2 * Copyright (c) 2000-2002 Silicon Graphics, Inc. All Rights Reserved.
4 * This program is free software; you can redistribute it and/or modify it
5 * under the terms of version 2 of the GNU General Public License as
6 * published by the Free Software Foundation.
8 * This program is distributed in the hope that it would be useful, but
9 * WITHOUT ANY WARRANTY; without even the implied warranty of
10 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
12 * Further, this software is distributed without any warranty that it is
13 * free of the rightful claim of any third person regarding infringement
14 * or the like. Any license provided herein, whether implied or
15 * otherwise, applies only to this software file. Patent licenses, if
16 * any, provided herein do not apply to combinations of this program with
17 * other software, or any other product whatsoever.
19 * You should have received a copy of the GNU General Public License along
20 * with this program; if not, write the Free Software Foundation, Inc., 59
21 * Temple Place - Suite 330, Boston MA 02111-1307, USA.
23 * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy,
24 * Mountain View, CA 94043, or:
28 * For further information regarding this notice, see:
30 * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/
34 #include "xfs_macros.h"
35 #include "xfs_types.h"
38 #include "xfs_trans.h"
42 #include "xfs_dmapi.h"
43 #include "xfs_mount.h"
45 #include "xfs_alloc_btree.h"
46 #include "xfs_bmap_btree.h"
47 #include "xfs_ialloc_btree.h"
48 #include "xfs_btree.h"
49 #include "xfs_attr_sf.h"
50 #include "xfs_dir_sf.h"
51 #include "xfs_dir2_sf.h"
52 #include "xfs_dinode.h"
53 #include "xfs_inode_item.h"
54 #include "xfs_inode.h"
56 #include "xfs_ialloc.h"
57 #include "xfs_itable.h"
58 #include "xfs_dfrag.h"
59 #include "xfs_error.h"
66 * Syssgi interface for swapext
73 xfs_inode_t *ip=NULL, *tip=NULL, *ips[2];
77 xfs_vnode_t *vp = NULL, *tvp = NULL;
78 struct vnode *bvp, *btvp;
79 bhv_desc_t *bdp, *tbdp;
80 vn_bhv_head_t *bhp, *tbhp;
82 int ilf_fields, tilf_fields;
84 xfs_ifork_t tempif, *ifp, *tifp;
95 if (copy_from_user(&sx, sxp, sizeof(sx)))
96 return XFS_ERROR(EFAULT);
98 /* Pull information for the target fd */
99 if (fgetvp(td, (int)sx.sx_fdtarget, &bvp) != 0) {
100 error = XFS_ERROR(EINVAL);
105 bhp = VN_BHV_HEAD(vp);
106 bdp = vn_bhv_lookup(bhp, &xfs_vnodeops);
108 error = XFS_ERROR(EBADF);
111 ip = XFS_BHVTOI(bdp);
114 if (fgetvp(td, (int)sx.sx_fdtmp, &btvp) != 0) {
115 error = XFS_ERROR(EINVAL);
119 tvp = VPTOXFSVP(btvp);
120 tbhp = VN_BHV_HEAD(tvp);
121 tbdp = vn_bhv_lookup(tbhp, &xfs_vnodeops);
123 error = XFS_ERROR(EBADF);
126 tip = XFS_BHVTOI(tbdp);
129 if (ip->i_ino == tip->i_ino) {
130 error = XFS_ERROR(EINVAL);
138 if (XFS_FORCED_SHUTDOWN(mp)) {
139 error = XFS_ERROR(EIO);
145 /* Lock in i_ino order */
146 if (ip->i_ino < tip->i_ino) {
153 lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
154 xfs_lock_inodes(ips, 2, 0, lock_flags);
156 /* Check permissions */
157 if ((error = _MAC_XFS_IACCESS(ip, MACWRITE, NULL))) {
160 if ((error = _MAC_XFS_IACCESS(tip, MACWRITE, NULL))) {
163 if ((current_fsuid(cred) != ip->i_d.di_uid) &&
164 (error = xfs_iaccess(ip, S_IWUSR, NULL)) &&
165 !capable_cred(NULL, CAP_FOWNER)) {
168 if ((current_fsuid(cred) != tip->i_d.di_uid) &&
169 (error = xfs_iaccess(tip, S_IWUSR, NULL)) &&
170 !capable_cred(NULL, CAP_FOWNER)) {
174 /* Verify both files are either real-time or non-realtime */
175 if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
176 (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
177 error = XFS_ERROR(EINVAL);
181 /* Should never get a local format */
182 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
183 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
184 error = XFS_ERROR(EINVAL);
188 if (VN_CACHED(tvp) != 0)
189 xfs_inval_cached_pages(XFS_ITOV(tip), &(tip->i_iocore),
192 /* Verify O_DIRECT for ftmp */
193 if (VN_CACHED(tvp) != 0) {
194 error = XFS_ERROR(EINVAL);
198 /* Verify all data are being swapped */
199 if (sx.sx_offset != 0 ||
200 sx.sx_length != ip->i_d.di_size ||
201 sx.sx_length != tip->i_d.di_size) {
202 error = XFS_ERROR(EFAULT);
207 * If the target has extended attributes, the tmp file
208 * must also in order to ensure the correct data fork
211 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
212 error = XFS_ERROR(EINVAL);
217 * Compare the current change & modify times with that
218 * passed in. If they differ, we abort this swap.
