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_mnt_user_ns(filp), file_dentry(filp),
701 error = xfs_update_prealloc_flags(ip, flags);
704 xfs_iunlock(ip, iolock);
705 mnt_drop_write_file(filp);
709 /* Return 0 on success or positive error */
711 xfs_fsbulkstat_one_fmt(
712 struct xfs_ibulk *breq,
713 const struct xfs_bulkstat *bstat)
715 struct xfs_bstat bs1;
717 xfs_bulkstat_to_bstat(breq->mp, &bs1, bstat);
718 if (copy_to_user(breq->ubuffer, &bs1, sizeof(bs1)))
720 return xfs_ibulk_advance(breq, sizeof(struct xfs_bstat));
725 struct xfs_ibulk *breq,
726 const struct xfs_inumbers *igrp)
728 struct xfs_inogrp ig1;
730 xfs_inumbers_to_inogrp(&ig1, igrp);
731 if (copy_to_user(breq->ubuffer, &ig1, sizeof(struct xfs_inogrp)))
733 return xfs_ibulk_advance(breq, sizeof(struct xfs_inogrp));
742 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
743 struct xfs_fsop_bulkreq bulkreq;
744 struct xfs_ibulk breq = {
746 .mnt_userns = file_mnt_user_ns(file),
752 /* done = 1 if there are more stats to get and if bulkstat */
753 /* should be called again (unused here, but used in dmapi) */
755 if (!capable(CAP_SYS_ADMIN))
758 if (XFS_FORCED_SHUTDOWN(mp))
761 if (copy_from_user(&bulkreq, arg, sizeof(struct xfs_fsop_bulkreq)))
764 if (copy_from_user(&lastino, bulkreq.lastip, sizeof(__s64)))
767 if (bulkreq.icount <= 0)
770 if (bulkreq.ubuffer == NULL)
773 breq.ubuffer = bulkreq.ubuffer;
774 breq.icount = bulkreq.icount;
777 * FSBULKSTAT_SINGLE expects that *lastip contains the inode number
778 * that we want to stat. However, FSINUMBERS and FSBULKSTAT expect
779 * that *lastip contains either zero or the number of the last inode to
780 * be examined by the previous call and return results starting with
781 * the next inode after that. The new bulk request back end functions
782 * take the inode to start with, so we have to compute the startino
783 * parameter from lastino to maintain correct function. lastino == 0
784 * is a special case because it has traditionally meant "first inode
787 if (cmd == XFS_IOC_FSINUMBERS) {
788 breq.startino = lastino ? lastino + 1 : 0;
789 error = xfs_inumbers(&breq, xfs_fsinumbers_fmt);
790 lastino = breq.startino - 1;
791 } else if (cmd == XFS_IOC_FSBULKSTAT_SINGLE) {
792 breq.startino = lastino;
794 error = xfs_bulkstat_one(&breq, xfs_fsbulkstat_one_fmt);
795 } else { /* XFS_IOC_FSBULKSTAT */
796 breq.startino = lastino ? lastino + 1 : 0;
797 error = xfs_bulkstat(&breq, xfs_fsbulkstat_one_fmt);
798 lastino = breq.startino - 1;
804 if (bulkreq.lastip != NULL &&
805 copy_to_user(bulkreq.lastip, &lastino, sizeof(xfs_ino_t)))
808 if (bulkreq.ocount != NULL &&
809 copy_to_user(bulkreq.ocount, &breq.ocount, sizeof(__s32)))
815 /* Return 0 on success or positive error */
818 struct xfs_ibulk *breq,
819 const struct xfs_bulkstat *bstat)
821 if (copy_to_user(breq->ubuffer, bstat, sizeof(struct xfs_bulkstat)))
823 return xfs_ibulk_advance(breq, sizeof(struct xfs_bulkstat));
827 * Check the incoming bulk request @hdr from userspace and initialize the
828 * internal @breq bulk request appropriately. Returns 0 if the bulk request
829 * should proceed; -ECANCELED if there's nothing to do; or the usual
830 * negative error code.
834 struct xfs_mount *mp,
835 struct xfs_bulk_ireq *hdr,
836 struct xfs_ibulk *breq,
837 void __user *ubuffer)
839 if (hdr->icount == 0 ||
840 (hdr->flags & ~XFS_BULK_IREQ_FLAGS_ALL) ||
841 memchr_inv(hdr->reserved, 0, sizeof(hdr->reserved)))
844 breq->startino = hdr->ino;
845 breq->ubuffer = ubuffer;
846 breq->icount = hdr->icount;
851 * The @ino parameter is a special value, so we must look it up here.
852 * We're not allowed to have IREQ_AGNO, and we only return one inode
855 if (hdr->flags & XFS_BULK_IREQ_SPECIAL) {
856 if (hdr->flags & XFS_BULK_IREQ_AGNO)
860 case XFS_BULK_IREQ_SPECIAL_ROOT:
861 hdr->ino = mp->m_sb.sb_rootino;
870 * The IREQ_AGNO flag means that we only want results from a given AG.
871 * If @hdr->ino is zero, we start iterating in that AG. If @hdr->ino is
872 * beyond the specified AG then we return no results.
874 if (hdr->flags & XFS_BULK_IREQ_AGNO) {
875 if (hdr->agno >= mp->m_sb.sb_agcount)
878 if (breq->startino == 0)
879 breq->startino = XFS_AGINO_TO_INO(mp, hdr->agno, 0);
880 else if (XFS_INO_TO_AGNO(mp, breq->startino) < hdr->agno)
883 breq->flags |= XFS_IBULK_SAME_AG;
885 /* Asking for an inode past the end of the AG? We're done! */
886 if (XFS_INO_TO_AGNO(mp, breq->startino) > hdr->agno)
888 } else if (hdr->agno)
891 /* Asking for an inode past the end of the FS? We're done! */
892 if (XFS_INO_TO_AGNO(mp, breq->startino) >= mp->m_sb.sb_agcount)
899 * Update the userspace bulk request @hdr to reflect the end state of the
900 * internal bulk request @breq.
