2 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
5 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
7 * published by the Free Software Foundation.
9 * This program is distributed in the hope that it would be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
14 * You should have received a copy of the GNU General Public License
15 * along with this program; if not, write the Free Software Foundation,
16 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
20 #include "xfs_types.h"
24 #include "xfs_trans.h"
28 #include "xfs_dmapi.h"
29 #include "xfs_mount.h"
30 #include "xfs_bmap_btree.h"
31 #include "xfs_alloc_btree.h"
32 #include "xfs_ialloc_btree.h"
33 #include "xfs_dir2_sf.h"
34 #include "xfs_attr_sf.h"
35 #include "xfs_dinode.h"
36 #include "xfs_inode.h"
37 #include "xfs_inode_item.h"
39 #include "xfs_btree.h"
40 #include "xfs_ialloc.h"
41 #include "xfs_itable.h"
42 #include "xfs_dfrag.h"
43 #include "xfs_error.h"
45 #include "xfs_vnodeops.h"
48 * Syssgi interface for swapext
54 xfs_inode_t *ip, *tip;
55 struct file *file, *target_file;
58 /* Pull information for the target fd */
59 file = fget((int)sxp->sx_fdtarget);
61 error = XFS_ERROR(EINVAL);
65 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) {
66 error = XFS_ERROR(EBADF);
70 target_file = fget((int)sxp->sx_fdtmp);
72 error = XFS_ERROR(EINVAL);
76 if (!(target_file->f_mode & FMODE_WRITE) ||
77 (target_file->f_flags & O_APPEND)) {
78 error = XFS_ERROR(EBADF);
79 goto out_put_target_file;
82 if (IS_SWAPFILE(file->f_path.dentry->d_inode) ||
83 IS_SWAPFILE(target_file->f_path.dentry->d_inode)) {
84 error = XFS_ERROR(EINVAL);
85 goto out_put_target_file;
88 ip = XFS_I(file->f_path.dentry->d_inode);
89 tip = XFS_I(target_file->f_path.dentry->d_inode);
91 if (ip->i_mount != tip->i_mount) {
92 error = XFS_ERROR(EINVAL);
93 goto out_put_target_file;
96 if (ip->i_ino == tip->i_ino) {
97 error = XFS_ERROR(EINVAL);
98 goto out_put_target_file;
101 if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
102 error = XFS_ERROR(EIO);
103 goto out_put_target_file;
106 error = xfs_swap_extents(ip, tip, sxp);
124 xfs_bstat_t *sbp = &sxp->sx_stat;
125 xfs_ifork_t *tempifp, *ifp, *tifp;
126 int ilf_fields, tilf_fields;
134 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
136 error = XFS_ERROR(ENOMEM);
143 * we have to do two separate lock calls here to keep lockdep
144 * happy. If we try to get all the locks in one call, lock will
145 * report false positives when we drop the ILOCK and regain them
148 xfs_lock_two_inodes(ip, tip, XFS_IOLOCK_EXCL);
149 xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
151 /* Verify that both files have the same format */
152 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
153 error = XFS_ERROR(EINVAL);
157 /* Verify both files are either real-time or non-realtime */
158 if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) {
159 error = XFS_ERROR(EINVAL);
163 /* Should never get a local format */
164 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
165 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
166 error = XFS_ERROR(EINVAL);
170 if (VN_CACHED(VFS_I(tip)) != 0) {
171 xfs_inval_cached_trace(tip, 0, -1, 0, -1);
172 error = xfs_flushinval_pages(tip, 0, -1,
178 /* Verify O_DIRECT for ftmp */
179 if (VN_CACHED(VFS_I(tip)) != 0) {
180 error = XFS_ERROR(EINVAL);
184 /* Verify all data are being swapped */
185 if (sxp->sx_offset != 0 ||
186 sxp->sx_length != ip->i_d.di_size ||
187 sxp->sx_length != tip->i_d.di_size) {
188 error = XFS_ERROR(EFAULT);
193 * If the target has extended attributes, the tmp file
194 * must also in order to ensure the correct data fork
197 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
198 error = XFS_ERROR(EINVAL);
203 * Compare the current change & modify times with that
204 * passed in. If they differ, we abort this swap.