219 * This is the mechanism used to ensure the calling
220 * process that the file was not changed out from
223 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
224 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
225 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
226 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
227 error = XFS_ERROR(EBUSY);
231 /* We need to fail if the file is memory mapped. Once we have tossed
232 * all existing pages, the page fault will have no option
233 * but to go to the filesystem for pages. By making the page fault call
234 * VOP_READ (or write in the case of autogrow) they block on the iolock
235 * until we have switched the extents.
238 error = XFS_ERROR(EBUSY);
242 xfs_iunlock(ip, XFS_ILOCK_EXCL);
243 xfs_iunlock(tip, XFS_ILOCK_EXCL);
246 * There is a race condition here since we gave up the
247 * ilock. However, the data fork will not change since
248 * we have the iolock (locked for truncation too) so we
249 * are safe. We don't really care if non-io related
253 XVOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
255 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
256 if ((error = xfs_trans_reserve(tp, 0,
257 XFS_ICHANGE_LOG_RES(mp), 0,
259 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
260 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
261 xfs_trans_cancel(tp, 0);
264 xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
267 * Count the number of extended attribute blocks
269 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
270 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
271 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
273 xfs_iunlock(ip, lock_flags);
274 xfs_iunlock(tip, lock_flags);
275 xfs_trans_cancel(tp, 0);
279 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
280 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
281 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
284 xfs_iunlock(ip, lock_flags);
285 xfs_iunlock(tip, lock_flags);
286 xfs_trans_cancel(tp, 0);
292 * Swap the data forks of the inodes
296 tempif = *ifp; /* struct copy */
297 *ifp = *tifp; /* struct copy */
298 *tifp = tempif; /* struct copy */
301 * Fix the on-disk inode values
303 tmp = (__uint64_t)ip->i_d.di_nblocks;
304 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
305 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
307 tmp = (__uint64_t) ip->i_d.di_nextents;
308 ip->i_d.di_nextents = tip->i_d.di_nextents;
309 tip->i_d.di_nextents = tmp;
311 tmp = (__uint64_t) ip->i_d.di_format;
312 ip->i_d.di_format = tip->i_d.di_format;
313 tip->i_d.di_format = tmp;
315 ilf_fields = XFS_ILOG_CORE;
317 switch(ip->i_d.di_format) {
318 case XFS_DINODE_FMT_EXTENTS:
319 /* If the extents fit in the inode, fix the
320 * pointer. Otherwise it's already NULL or
321 * pointing to the extent.
323 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
324 ifp->if_u1.if_extents =
325 ifp->if_u2.if_inline_ext;
327 ilf_fields |= XFS_ILOG_DEXT;
329 case XFS_DINODE_FMT_BTREE:
330 ilf_fields |= XFS_ILOG_DBROOT;
334 tilf_fields = XFS_ILOG_CORE;
336 switch(tip->i_d.di_format) {
337 case XFS_DINODE_FMT_EXTENTS:
338 /* If the extents fit in the inode, fix the
339 * pointer. Otherwise it's already NULL or
340 * pointing to the extent.
342 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
343 tifp->if_u1.if_extents =
344 tifp->if_u2.if_inline_ext;
346 tilf_fields |= XFS_ILOG_DEXT;
348 case XFS_DINODE_FMT_BTREE:
349 tilf_fields |= XFS_ILOG_DBROOT;
353 #if XXXKAN /* Not necessary, vnodes are vrefed already by fgetvp */
355 * Increment vnode ref counts since xfs_trans_commit &
356 * xfs_trans_cancel will both unlock the inodes and
357 * decrement the associated ref counts.
363 xfs_trans_ijoin(tp, ip, lock_flags);
364 xfs_trans_ijoin(tp, tip, lock_flags);
366 xfs_trans_log_inode(tp, ip, ilf_fields);
367 xfs_trans_log_inode(tp, tip, tilf_fields);
370 * If this is a synchronous mount, make sure that the
371 * transaction goes to disk before returning to the user.
373 if (mp->m_flags & XFS_MOUNT_WSYNC) {
374 xfs_trans_set_sync(tp);
377 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
383 xfs_iunlock(ip, lock_flags);
384 xfs_iunlock(tip, lock_flags);
387 if (vp != NULL) VN_RELE(vp);
388 if (tvp != NULL) VN_RELE(tvp);