903 xfs_bulk_ireq_teardown(
904 struct xfs_bulk_ireq *hdr,
905 struct xfs_ibulk *breq)
907 hdr->ino = breq->startino;
908 hdr->ocount = breq->ocount;
911 /* Handle the v5 bulkstat ioctl. */
916 struct xfs_bulkstat_req __user *arg)
918 struct xfs_mount *mp = XFS_I(file_inode(file))->i_mount;
919 struct xfs_bulk_ireq hdr;
920 struct xfs_ibulk breq = {
922 .mnt_userns = file_mnt_user_ns(file),
926 if (!capable(CAP_SYS_ADMIN))
929 if (XFS_FORCED_SHUTDOWN(mp))
932 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
935 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->bulkstat);
936 if (error == -ECANCELED)
941 error = xfs_bulkstat(&breq, xfs_bulkstat_fmt);
946 xfs_bulk_ireq_teardown(&hdr, &breq);
947 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
955 struct xfs_ibulk *breq,
956 const struct xfs_inumbers *igrp)
958 if (copy_to_user(breq->ubuffer, igrp, sizeof(struct xfs_inumbers)))
960 return xfs_ibulk_advance(breq, sizeof(struct xfs_inumbers));
963 /* Handle the v5 inumbers ioctl. */
966 struct xfs_mount *mp,
968 struct xfs_inumbers_req __user *arg)
970 struct xfs_bulk_ireq hdr;
971 struct xfs_ibulk breq = {
976 if (!capable(CAP_SYS_ADMIN))
979 if (XFS_FORCED_SHUTDOWN(mp))
982 if (copy_from_user(&hdr, &arg->hdr, sizeof(hdr)))
985 error = xfs_bulk_ireq_setup(mp, &hdr, &breq, arg->inumbers);
986 if (error == -ECANCELED)
991 error = xfs_inumbers(&breq, xfs_inumbers_fmt);
996 xfs_bulk_ireq_teardown(&hdr, &breq);
997 if (copy_to_user(&arg->hdr, &hdr, sizeof(hdr)))
1005 struct xfs_mount *mp,
1009 struct xfs_fsop_geom fsgeo;
1012 xfs_fs_geometry(&mp->m_sb, &fsgeo, struct_version);
1014 if (struct_version <= 3)
1015 len = sizeof(struct xfs_fsop_geom_v1);
1016 else if (struct_version == 4)
1017 len = sizeof(struct xfs_fsop_geom_v4);
1019 xfs_fsop_geom_health(mp, &fsgeo);
1020 len = sizeof(fsgeo);
1023 if (copy_to_user(arg, &fsgeo, len))
1029 xfs_ioc_ag_geometry(
1030 struct xfs_mount *mp,
1033 struct xfs_ag_geometry ageo;
1036 if (copy_from_user(&ageo, arg, sizeof(ageo)))
1040 if (memchr_inv(&ageo.ag_reserved, 0, sizeof(ageo.ag_reserved)))
1043 error = xfs_ag_get_geometry(mp, ageo.ag_number, &ageo);
1047 if (copy_to_user(arg, &ageo, sizeof(ageo)))
1053 * Linux extended inode flags interface.
1057 xfs_merge_ioc_xflags(
1061 unsigned int xflags = start;
1063 if (flags & FS_IMMUTABLE_FL)
1064 xflags |= FS_XFLAG_IMMUTABLE;
1066 xflags &= ~FS_XFLAG_IMMUTABLE;
1067 if (flags & FS_APPEND_FL)
1068 xflags |= FS_XFLAG_APPEND;
1070 xflags &= ~FS_XFLAG_APPEND;
1071 if (flags & FS_SYNC_FL)
1072 xflags |= FS_XFLAG_SYNC;
1074 xflags &= ~FS_XFLAG_SYNC;
1075 if (flags & FS_NOATIME_FL)
1076 xflags |= FS_XFLAG_NOATIME;
1078 xflags &= ~FS_XFLAG_NOATIME;
1079 if (flags & FS_NODUMP_FL)
1080 xflags |= FS_XFLAG_NODUMP;
1082 xflags &= ~FS_XFLAG_NODUMP;
1083 if (flags & FS_DAX_FL)
1084 xflags |= FS_XFLAG_DAX;
1086 xflags &= ~FS_XFLAG_DAX;
1096 unsigned int flags = 0;
1098 if (di_flags & XFS_DIFLAG_IMMUTABLE)
1099 flags |= FS_IMMUTABLE_FL;
1100 if (di_flags & XFS_DIFLAG_APPEND)
1101 flags |= FS_APPEND_FL;
1102 if (di_flags & XFS_DIFLAG_SYNC)
1103 flags |= FS_SYNC_FL;
1104 if (di_flags & XFS_DIFLAG_NOATIME)
1105 flags |= FS_NOATIME_FL;
1106 if (di_flags & XFS_DIFLAG_NODUMP)
1107 flags |= FS_NODUMP_FL;
1108 if (di_flags2 & XFS_DIFLAG2_DAX) {
1116 struct xfs_inode *ip,
1120 struct xfs_ifork *ifp = attr ? ip->i_afp : &ip->i_df;
1122 simple_fill_fsxattr(fa, xfs_ip2xflags(ip));
1123 fa->fsx_extsize = ip->i_d.di_extsize << ip->i_mount->m_sb.sb_blocklog;
1124 fa->fsx_cowextsize = ip->i_d.di_cowextsize <<
1125 ip->i_mount->m_sb.sb_blocklog;
1126 fa->fsx_projid = ip->i_d.di_projid;
1127 if (ifp && (ifp->if_flags & XFS_IFEXTENTS))
1128 fa->fsx_nextents = xfs_iext_count(ifp);
1130 fa->fsx_nextents = xfs_ifork_nextents(ifp);
1141 xfs_ilock(ip, XFS_ILOCK_SHARED);
1142 xfs_fill_fsxattr(ip, attr, &fa);
1143 xfs_iunlock(ip, XFS_ILOCK_SHARED);
1145 if (copy_to_user(arg, &fa, sizeof(fa)))
1152 struct xfs_inode *ip,
1153 unsigned int xflags)
1155 /* can't set PREALLOC this way, just preserve it */
1157 (ip->i_d.di_flags & XFS_DIFLAG_PREALLOC);
