1244 !(ip->i_diflags2 & XFS_DIFLAG2_DAX)) ||
1245 (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
1246 (ip->i_diflags2 & XFS_DIFLAG2_DAX)))
1247 d_mark_dontcache(inode);
1251 * Set up the transaction structure for the setattr operation, checking that we
1252 * have permission to do so. On success, return a clean transaction and the
1253 * inode locked exclusively ready for further operation specific checks. On
1254 * failure, return an error without modifying or locking the inode.
1256 static struct xfs_trans *
1257 xfs_ioctl_setattr_get_trans(
1258 struct xfs_inode *ip,
1259 struct xfs_dquot *pdqp)
1261 struct xfs_mount *mp = ip->i_mount;
1262 struct xfs_trans *tp;
1265 if (xfs_is_readonly(mp))
1268 if (xfs_is_shutdown(mp))
1271 error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
1272 capable(CAP_FOWNER), &tp);
1276 if (xfs_has_wsync(mp))
1277 xfs_trans_set_sync(tp);
1282 return ERR_PTR(error);
1286 * Validate a proposed extent size hint. For regular files, the hint can only
1287 * be changed if no extents are allocated.
1290 xfs_ioctl_setattr_check_extsize(
1291 struct xfs_inode *ip,
1292 struct fileattr *fa)
1294 struct xfs_mount *mp = ip->i_mount;
1295 xfs_failaddr_t failaddr;
1296 uint16_t new_diflags;
1301 if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
1302 XFS_FSB_TO_B(mp, ip->i_extsize) != fa->fsx_extsize)
1305 if (fa->fsx_extsize & mp->m_blockmask)
1308 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1311 * Inode verifiers do not check that the extent size hint is an integer
1312 * multiple of the rt extent size on a directory with both rtinherit
1313 * and extszinherit flags set. Don't let sysadmins misconfigure
1316 if ((new_diflags & XFS_DIFLAG_RTINHERIT) &&
1317 (new_diflags & XFS_DIFLAG_EXTSZINHERIT)) {
1318 unsigned int rtextsize_bytes;
1320 rtextsize_bytes = XFS_FSB_TO_B(mp, mp->m_sb.sb_rextsize);
1321 if (fa->fsx_extsize % rtextsize_bytes)
1325 failaddr = xfs_inode_validate_extsize(ip->i_mount,
1326 XFS_B_TO_FSB(mp, fa->fsx_extsize),
1327 VFS_I(ip)->i_mode, new_diflags);
1328 return failaddr != NULL ? -EINVAL : 0;
1332 xfs_ioctl_setattr_check_cowextsize(
1333 struct xfs_inode *ip,
1334 struct fileattr *fa)
1336 struct xfs_mount *mp = ip->i_mount;
1337 xfs_failaddr_t failaddr;
1338 uint64_t new_diflags2;
1339 uint16_t new_diflags;
1344 if (fa->fsx_cowextsize & mp->m_blockmask)
1347 new_diflags = xfs_flags2diflags(ip, fa->fsx_xflags);
1348 new_diflags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1350 failaddr = xfs_inode_validate_cowextsize(ip->i_mount,
1351 XFS_B_TO_FSB(mp, fa->fsx_cowextsize),
1352 VFS_I(ip)->i_mode, new_diflags, new_diflags2);
1353 return failaddr != NULL ? -EINVAL : 0;
1357 xfs_ioctl_setattr_check_projid(
1358 struct xfs_inode *ip,
1359 struct fileattr *fa)
1364 /* Disallow 32bit project ids if 32bit IDs are not enabled. */
1365 if (fa->fsx_projid > (uint16_t)-1 &&
1366 !xfs_has_projid32(ip->i_mount))
1373 struct user_namespace *mnt_userns,
1374 struct dentry *dentry,
1375 struct fileattr *fa)
1377 struct xfs_inode *ip = XFS_I(d_inode(dentry));
1378 struct xfs_mount *mp = ip->i_mount;
1379 struct xfs_trans *tp;
1380 struct xfs_dquot *pdqp = NULL;
1381 struct xfs_dquot *olddquot = NULL;
1384 trace_xfs_ioctl_setattr(ip);
1386 if (d_is_special(dentry))
1389 if (!fa->fsx_valid) {
1390 if (fa->flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL |
1391 FS_NOATIME_FL | FS_NODUMP_FL |
1392 FS_SYNC_FL | FS_DAX_FL | FS_PROJINHERIT_FL))
1396 error = xfs_ioctl_setattr_check_projid(ip, fa);
1401 * If disk quotas is on, we make sure that the dquots do exist on disk,
1402 * before we start any other transactions. Trying to do this later
1403 * is messy. We don't care to take a readlock to look at the ids
1404 * in inode here, because we can't hold it across the trans_reserve.