205 * This is the mechanism used to ensure the calling
206 * process that the file was not changed out from
209 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
210 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
211 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
212 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
213 error = XFS_ERROR(EBUSY);
217 /* We need to fail if the file is memory mapped. Once we have tossed
218 * all existing pages, the page fault will have no option
219 * but to go to the filesystem for pages. By making the page fault call
220 * vop_read (or write in the case of autogrow) they block on the iolock
221 * until we have switched the extents.
223 if (VN_MAPPED(VFS_I(ip))) {
224 error = XFS_ERROR(EBUSY);
228 xfs_iunlock(ip, XFS_ILOCK_EXCL);
229 xfs_iunlock(tip, XFS_ILOCK_EXCL);
232 * There is a race condition here since we gave up the
233 * ilock. However, the data fork will not change since
234 * we have the iolock (locked for truncation too) so we
235 * are safe. We don't really care if non-io related
239 xfs_tosspages(ip, 0, -1, FI_REMAPF);
241 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
242 if ((error = xfs_trans_reserve(tp, 0,
243 XFS_ICHANGE_LOG_RES(mp), 0,
245 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
246 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
247 xfs_trans_cancel(tp, 0);
250 xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
253 * Count the number of extended attribute blocks
255 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
256 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
257 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
259 goto out_trans_cancel;
261 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
262 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
263 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
266 goto out_trans_cancel;
270 * Swap the data forks of the inodes
274 *tempifp = *ifp; /* struct copy */
275 *ifp = *tifp; /* struct copy */
276 *tifp = *tempifp; /* struct copy */
279 * Fix the on-disk inode values
281 tmp = (__uint64_t)ip->i_d.di_nblocks;
282 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
283 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
285 tmp = (__uint64_t) ip->i_d.di_nextents;
286 ip->i_d.di_nextents = tip->i_d.di_nextents;
287 tip->i_d.di_nextents = tmp;
289 tmp = (__uint64_t) ip->i_d.di_format;
290 ip->i_d.di_format = tip->i_d.di_format;
291 tip->i_d.di_format = tmp;
293 ilf_fields = XFS_ILOG_CORE;
295 switch(ip->i_d.di_format) {
296 case XFS_DINODE_FMT_EXTENTS:
297 /* If the extents fit in the inode, fix the
298 * pointer. Otherwise it's already NULL or
299 * pointing to the extent.
301 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
302 ifp->if_u1.if_extents =
303 ifp->if_u2.if_inline_ext;
305 ilf_fields |= XFS_ILOG_DEXT;
307 case XFS_DINODE_FMT_BTREE:
308 ilf_fields |= XFS_ILOG_DBROOT;
312 tilf_fields = XFS_ILOG_CORE;
314 switch(tip->i_d.di_format) {
315 case XFS_DINODE_FMT_EXTENTS:
316 /* If the extents fit in the inode, fix the
317 * pointer. Otherwise it's already NULL or
318 * pointing to the extent.
320 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
321 tifp->if_u1.if_extents =
322 tifp->if_u2.if_inline_ext;
324 tilf_fields |= XFS_ILOG_DEXT;
326 case XFS_DINODE_FMT_BTREE:
327 tilf_fields |= XFS_ILOG_DBROOT;
333 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
336 xfs_trans_ijoin(tp, tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
338 xfs_trans_log_inode(tp, ip, ilf_fields);
339 xfs_trans_log_inode(tp, tip, tilf_fields);
342 * If this is a synchronous mount, make sure that the
343 * transaction goes to disk before returning to the user.
345 if (mp->m_flags & XFS_MOUNT_WSYNC)
346 xfs_trans_set_sync(tp);
348 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
351 xfs_iunlock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
352 xfs_iunlock(tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
358 xfs_trans_cancel(tp, 0);