1159 if (xflags & FS_XFLAG_IMMUTABLE)
1160 di_flags |= XFS_DIFLAG_IMMUTABLE;
1161 if (xflags & FS_XFLAG_APPEND)
1162 di_flags |= XFS_DIFLAG_APPEND;
1163 if (xflags & FS_XFLAG_SYNC)
1164 di_flags |= XFS_DIFLAG_SYNC;
1165 if (xflags & FS_XFLAG_NOATIME)
1166 di_flags |= XFS_DIFLAG_NOATIME;
1167 if (xflags & FS_XFLAG_NODUMP)
1168 di_flags |= XFS_DIFLAG_NODUMP;
1169 if (xflags & FS_XFLAG_NODEFRAG)
1170 di_flags |= XFS_DIFLAG_NODEFRAG;
1171 if (xflags & FS_XFLAG_FILESTREAM)
1172 di_flags |= XFS_DIFLAG_FILESTREAM;
1173 if (S_ISDIR(VFS_I(ip)->i_mode)) {
1174 if (xflags & FS_XFLAG_RTINHERIT)
1175 di_flags |= XFS_DIFLAG_RTINHERIT;
1176 if (xflags & FS_XFLAG_NOSYMLINKS)
1177 di_flags |= XFS_DIFLAG_NOSYMLINKS;
1178 if (xflags & FS_XFLAG_EXTSZINHERIT)
1179 di_flags |= XFS_DIFLAG_EXTSZINHERIT;
1180 if (xflags & FS_XFLAG_PROJINHERIT)
1181 di_flags |= XFS_DIFLAG_PROJINHERIT;
1182 } else if (S_ISREG(VFS_I(ip)->i_mode)) {
1183 if (xflags & FS_XFLAG_REALTIME)
1184 di_flags |= XFS_DIFLAG_REALTIME;
1185 if (xflags & FS_XFLAG_EXTSIZE)
1186 di_flags |= XFS_DIFLAG_EXTSIZE;
1194 struct xfs_inode *ip,
1195 unsigned int xflags)
1197 uint64_t di_flags2 =
1198 (ip->i_d.di_flags2 & (XFS_DIFLAG2_REFLINK |
1199 XFS_DIFLAG2_BIGTIME));
1201 if (xflags & FS_XFLAG_DAX)
1202 di_flags2 |= XFS_DIFLAG2_DAX;
1203 if (xflags & FS_XFLAG_COWEXTSIZE)
1204 di_flags2 |= XFS_DIFLAG2_COWEXTSIZE;
1210 xfs_ioctl_setattr_xflags(
1211 struct xfs_trans *tp,
1212 struct xfs_inode *ip,
1215 struct xfs_mount *mp = ip->i_mount;
1218 /* Can't change realtime flag if any extents are allocated. */
1219 if ((ip->i_df.if_nextents || ip->i_delayed_blks) &&
1220 XFS_IS_REALTIME_INODE(ip) != (fa->fsx_xflags & FS_XFLAG_REALTIME))
1223 /* If realtime flag is set then must have realtime device */
1224 if (fa->fsx_xflags & FS_XFLAG_REALTIME) {
1225 if (mp->m_sb.sb_rblocks == 0 || mp->m_sb.sb_rextsize == 0 ||
1226 (ip->i_d.di_extsize % mp->m_sb.sb_rextsize))
1230 /* Clear reflink if we are actually able to set the rt flag. */
1231 if ((fa->fsx_xflags & FS_XFLAG_REALTIME) && xfs_is_reflink_inode(ip))
1232 ip->i_d.di_flags2 &= ~XFS_DIFLAG2_REFLINK;
1234 /* Don't allow us to set DAX mode for a reflinked file for now. */
1235 if ((fa->fsx_xflags & FS_XFLAG_DAX) && xfs_is_reflink_inode(ip))
1238 /* diflags2 only valid for v3 inodes. */
1239 di_flags2 = xfs_flags2diflags2(ip, fa->fsx_xflags);
1240 if (di_flags2 && !xfs_sb_version_has_v3inode(&mp->m_sb))
1243 ip->i_d.di_flags = xfs_flags2diflags(ip, fa->fsx_xflags);
1244 ip->i_d.di_flags2 = di_flags2;
1246 xfs_diflags_to_iflags(ip, false);
1247 xfs_trans_ichgtime(tp, ip, XFS_ICHGTIME_CHG);
1248 xfs_trans_log_inode(tp, ip, XFS_ILOG_CORE);
1249 XFS_STATS_INC(mp, xs_ig_attrchg);
1254 xfs_ioctl_setattr_prepare_dax(
1255 struct xfs_inode *ip,
1258 struct xfs_mount *mp = ip->i_mount;
1259 struct inode *inode = VFS_I(ip);
1261 if (S_ISDIR(inode->i_mode))
1264 if ((mp->m_flags & XFS_MOUNT_DAX_ALWAYS) ||
1265 (mp->m_flags & XFS_MOUNT_DAX_NEVER))
1268 if (((fa->fsx_xflags & FS_XFLAG_DAX) &&
1269 !(ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)) ||
1270 (!(fa->fsx_xflags & FS_XFLAG_DAX) &&
1271 (ip->i_d.di_flags2 & XFS_DIFLAG2_DAX)))
1272 d_mark_dontcache(inode);
1276 * Set up the transaction structure for the setattr operation, checking that we
1277 * have permission to do so. On success, return a clean transaction and the
1278 * inode locked exclusively ready for further operation specific checks. On
1279 * failure, return an error without modifying or locking the inode.
1281 static struct xfs_trans *
1282 xfs_ioctl_setattr_get_trans(
1284 struct xfs_dquot *pdqp)
1286 struct xfs_inode *ip = XFS_I(file_inode(file));
1287 struct xfs_mount *mp = ip->i_mount;
1288 struct xfs_trans *tp;
1291 if (mp->m_flags & XFS_MOUNT_RDONLY)
1294 if (XFS_FORCED_SHUTDOWN(mp))
1297 error = xfs_trans_alloc_ichange(ip, NULL, NULL, pdqp,
1298 capable(CAP_FOWNER), &tp);
1303 * CAP_FOWNER overrides the following restrictions:
1305 * The user ID of the calling process must be equal to the file owner
1306 * ID, except in cases where the CAP_FSETID capability is applicable.