1405 * If the IDs do change before we take the ilock, we're covered
1406 * because the i_*dquot fields will get updated anyway.
1408 if (fa->fsx_valid && XFS_IS_QUOTA_ON(mp)) {
1409 error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
1410 VFS_I(ip)->i_gid, fa->fsx_projid,
1411 XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
1416 xfs_ioctl_setattr_prepare_dax(ip, fa);
1418 tp = xfs_ioctl_setattr_get_trans(ip, pdqp);
1420 error = PTR_ERR(tp);
1421 goto error_free_dquots;
1424 error = xfs_ioctl_setattr_check_extsize(ip, fa);
1426 goto error_trans_cancel;
1428 error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1430 goto error_trans_cancel;
1432 error = xfs_ioctl_setattr_xflags(tp, ip, fa);
1434 goto error_trans_cancel;
1439 * Change file ownership. Must be the owner or privileged. CAP_FSETID
1440 * overrides the following restrictions:
1442 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1443 * successful return from chown()
1446 if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1447 !capable_wrt_inode_uidgid(mnt_userns, VFS_I(ip), CAP_FSETID))
1448 VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1450 /* Change the ownerships and register project quota modifications */
1451 if (ip->i_projid != fa->fsx_projid) {
1452 if (XFS_IS_PQUOTA_ON(mp)) {
1453 olddquot = xfs_qm_vop_chown(tp, ip,
1454 &ip->i_pdquot, pdqp);
1456 ip->i_projid = fa->fsx_projid;
1460 * Only set the extent size hint if we've already determined that the
1461 * extent size hint should be set on the inode. If no extent size flags
1462 * are set on the inode then unconditionally clear the extent size hint.
1464 if (ip->i_diflags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1465 ip->i_extsize = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1469 if (xfs_has_v3inodes(mp)) {
1470 if (ip->i_diflags2 & XFS_DIFLAG2_COWEXTSIZE)
1471 ip->i_cowextsize = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1473 ip->i_cowextsize = 0;
1477 error = xfs_trans_commit(tp);
1480 * Release any dquot(s) the inode had kept before chown.
1482 xfs_qm_dqrele(olddquot);
1483 xfs_qm_dqrele(pdqp);
1488 xfs_trans_cancel(tp);
1490 xfs_qm_dqrele(pdqp);
1497 struct getbmapx __user *u,
1500 if (put_user(p->bmv_offset, &u->bmv_offset) ||
1501 put_user(p->bmv_block, &u->bmv_block) ||
1502 put_user(p->bmv_length, &u->bmv_length) ||
1503 put_user(0, &u->bmv_count) ||
1504 put_user(0, &u->bmv_entries))
1506 if (recsize < sizeof(struct getbmapx))
1508 if (put_user(0, &u->bmv_iflags) ||
1509 put_user(p->bmv_oflags, &u->bmv_oflags) ||
1510 put_user(0, &u->bmv_unused1) ||
1511 put_user(0, &u->bmv_unused2))
1522 struct getbmapx bmx = { 0 };
1523 struct kgetbmap *buf;
1528 case XFS_IOC_GETBMAPA:
1529 bmx.bmv_iflags = BMV_IF_ATTRFORK;
1531 case XFS_IOC_GETBMAP:
1532 /* struct getbmap is a strict subset of struct getbmapx. */
1533 recsize = sizeof(struct getbmap);
1535 case XFS_IOC_GETBMAPX:
1536 recsize = sizeof(struct getbmapx);
1542 if (copy_from_user(&bmx, arg, recsize))
1545 if (bmx.bmv_count < 2)
1547 if (bmx.bmv_count > ULONG_MAX / recsize)
1550 buf = kvzalloc(bmx.bmv_count * sizeof(*buf), GFP_KERNEL);
1554 error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1559 if (copy_to_user(arg, &bmx, recsize))
1563 for (i = 0; i < bmx.bmv_entries; i++) {
1564 if (!xfs_getbmap_format(buf + i, arg, recsize))
1577 struct xfs_inode *ip,
1578 struct fsmap_head __user *arg)
1580 struct xfs_fsmap_head xhead = {0};
1581 struct fsmap_head head;
1584 __u32 last_flags = 0;
1588 if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1590 if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1591 memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1592 sizeof(head.fmh_keys[0].fmr_reserved)) ||
1593 memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1594 sizeof(head.fmh_keys[1].fmr_reserved)))
1598 * Use an internal memory buffer so that we don't have to copy fsmap
1599 * data to userspace while holding locks. Start by trying to allocate
1600 * up to 128k for the buffer, but fall back to a single page if needed.
1602 count = min_t(unsigned int, head.fmh_count,
1603 131072 / sizeof(struct fsmap));
1604 recs = kvzalloc(count * sizeof(struct fsmap), GFP_KERNEL);
1606 count = min_t(unsigned int, head.fmh_count,
1607 PAGE_SIZE / sizeof(struct fsmap));
1608 recs = kvzalloc(count * sizeof(struct fsmap), GFP_KERNEL);
1613 xhead.fmh_iflags = head.fmh_iflags;
1614 xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1615 xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1617 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1618 trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1620 head.fmh_entries = 0;
1622 struct fsmap __user *user_recs;
1623 struct fsmap *last_rec;
1625 user_recs = &arg->fmh_recs[head.fmh_entries];
1626 xhead.fmh_entries = 0;
1627 xhead.fmh_count = min_t(unsigned int, count,
1628 head.fmh_count - head.fmh_entries);
1630 /* Run query, record how many entries we got. */
1631 error = xfs_getfsmap(ip->i_mount, &xhead, recs);
1635 * There are no more records in the result set. Copy
1636 * whatever we got to userspace and break out.
1642 * The internal memory buffer is full. Copy whatever
1643 * records we got to userspace and go again if we have
1644 * not yet filled the userspace buffer.
1651 head.fmh_entries += xhead.fmh_entries;
1652 head.fmh_oflags = xhead.fmh_oflags;
1655 * If the caller wanted a record count or there aren't any
1656 * new records to return, we're done.
1658 if (head.fmh_count == 0 || xhead.fmh_entries == 0)
1661 /* Copy all the records we got out to userspace. */
1662 if (copy_to_user(user_recs, recs,
1663 xhead.fmh_entries * sizeof(struct fsmap))) {
1668 /* Remember the last record flags we copied to userspace. */
1669 last_rec = &recs[xhead.fmh_entries - 1];
1670 last_flags = last_rec->fmr_flags;
1672 /* Set up the low key for the next iteration. */
1673 xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
1674 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1675 } while (!done && head.fmh_entries < head.fmh_count);
1678 * If there are no more records in the query result set and we're not
1679 * in counting mode, mark the last record returned with the LAST flag.
1681 if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
1682 struct fsmap __user *user_rec;
1684 last_flags |= FMR_OF_LAST;
1685 user_rec = &arg->fmh_recs[head.fmh_entries - 1];
1687 if (copy_to_user(&user_rec->fmr_flags, &last_flags,
1688 sizeof(last_flags))) {
1694 /* copy back header */
1695 if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1706 xfs_ioc_scrub_metadata(
1710 struct xfs_scrub_metadata scrub;
1713 if (!capable(CAP_SYS_ADMIN))
1716 if (copy_from_user(&scrub, arg, sizeof(scrub)))
1719 error = xfs_scrub_metadata(file, &scrub);
1723 if (copy_to_user(arg, &scrub, sizeof(scrub)))
1733 xfs_inode_t *ip, *tip;
1737 /* Pull information for the target fd */
1738 f = fdget((int)sxp->sx_fdtarget);
1744 if (!(f.file->f_mode & FMODE_WRITE) ||
1745 !(f.file->f_mode & FMODE_READ) ||
1746 (f.file->f_flags & O_APPEND)) {
1751 tmp = fdget((int)sxp->sx_fdtmp);
1757 if (!(tmp.file->f_mode & FMODE_WRITE) ||
1758 !(tmp.file->f_mode & FMODE_READ) ||
1759 (tmp.file->f_flags & O_APPEND)) {
1761 goto out_put_tmp_file;
1764 if (IS_SWAPFILE(file_inode(f.file)) ||
1765 IS_SWAPFILE(file_inode(tmp.file))) {
1767 goto out_put_tmp_file;
1771 * We need to ensure that the fds passed in point to XFS inodes
1772 * before we cast and access them as XFS structures as we have no
1773 * control over what the user passes us here.