1308 if (!inode_owner_or_capable(file_mnt_user_ns(file), VFS_I(ip))) {
1313 if (mp->m_flags & XFS_MOUNT_WSYNC)
1314 xfs_trans_set_sync(tp);
1319 xfs_trans_cancel(tp);
1321 return ERR_PTR(error);
1325 * extent size hint validation is somewhat cumbersome. Rules are:
1327 * 1. extent size hint is only valid for directories and regular files
1328 * 2. FS_XFLAG_EXTSIZE is only valid for regular files
1329 * 3. FS_XFLAG_EXTSZINHERIT is only valid for directories.
1330 * 4. can only be changed on regular files if no extents are allocated
1331 * 5. can be changed on directories at any time
1332 * 6. extsize hint of 0 turns off hints, clears inode flags.
1333 * 7. Extent size must be a multiple of the appropriate block size.
1334 * 8. for non-realtime files, the extent size hint must be limited
1335 * to half the AG size to avoid alignment extending the extent beyond the
1338 * Please keep this function in sync with xfs_scrub_inode_extsize.
1341 xfs_ioctl_setattr_check_extsize(
1342 struct xfs_inode *ip,
1345 struct xfs_mount *mp = ip->i_mount;
1347 xfs_fsblock_t extsize_fsb;
1349 if (S_ISREG(VFS_I(ip)->i_mode) && ip->i_df.if_nextents &&
1350 ((ip->i_d.di_extsize << mp->m_sb.sb_blocklog) != fa->fsx_extsize))
1353 if (fa->fsx_extsize == 0)
1356 extsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_extsize);
1357 if (extsize_fsb > MAXEXTLEN)
1360 if (XFS_IS_REALTIME_INODE(ip) ||
1361 (fa->fsx_xflags & FS_XFLAG_REALTIME)) {
1362 size = mp->m_sb.sb_rextsize << mp->m_sb.sb_blocklog;
1364 size = mp->m_sb.sb_blocksize;
1365 if (extsize_fsb > mp->m_sb.sb_agblocks / 2)
1369 if (fa->fsx_extsize % size)
1376 * CoW extent size hint validation rules are:
1378 * 1. CoW extent size hint can only be set if reflink is enabled on the fs.
1379 * The inode does not have to have any shared blocks, but it must be a v3.
1380 * 2. FS_XFLAG_COWEXTSIZE is only valid for directories and regular files;
1381 * for a directory, the hint is propagated to new files.
1382 * 3. Can be changed on files & directories at any time.
1383 * 4. CoW extsize hint of 0 turns off hints, clears inode flags.
1384 * 5. Extent size must be a multiple of the appropriate block size.
1385 * 6. The extent size hint must be limited to half the AG size to avoid
1386 * alignment extending the extent beyond the limits of the AG.
1388 * Please keep this function in sync with xfs_scrub_inode_cowextsize.
1391 xfs_ioctl_setattr_check_cowextsize(
1392 struct xfs_inode *ip,
1395 struct xfs_mount *mp = ip->i_mount;
1397 xfs_fsblock_t cowextsize_fsb;
1399 if (!(fa->fsx_xflags & FS_XFLAG_COWEXTSIZE))
1402 if (!xfs_sb_version_hasreflink(&ip->i_mount->m_sb))
1405 if (fa->fsx_cowextsize == 0)
1408 cowextsize_fsb = XFS_B_TO_FSB(mp, fa->fsx_cowextsize);
1409 if (cowextsize_fsb > MAXEXTLEN)
1412 size = mp->m_sb.sb_blocksize;
1413 if (cowextsize_fsb > mp->m_sb.sb_agblocks / 2)
1416 if (fa->fsx_cowextsize % size)
1423 xfs_ioctl_setattr_check_projid(
1424 struct xfs_inode *ip,
1427 /* Disallow 32bit project ids if projid32bit feature is not enabled. */
1428 if (fa->fsx_projid > (uint16_t)-1 &&
1429 !xfs_sb_version_hasprojid32bit(&ip->i_mount->m_sb))
1439 struct user_namespace *mnt_userns = file_mnt_user_ns(file);
1440 struct xfs_inode *ip = XFS_I(file_inode(file));
1441 struct fsxattr old_fa;
1442 struct xfs_mount *mp = ip->i_mount;
1443 struct xfs_trans *tp;
1444 struct xfs_dquot *pdqp = NULL;
1445 struct xfs_dquot *olddquot = NULL;
1448 trace_xfs_ioctl_setattr(ip);
1450 error = xfs_ioctl_setattr_check_projid(ip, fa);
1455 * If disk quotas is on, we make sure that the dquots do exist on disk,
1456 * before we start any other transactions. Trying to do this later
1457 * is messy. We don't care to take a readlock to look at the ids
1458 * in inode here, because we can't hold it across the trans_reserve.
1459 * If the IDs do change before we take the ilock, we're covered
1460 * because the i_*dquot fields will get updated anyway.
1462 if (XFS_IS_QUOTA_ON(mp)) {
1463 error = xfs_qm_vop_dqalloc(ip, VFS_I(ip)->i_uid,
1464 VFS_I(ip)->i_gid, fa->fsx_projid,
1465 XFS_QMOPT_PQUOTA, NULL, NULL, &pdqp);
1470 xfs_ioctl_setattr_prepare_dax(ip, fa);
1472 tp = xfs_ioctl_setattr_get_trans(file, pdqp);
1474 error = PTR_ERR(tp);
1475 goto error_free_dquots;
1478 xfs_fill_fsxattr(ip, false, &old_fa);
1479 error = vfs_ioc_fssetxattr_check(VFS_I(ip), &old_fa, fa);
1481 goto error_trans_cancel;
1483 error = xfs_ioctl_setattr_check_extsize(ip, fa);
1485 goto error_trans_cancel;
1487 error = xfs_ioctl_setattr_check_cowextsize(ip, fa);
1489 goto error_trans_cancel;
1491 error = xfs_ioctl_setattr_xflags(tp, ip, fa);
1493 goto error_trans_cancel;
1496 * Change file ownership. Must be the owner or privileged. CAP_FSETID
1497 * overrides the following restrictions:
1499 * The set-user-ID and set-group-ID bits of a file will be cleared upon
1500 * successful return from chown()
1503 if ((VFS_I(ip)->i_mode & (S_ISUID|S_ISGID)) &&
1504 !capable_wrt_inode_uidgid(mnt_userns, VFS_I(ip), CAP_FSETID))
1505 VFS_I(ip)->i_mode &= ~(S_ISUID|S_ISGID);
1507 /* Change the ownerships and register project quota modifications */
1508 if (ip->i_d.di_projid != fa->fsx_projid) {
1509 if (XFS_IS_QUOTA_RUNNING(mp) && XFS_IS_PQUOTA_ON(mp)) {
1510 olddquot = xfs_qm_vop_chown(tp, ip,
1511 &ip->i_pdquot, pdqp);
1513 ip->i_d.di_projid = fa->fsx_projid;
1517 * Only set the extent size hint if we've already determined that the
1518 * extent size hint should be set on the inode. If no extent size flags
1519 * are set on the inode then unconditionally clear the extent size hint.
1521 if (ip->i_d.di_flags & (XFS_DIFLAG_EXTSIZE | XFS_DIFLAG_EXTSZINHERIT))
1522 ip->i_d.di_extsize = fa->fsx_extsize >> mp->m_sb.sb_blocklog;
1524 ip->i_d.di_extsize = 0;
1525 if (xfs_sb_version_has_v3inode(&mp->m_sb) &&
1526 (ip->i_d.di_flags2 & XFS_DIFLAG2_COWEXTSIZE))
1527 ip->i_d.di_cowextsize = fa->fsx_cowextsize >>
1528 mp->m_sb.sb_blocklog;
1530 ip->i_d.di_cowextsize = 0;
1532 error = xfs_trans_commit(tp);
1535 * Release any dquot(s) the inode had kept before chown.
1537 xfs_qm_dqrele(olddquot);
1538 xfs_qm_dqrele(pdqp);
1543 xfs_trans_cancel(tp);
1545 xfs_qm_dqrele(pdqp);
1557 if (copy_from_user(&fa, arg, sizeof(fa)))
1560 error = mnt_want_write_file(filp);
1563 error = xfs_ioctl_setattr(filp, &fa);
1564 mnt_drop_write_file(filp);
1575 flags = xfs_di2lxflags(ip->i_d.di_flags, ip->i_d.di_flags2);
1576 if (copy_to_user(arg, &flags, sizeof(flags)))
1583 struct xfs_inode *ip,
1587 struct xfs_trans *tp;
1589 struct fsxattr old_fa;
1593 if (copy_from_user(&flags, arg, sizeof(flags)))
1596 if (flags & ~(FS_IMMUTABLE_FL | FS_APPEND_FL | \
1597 FS_NOATIME_FL | FS_NODUMP_FL | \
1598 FS_SYNC_FL | FS_DAX_FL))
1601 fa.fsx_xflags = xfs_merge_ioc_xflags(flags, xfs_ip2xflags(ip));
1603 error = mnt_want_write_file(filp);
1607 xfs_ioctl_setattr_prepare_dax(ip, &fa);
1609 tp = xfs_ioctl_setattr_get_trans(filp, NULL);
1611 error = PTR_ERR(tp);
1612 goto out_drop_write;
1615 xfs_fill_fsxattr(ip, false, &old_fa);
1616 error = vfs_ioc_fssetxattr_check(VFS_I(ip), &old_fa, &fa);
1618 xfs_trans_cancel(tp);
1619 goto out_drop_write;
1622 error = xfs_ioctl_setattr_xflags(tp, ip, &fa);
1624 xfs_trans_cancel(tp);
1625 goto out_drop_write;
1628 error = xfs_trans_commit(tp);
1630 mnt_drop_write_file(filp);
1637 struct getbmapx __user *u,
1640 if (put_user(p->bmv_offset, &u->bmv_offset) ||
1641 put_user(p->bmv_block, &u->bmv_block) ||
1642 put_user(p->bmv_length, &u->bmv_length) ||
1643 put_user(0, &u->bmv_count) ||
1644 put_user(0, &u->bmv_entries))
1646 if (recsize < sizeof(struct getbmapx))
1648 if (put_user(0, &u->bmv_iflags) ||
1649 put_user(p->bmv_oflags, &u->bmv_oflags) ||
1650 put_user(0, &u->bmv_unused1) ||
1651 put_user(0, &u->bmv_unused2))
1662 struct getbmapx bmx = { 0 };
1663 struct kgetbmap *buf;
1668 case XFS_IOC_GETBMAPA:
1669 bmx.bmv_iflags = BMV_IF_ATTRFORK;
1671 case XFS_IOC_GETBMAP:
1672 if (file->f_mode & FMODE_NOCMTIME)
1673 bmx.bmv_iflags |= BMV_IF_NO_DMAPI_READ;
1674 /* struct getbmap is a strict subset of struct getbmapx. */
1675 recsize = sizeof(struct getbmap);
1677 case XFS_IOC_GETBMAPX:
1678 recsize = sizeof(struct getbmapx);
1684 if (copy_from_user(&bmx, arg, recsize))
1687 if (bmx.bmv_count < 2)
1689 if (bmx.bmv_count > ULONG_MAX / recsize)
1692 buf = kvzalloc(bmx.bmv_count * sizeof(*buf), GFP_KERNEL);
1696 error = xfs_getbmap(XFS_I(file_inode(file)), &bmx, buf);
1701 if (copy_to_user(arg, &bmx, recsize))
1705 for (i = 0; i < bmx.bmv_entries; i++) {
1706 if (!xfs_getbmap_format(buf + i, arg, recsize))
1719 struct xfs_inode *ip,
1720 struct fsmap_head __user *arg)
1722 struct xfs_fsmap_head xhead = {0};
1723 struct fsmap_head head;
1726 __u32 last_flags = 0;
1730 if (copy_from_user(&head, arg, sizeof(struct fsmap_head)))
1732 if (memchr_inv(head.fmh_reserved, 0, sizeof(head.fmh_reserved)) ||
1733 memchr_inv(head.fmh_keys[0].fmr_reserved, 0,
1734 sizeof(head.fmh_keys[0].fmr_reserved)) ||
1735 memchr_inv(head.fmh_keys[1].fmr_reserved, 0,
1736 sizeof(head.fmh_keys[1].fmr_reserved)))
1740 * Use an internal memory buffer so that we don't have to copy fsmap
1741 * data to userspace while holding locks. Start by trying to allocate
1742 * up to 128k for the buffer, but fall back to a single page if needed.