1775 if (f.file->f_op != &xfs_file_operations ||
1776 tmp.file->f_op != &xfs_file_operations) {
1778 goto out_put_tmp_file;
1781 ip = XFS_I(file_inode(f.file));
1782 tip = XFS_I(file_inode(tmp.file));
1784 if (ip->i_mount != tip->i_mount) {
1786 goto out_put_tmp_file;
1789 if (ip->i_ino == tip->i_ino) {
1791 goto out_put_tmp_file;
1794 if (xfs_is_shutdown(ip->i_mount)) {
1796 goto out_put_tmp_file;
1799 error = xfs_swap_extents(ip, tip, sxp);
1811 struct xfs_mount *mp,
1812 char __user *user_label)
1814 struct xfs_sb *sbp = &mp->m_sb;
1815 char label[XFSLABEL_MAX + 1];
1818 BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1820 /* 1 larger than sb_fname, so this ensures a trailing NUL char */
1821 memset(label, 0, sizeof(label));
1822 spin_lock(&mp->m_sb_lock);
1823 strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1824 spin_unlock(&mp->m_sb_lock);
1826 if (copy_to_user(user_label, label, sizeof(label)))
1834 struct xfs_mount *mp,
1835 char __user *newlabel)
1837 struct xfs_sb *sbp = &mp->m_sb;
1838 char label[XFSLABEL_MAX + 1];
1842 if (!capable(CAP_SYS_ADMIN))
1845 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1846 * smaller, at 12 bytes. We copy one more to be sure we find the
1847 * (required) NULL character to test the incoming label length.
1848 * NB: The on disk label doesn't need to be null terminated.
1850 if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1852 len = strnlen(label, XFSLABEL_MAX + 1);
1853 if (len > sizeof(sbp->sb_fname))
1856 error = mnt_want_write_file(filp);
1860 spin_lock(&mp->m_sb_lock);
1861 memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
1862 memcpy(sbp->sb_fname, label, len);
1863 spin_unlock(&mp->m_sb_lock);
1866 * Now we do several things to satisfy userspace.
1867 * In addition to normal logging of the primary superblock, we also
1868 * immediately write these changes to sector zero for the primary, then
1869 * update all backup supers (as xfs_db does for a label change), then
1870 * invalidate the block device page cache. This is so that any prior
1871 * buffered reads from userspace (i.e. from blkid) are invalidated,
1872 * and userspace will see the newly-written label.
1874 error = xfs_sync_sb_buf(mp);
1878 * growfs also updates backup supers so lock against that.
1880 mutex_lock(&mp->m_growlock);
1881 error = xfs_update_secondary_sbs(mp);
1882 mutex_unlock(&mp->m_growlock);
1884 invalidate_bdev(mp->m_ddev_targp->bt_bdev);
1887 mnt_drop_write_file(filp);
1892 xfs_fs_eofblocks_from_user(
1893 struct xfs_fs_eofblocks *src,
1894 struct xfs_icwalk *dst)
1896 if (src->eof_version != XFS_EOFBLOCKS_VERSION)
1899 if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
1902 if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
1903 memchr_inv(src->pad64, 0, sizeof(src->pad64)))
1907 if (src->eof_flags & XFS_EOF_FLAGS_SYNC)
1908 dst->icw_flags |= XFS_ICWALK_FLAG_SYNC;
1909 if (src->eof_flags & XFS_EOF_FLAGS_UID)
1910 dst->icw_flags |= XFS_ICWALK_FLAG_UID;
1911 if (src->eof_flags & XFS_EOF_FLAGS_GID)
1912 dst->icw_flags |= XFS_ICWALK_FLAG_GID;
1913 if (src->eof_flags & XFS_EOF_FLAGS_PRID)
1914 dst->icw_flags |= XFS_ICWALK_FLAG_PRID;
1915 if (src->eof_flags & XFS_EOF_FLAGS_MINFILESIZE)
1916 dst->icw_flags |= XFS_ICWALK_FLAG_MINFILESIZE;
1918 dst->icw_prid = src->eof_prid;
1919 dst->icw_min_file_size = src->eof_min_file_size;
1921 dst->icw_uid = INVALID_UID;
1922 if (src->eof_flags & XFS_EOF_FLAGS_UID) {
1923 dst->icw_uid = make_kuid(current_user_ns(), src->eof_uid);
1924 if (!uid_valid(dst->icw_uid))
1928 dst->icw_gid = INVALID_GID;
1929 if (src->eof_flags & XFS_EOF_FLAGS_GID) {
1930 dst->icw_gid = make_kgid(current_user_ns(), src->eof_gid);
1931 if (!gid_valid(dst->icw_gid))
1938 * Note: some of the ioctl's return positive numbers as a
1939 * byte count indicating success, such as readlink_by_handle.