1744 count = min_t(unsigned int, head.fmh_count,
1745 131072 / sizeof(struct fsmap));
1746 recs = kvzalloc(count * sizeof(struct fsmap), GFP_KERNEL);
1748 count = min_t(unsigned int, head.fmh_count,
1749 PAGE_SIZE / sizeof(struct fsmap));
1750 recs = kvzalloc(count * sizeof(struct fsmap), GFP_KERNEL);
1755 xhead.fmh_iflags = head.fmh_iflags;
1756 xfs_fsmap_to_internal(&xhead.fmh_keys[0], &head.fmh_keys[0]);
1757 xfs_fsmap_to_internal(&xhead.fmh_keys[1], &head.fmh_keys[1]);
1759 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1760 trace_xfs_getfsmap_high_key(ip->i_mount, &xhead.fmh_keys[1]);
1762 head.fmh_entries = 0;
1764 struct fsmap __user *user_recs;
1765 struct fsmap *last_rec;
1767 user_recs = &arg->fmh_recs[head.fmh_entries];
1768 xhead.fmh_entries = 0;
1769 xhead.fmh_count = min_t(unsigned int, count,
1770 head.fmh_count - head.fmh_entries);
1772 /* Run query, record how many entries we got. */
1773 error = xfs_getfsmap(ip->i_mount, &xhead, recs);
1777 * There are no more records in the result set. Copy
1778 * whatever we got to userspace and break out.
1784 * The internal memory buffer is full. Copy whatever
1785 * records we got to userspace and go again if we have
1786 * not yet filled the userspace buffer.
1793 head.fmh_entries += xhead.fmh_entries;
1794 head.fmh_oflags = xhead.fmh_oflags;
1797 * If the caller wanted a record count or there aren't any
1798 * new records to return, we're done.
1800 if (head.fmh_count == 0 || xhead.fmh_entries == 0)
1803 /* Copy all the records we got out to userspace. */
1804 if (copy_to_user(user_recs, recs,
1805 xhead.fmh_entries * sizeof(struct fsmap))) {
1810 /* Remember the last record flags we copied to userspace. */
1811 last_rec = &recs[xhead.fmh_entries - 1];
1812 last_flags = last_rec->fmr_flags;
1814 /* Set up the low key for the next iteration. */
1815 xfs_fsmap_to_internal(&xhead.fmh_keys[0], last_rec);
1816 trace_xfs_getfsmap_low_key(ip->i_mount, &xhead.fmh_keys[0]);
1817 } while (!done && head.fmh_entries < head.fmh_count);
1820 * If there are no more records in the query result set and we're not
1821 * in counting mode, mark the last record returned with the LAST flag.
1823 if (done && head.fmh_count > 0 && head.fmh_entries > 0) {
1824 struct fsmap __user *user_rec;
1826 last_flags |= FMR_OF_LAST;
1827 user_rec = &arg->fmh_recs[head.fmh_entries - 1];
1829 if (copy_to_user(&user_rec->fmr_flags, &last_flags,
1830 sizeof(last_flags))) {
1836 /* copy back header */
1837 if (copy_to_user(arg, &head, sizeof(struct fsmap_head))) {
1848 xfs_ioc_scrub_metadata(
1849 struct xfs_inode *ip,
1852 struct xfs_scrub_metadata scrub;
1855 if (!capable(CAP_SYS_ADMIN))
1858 if (copy_from_user(&scrub, arg, sizeof(scrub)))
1861 error = xfs_scrub_metadata(ip, &scrub);
1865 if (copy_to_user(arg, &scrub, sizeof(scrub)))
1875 xfs_inode_t *ip, *tip;
1879 /* Pull information for the target fd */
1880 f = fdget((int)sxp->sx_fdtarget);
1886 if (!(f.file->f_mode & FMODE_WRITE) ||
1887 !(f.file->f_mode & FMODE_READ) ||
1888 (f.file->f_flags & O_APPEND)) {
1893 tmp = fdget((int)sxp->sx_fdtmp);
1899 if (!(tmp.file->f_mode & FMODE_WRITE) ||
1900 !(tmp.file->f_mode & FMODE_READ) ||
1901 (tmp.file->f_flags & O_APPEND)) {
1903 goto out_put_tmp_file;
1906 if (IS_SWAPFILE(file_inode(f.file)) ||
1907 IS_SWAPFILE(file_inode(tmp.file))) {
1909 goto out_put_tmp_file;
1913 * We need to ensure that the fds passed in point to XFS inodes
1914 * before we cast and access them as XFS structures as we have no
1915 * control over what the user passes us here.