1940 * So we don't "sign flip" like most other routines. This means
1941 * true errors need to be returned as a negative value.
1949 struct inode *inode = file_inode(filp);
1950 struct xfs_inode *ip = XFS_I(inode);
1951 struct xfs_mount *mp = ip->i_mount;
1952 void __user *arg = (void __user *)p;
1955 trace_xfs_file_ioctl(ip);
1959 return xfs_ioc_trim(mp, arg);
1960 case FS_IOC_GETFSLABEL:
1961 return xfs_ioc_getlabel(mp, arg);
1962 case FS_IOC_SETFSLABEL:
1963 return xfs_ioc_setlabel(filp, mp, arg);
1964 case XFS_IOC_ALLOCSP:
1965 case XFS_IOC_FREESP:
1966 case XFS_IOC_ALLOCSP64:
1967 case XFS_IOC_FREESP64: {
1970 if (copy_from_user(&bf, arg, sizeof(bf)))
1972 return xfs_ioc_space(filp, &bf);
1974 case XFS_IOC_DIOINFO: {
1975 struct xfs_buftarg *target = xfs_inode_buftarg(ip);
1978 da.d_mem = da.d_miniosz = target->bt_logical_sectorsize;
1979 da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
1981 if (copy_to_user(arg, &da, sizeof(da)))
1986 case XFS_IOC_FSBULKSTAT_SINGLE:
1987 case XFS_IOC_FSBULKSTAT:
1988 case XFS_IOC_FSINUMBERS:
1989 return xfs_ioc_fsbulkstat(filp, cmd, arg);
1991 case XFS_IOC_BULKSTAT:
1992 return xfs_ioc_bulkstat(filp, cmd, arg);
1993 case XFS_IOC_INUMBERS:
1994 return xfs_ioc_inumbers(mp, cmd, arg);
1996 case XFS_IOC_FSGEOMETRY_V1:
1997 return xfs_ioc_fsgeometry(mp, arg, 3);
1998 case XFS_IOC_FSGEOMETRY_V4:
1999 return xfs_ioc_fsgeometry(mp, arg, 4);
2000 case XFS_IOC_FSGEOMETRY:
2001 return xfs_ioc_fsgeometry(mp, arg, 5);
2003 case XFS_IOC_AG_GEOMETRY:
2004 return xfs_ioc_ag_geometry(mp, arg);
2006 case XFS_IOC_GETVERSION:
2007 return put_user(inode->i_generation, (int __user *)arg);
2009 case XFS_IOC_FSGETXATTRA:
2010 return xfs_ioc_fsgetxattra(ip, arg);
2012 case XFS_IOC_GETBMAP:
2013 case XFS_IOC_GETBMAPA:
2014 case XFS_IOC_GETBMAPX:
2015 return xfs_ioc_getbmap(filp, cmd, arg);
2017 case FS_IOC_GETFSMAP:
2018 return xfs_ioc_getfsmap(ip, arg);
2020 case XFS_IOC_SCRUB_METADATA:
2021 return xfs_ioc_scrub_metadata(filp, arg);
2023 case XFS_IOC_FD_TO_HANDLE:
2024 case XFS_IOC_PATH_TO_HANDLE:
2025 case XFS_IOC_PATH_TO_FSHANDLE: {
2026 xfs_fsop_handlereq_t hreq;
2028 if (copy_from_user(&hreq, arg, sizeof(hreq)))
2030 return xfs_find_handle(cmd, &hreq);
2032 case XFS_IOC_OPEN_BY_HANDLE: {
2033 xfs_fsop_handlereq_t hreq;
2035 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2037 return xfs_open_by_handle(filp, &hreq);
2040 case XFS_IOC_READLINK_BY_HANDLE: {
2041 xfs_fsop_handlereq_t hreq;
2043 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2045 return xfs_readlink_by_handle(filp, &hreq);
2047 case XFS_IOC_ATTRLIST_BY_HANDLE:
2048 return xfs_attrlist_by_handle(filp, arg);
2050 case XFS_IOC_ATTRMULTI_BY_HANDLE:
2051 return xfs_attrmulti_by_handle(filp, arg);
2053 case XFS_IOC_SWAPEXT: {
2054 struct xfs_swapext sxp;
2056 if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
2058 error = mnt_want_write_file(filp);
2061 error = xfs_ioc_swapext(&sxp);
2062 mnt_drop_write_file(filp);
2066 case XFS_IOC_FSCOUNTS: {
2067 xfs_fsop_counts_t out;
2069 xfs_fs_counts(mp, &out);
2071 if (copy_to_user(arg, &out, sizeof(out)))
2076 case XFS_IOC_SET_RESBLKS: {
2077 xfs_fsop_resblks_t inout;
2080 if (!capable(CAP_SYS_ADMIN))
2083 if (xfs_is_readonly(mp))
2086 if (copy_from_user(&inout, arg, sizeof(inout)))
2089 error = mnt_want_write_file(filp);
2093 /* input parameter is passed in resblks field of structure */
2095 error = xfs_reserve_blocks(mp, &in, &inout);
2096 mnt_drop_write_file(filp);
2100 if (copy_to_user(arg, &inout, sizeof(inout)))
2105 case XFS_IOC_GET_RESBLKS: {
2106 xfs_fsop_resblks_t out;
2108 if (!capable(CAP_SYS_ADMIN))
2111 error = xfs_reserve_blocks(mp, NULL, &out);
2115 if (copy_to_user(arg, &out, sizeof(out)))
2121 case XFS_IOC_FSGROWFSDATA: {
2122 struct xfs_growfs_data in;
2124 if (copy_from_user(&in, arg, sizeof(in)))
2127 error = mnt_want_write_file(filp);
2130 error = xfs_growfs_data(mp, &in);
2131 mnt_drop_write_file(filp);
2135 case XFS_IOC_FSGROWFSLOG: {
2136 struct xfs_growfs_log in;
2138 if (copy_from_user(&in, arg, sizeof(in)))
2141 error = mnt_want_write_file(filp);
2144 error = xfs_growfs_log(mp, &in);
2145 mnt_drop_write_file(filp);
2149 case XFS_IOC_FSGROWFSRT: {
2152 if (copy_from_user(&in, arg, sizeof(in)))
2155 error = mnt_want_write_file(filp);
2158 error = xfs_growfs_rt(mp, &in);
2159 mnt_drop_write_file(filp);
2163 case XFS_IOC_GOINGDOWN: {
2166 if (!capable(CAP_SYS_ADMIN))
2169 if (get_user(in, (uint32_t __user *)arg))
2172 return xfs_fs_goingdown(mp, in);
2175 case XFS_IOC_ERROR_INJECTION: {
2176 xfs_error_injection_t in;
2178 if (!capable(CAP_SYS_ADMIN))
2181 if (copy_from_user(&in, arg, sizeof(in)))
2184 return xfs_errortag_add(mp, in.errtag);
2187 case XFS_IOC_ERROR_CLEARALL:
2188 if (!capable(CAP_SYS_ADMIN))
2191 return xfs_errortag_clearall(mp);
2193 case XFS_IOC_FREE_EOFBLOCKS: {
2194 struct xfs_fs_eofblocks eofb;
2195 struct xfs_icwalk icw;
2197 if (!capable(CAP_SYS_ADMIN))
2200 if (xfs_is_readonly(mp))
2203 if (copy_from_user(&eofb, arg, sizeof(eofb)))
2206 error = xfs_fs_eofblocks_from_user(&eofb, &icw);
2210 trace_xfs_ioc_free_eofblocks(mp, &icw, _RET_IP_);
2212 sb_start_write(mp->m_super);
2213 error = xfs_blockgc_free_space(mp, &icw);
2214 sb_end_write(mp->m_super);