1917 if (f.file->f_op != &xfs_file_operations ||
1918 tmp.file->f_op != &xfs_file_operations) {
1920 goto out_put_tmp_file;
1923 ip = XFS_I(file_inode(f.file));
1924 tip = XFS_I(file_inode(tmp.file));
1926 if (ip->i_mount != tip->i_mount) {
1928 goto out_put_tmp_file;
1931 if (ip->i_ino == tip->i_ino) {
1933 goto out_put_tmp_file;
1936 if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
1938 goto out_put_tmp_file;
1941 error = xfs_swap_extents(ip, tip, sxp);
1953 struct xfs_mount *mp,
1954 char __user *user_label)
1956 struct xfs_sb *sbp = &mp->m_sb;
1957 char label[XFSLABEL_MAX + 1];
1960 BUILD_BUG_ON(sizeof(sbp->sb_fname) > FSLABEL_MAX);
1962 /* 1 larger than sb_fname, so this ensures a trailing NUL char */
1963 memset(label, 0, sizeof(label));
1964 spin_lock(&mp->m_sb_lock);
1965 strncpy(label, sbp->sb_fname, XFSLABEL_MAX);
1966 spin_unlock(&mp->m_sb_lock);
1968 if (copy_to_user(user_label, label, sizeof(label)))
1976 struct xfs_mount *mp,
1977 char __user *newlabel)
1979 struct xfs_sb *sbp = &mp->m_sb;
1980 char label[XFSLABEL_MAX + 1];
1984 if (!capable(CAP_SYS_ADMIN))
1987 * The generic ioctl allows up to FSLABEL_MAX chars, but XFS is much
1988 * smaller, at 12 bytes. We copy one more to be sure we find the
1989 * (required) NULL character to test the incoming label length.
1990 * NB: The on disk label doesn't need to be null terminated.
1992 if (copy_from_user(label, newlabel, XFSLABEL_MAX + 1))
1994 len = strnlen(label, XFSLABEL_MAX + 1);
1995 if (len > sizeof(sbp->sb_fname))
1998 error = mnt_want_write_file(filp);
2002 spin_lock(&mp->m_sb_lock);
2003 memset(sbp->sb_fname, 0, sizeof(sbp->sb_fname));
2004 memcpy(sbp->sb_fname, label, len);
2005 spin_unlock(&mp->m_sb_lock);
2008 * Now we do several things to satisfy userspace.
2009 * In addition to normal logging of the primary superblock, we also
2010 * immediately write these changes to sector zero for the primary, then
2011 * update all backup supers (as xfs_db does for a label change), then
2012 * invalidate the block device page cache. This is so that any prior
2013 * buffered reads from userspace (i.e. from blkid) are invalidated,
2014 * and userspace will see the newly-written label.
2016 error = xfs_sync_sb_buf(mp);
2020 * growfs also updates backup supers so lock against that.
2022 mutex_lock(&mp->m_growlock);
2023 error = xfs_update_secondary_sbs(mp);
2024 mutex_unlock(&mp->m_growlock);
2026 invalidate_bdev(mp->m_ddev_targp->bt_bdev);
2029 mnt_drop_write_file(filp);
2034 xfs_fs_eofblocks_from_user(
2035 struct xfs_fs_eofblocks *src,
2036 struct xfs_eofblocks *dst)
2038 if (src->eof_version != XFS_EOFBLOCKS_VERSION)
2041 if (src->eof_flags & ~XFS_EOF_FLAGS_VALID)
2044 if (memchr_inv(&src->pad32, 0, sizeof(src->pad32)) ||
2045 memchr_inv(src->pad64, 0, sizeof(src->pad64)))
2048 dst->eof_flags = src->eof_flags;
2049 dst->eof_prid = src->eof_prid;
2050 dst->eof_min_file_size = src->eof_min_file_size;
2052 dst->eof_uid = INVALID_UID;
2053 if (src->eof_flags & XFS_EOF_FLAGS_UID) {
2054 dst->eof_uid = make_kuid(current_user_ns(), src->eof_uid);
2055 if (!uid_valid(dst->eof_uid))
2059 dst->eof_gid = INVALID_GID;
2060 if (src->eof_flags & XFS_EOF_FLAGS_GID) {
2061 dst->eof_gid = make_kgid(current_user_ns(), src->eof_gid);
2062 if (!gid_valid(dst->eof_gid))
2069 * Note: some of the ioctl's return positive numbers as a
2070 * byte count indicating success, such as readlink_by_handle.
2071 * So we don't "sign flip" like most other routines. This means
2072 * true errors need to be returned as a negative value.
2080 struct inode *inode = file_inode(filp);
2081 struct xfs_inode *ip = XFS_I(inode);
2082 struct xfs_mount *mp = ip->i_mount;
2083 void __user *arg = (void __user *)p;
2086 trace_xfs_file_ioctl(ip);
2090 return xfs_ioc_trim(mp, arg);
2091 case FS_IOC_GETFSLABEL:
2092 return xfs_ioc_getlabel(mp, arg);
2093 case FS_IOC_SETFSLABEL:
2094 return xfs_ioc_setlabel(filp, mp, arg);
2095 case XFS_IOC_ALLOCSP:
2096 case XFS_IOC_FREESP:
2097 case XFS_IOC_ALLOCSP64:
2098 case XFS_IOC_FREESP64: {
2101 if (copy_from_user(&bf, arg, sizeof(bf)))
2103 return xfs_ioc_space(filp, &bf);
2105 case XFS_IOC_DIOINFO: {
2106 struct xfs_buftarg *target = xfs_inode_buftarg(ip);
2109 da.d_mem = da.d_miniosz = target->bt_logical_sectorsize;
2110 da.d_maxiosz = INT_MAX & ~(da.d_miniosz - 1);
2112 if (copy_to_user(arg, &da, sizeof(da)))
2117 case XFS_IOC_FSBULKSTAT_SINGLE:
2118 case XFS_IOC_FSBULKSTAT:
2119 case XFS_IOC_FSINUMBERS:
2120 return xfs_ioc_fsbulkstat(filp, cmd, arg);
2122 case XFS_IOC_BULKSTAT:
2123 return xfs_ioc_bulkstat(filp, cmd, arg);
2124 case XFS_IOC_INUMBERS:
2125 return xfs_ioc_inumbers(mp, cmd, arg);
2127 case XFS_IOC_FSGEOMETRY_V1:
2128 return xfs_ioc_fsgeometry(mp, arg, 3);
2129 case XFS_IOC_FSGEOMETRY_V4:
2130 return xfs_ioc_fsgeometry(mp, arg, 4);
2131 case XFS_IOC_FSGEOMETRY:
2132 return xfs_ioc_fsgeometry(mp, arg, 5);
2134 case XFS_IOC_AG_GEOMETRY:
2135 return xfs_ioc_ag_geometry(mp, arg);
2137 case XFS_IOC_GETVERSION:
2138 return put_user(inode->i_generation, (int __user *)arg);
2140 case XFS_IOC_FSGETXATTR:
2141 return xfs_ioc_fsgetxattr(ip, 0, arg);
2142 case XFS_IOC_FSGETXATTRA:
2143 return xfs_ioc_fsgetxattr(ip, 1, arg);
2144 case XFS_IOC_FSSETXATTR:
2145 return xfs_ioc_fssetxattr(filp, arg);
2146 case XFS_IOC_GETXFLAGS:
2147 return xfs_ioc_getxflags(ip, arg);
2148 case XFS_IOC_SETXFLAGS:
2149 return xfs_ioc_setxflags(ip, filp, arg);
2151 case XFS_IOC_GETBMAP:
2152 case XFS_IOC_GETBMAPA:
2153 case XFS_IOC_GETBMAPX:
2154 return xfs_ioc_getbmap(filp, cmd, arg);
2156 case FS_IOC_GETFSMAP:
2157 return xfs_ioc_getfsmap(ip, arg);
2159 case XFS_IOC_SCRUB_METADATA:
2160 return xfs_ioc_scrub_metadata(ip, arg);
2162 case XFS_IOC_FD_TO_HANDLE:
2163 case XFS_IOC_PATH_TO_HANDLE:
2164 case XFS_IOC_PATH_TO_FSHANDLE: {
2165 xfs_fsop_handlereq_t hreq;
2167 if (copy_from_user(&hreq, arg, sizeof(hreq)))
2169 return xfs_find_handle(cmd, &hreq);
2171 case XFS_IOC_OPEN_BY_HANDLE: {
2172 xfs_fsop_handlereq_t hreq;
2174 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2176 return xfs_open_by_handle(filp, &hreq);
2179 case XFS_IOC_READLINK_BY_HANDLE: {
2180 xfs_fsop_handlereq_t hreq;
2182 if (copy_from_user(&hreq, arg, sizeof(xfs_fsop_handlereq_t)))
2184 return xfs_readlink_by_handle(filp, &hreq);
2186 case XFS_IOC_ATTRLIST_BY_HANDLE:
2187 return xfs_attrlist_by_handle(filp, arg);
2189 case XFS_IOC_ATTRMULTI_BY_HANDLE:
2190 return xfs_attrmulti_by_handle(filp, arg);
2192 case XFS_IOC_SWAPEXT: {
2193 struct xfs_swapext sxp;
2195 if (copy_from_user(&sxp, arg, sizeof(xfs_swapext_t)))
2197 error = mnt_want_write_file(filp);
2200 error = xfs_ioc_swapext(&sxp);
2201 mnt_drop_write_file(filp);
2205 case XFS_IOC_FSCOUNTS: {
2206 xfs_fsop_counts_t out;
2208 xfs_fs_counts(mp, &out);
2210 if (copy_to_user(arg, &out, sizeof(out)))
2215 case XFS_IOC_SET_RESBLKS: {
2216 xfs_fsop_resblks_t inout;
2219 if (!capable(CAP_SYS_ADMIN))
2222 if (mp->m_flags & XFS_MOUNT_RDONLY)
2225 if (copy_from_user(&inout, arg, sizeof(inout)))
2228 error = mnt_want_write_file(filp);
2232 /* input parameter is passed in resblks field of structure */
2234 error = xfs_reserve_blocks(mp, &in, &inout);
2235 mnt_drop_write_file(filp);
2239 if (copy_to_user(arg, &inout, sizeof(inout)))
2244 case XFS_IOC_GET_RESBLKS: {
2245 xfs_fsop_resblks_t out;
2247 if (!capable(CAP_SYS_ADMIN))
2250 error = xfs_reserve_blocks(mp, NULL, &out);
2254 if (copy_to_user(arg, &out, sizeof(out)))
2260 case XFS_IOC_FSGROWFSDATA: {
2261 struct xfs_growfs_data in;
2263 if (copy_from_user(&in, arg, sizeof(in)))
2266 error = mnt_want_write_file(filp);
2269 error = xfs_growfs_data(mp, &in);
2270 mnt_drop_write_file(filp);
2274 case XFS_IOC_FSGROWFSLOG: {
2275 struct xfs_growfs_log in;
2277 if (copy_from_user(&in, arg, sizeof(in)))
2280 error = mnt_want_write_file(filp);
2283 error = xfs_growfs_log(mp, &in);
2284 mnt_drop_write_file(filp);
2288 case XFS_IOC_FSGROWFSRT: {
2291 if (copy_from_user(&in, arg, sizeof(in)))
2294 error = mnt_want_write_file(filp);
2297 error = xfs_growfs_rt(mp, &in);
2298 mnt_drop_write_file(filp);
2302 case XFS_IOC_GOINGDOWN: {
2305 if (!capable(CAP_SYS_ADMIN))
2308 if (get_user(in, (uint32_t __user *)arg))
2311 return xfs_fs_goingdown(mp, in);
2314 case XFS_IOC_ERROR_INJECTION: {
2315 xfs_error_injection_t in;
2317 if (!capable(CAP_SYS_ADMIN))
2320 if (copy_from_user(&in, arg, sizeof(in)))
2323 return xfs_errortag_add(mp, in.errtag);
2326 case XFS_IOC_ERROR_CLEARALL:
2327 if (!capable(CAP_SYS_ADMIN))
2330 return xfs_errortag_clearall(mp);
2332 case XFS_IOC_FREE_EOFBLOCKS: {
2333 struct xfs_fs_eofblocks eofb;
2334 struct xfs_eofblocks keofb;
2336 if (!capable(CAP_SYS_ADMIN))
2339 if (mp->m_flags & XFS_MOUNT_RDONLY)
2342 if (copy_from_user(&eofb, arg, sizeof(eofb)))
2345 error = xfs_fs_eofblocks_from_user(&eofb, &keofb);
2349 trace_xfs_ioc_free_eofblocks(mp, &keofb, _RET_IP_);
2351 sb_start_write(mp->m_super);
2352 error = xfs_blockgc_free_space(mp, &keofb);
2353 sb_end_write(mp->m_super);