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 ip = XFS_I(file->f_path.dentry->d_inode);
83 tip = XFS_I(target_file->f_path.dentry->d_inode);
85 if (ip->i_mount != tip->i_mount) {
86 error = XFS_ERROR(EINVAL);
87 goto out_put_target_file;
90 if (ip->i_ino == tip->i_ino) {
91 error = XFS_ERROR(EINVAL);
92 goto out_put_target_file;
95 if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
96 error = XFS_ERROR(EIO);
97 goto out_put_target_file;
100 error = xfs_swap_extents(ip, tip, sxp);
118 xfs_bstat_t *sbp = &sxp->sx_stat;
119 xfs_ifork_t *tempifp, *ifp, *tifp;
120 int ilf_fields, tilf_fields;
128 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
130 error = XFS_ERROR(ENOMEM);
137 * we have to do two separate lock calls here to keep lockdep
138 * happy. If we try to get all the locks in one call, lock will
139 * report false positives when we drop the ILOCK and regain them
142 xfs_lock_two_inodes(ip, tip, XFS_IOLOCK_EXCL);
143 xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
145 /* Verify that both files have the same format */
146 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
147 error = XFS_ERROR(EINVAL);
151 /* Verify both files are either real-time or non-realtime */
152 if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) {
153 error = XFS_ERROR(EINVAL);
157 /* Should never get a local format */
158 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
159 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
160 error = XFS_ERROR(EINVAL);
164 if (VN_CACHED(VFS_I(tip)) != 0) {
165 xfs_inval_cached_trace(tip, 0, -1, 0, -1);
166 error = xfs_flushinval_pages(tip, 0, -1,
172 /* Verify O_DIRECT for ftmp */
173 if (VN_CACHED(VFS_I(tip)) != 0) {
174 error = XFS_ERROR(EINVAL);
178 /* Verify all data are being swapped */
179 if (sxp->sx_offset != 0 ||
180 sxp->sx_length != ip->i_d.di_size ||
181 sxp->sx_length != tip->i_d.di_size) {
182 error = XFS_ERROR(EFAULT);
187 * If the target has extended attributes, the tmp file
188 * must also in order to ensure the correct data fork
191 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
192 error = XFS_ERROR(EINVAL);
197 * Compare the current change & modify times with that
198 * passed in. If they differ, we abort this swap.
199 * This is the mechanism used to ensure the calling
200 * process that the file was not changed out from
203 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
204 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
205 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
206 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
207 error = XFS_ERROR(EBUSY);
211 /* We need to fail if the file is memory mapped. Once we have tossed
212 * all existing pages, the page fault will have no option
213 * but to go to the filesystem for pages. By making the page fault call
214 * vop_read (or write in the case of autogrow) they block on the iolock
215 * until we have switched the extents.
217 if (VN_MAPPED(VFS_I(ip))) {
218 error = XFS_ERROR(EBUSY);
222 xfs_iunlock(ip, XFS_ILOCK_EXCL);
223 xfs_iunlock(tip, XFS_ILOCK_EXCL);
226 * There is a race condition here since we gave up the
227 * ilock. However, the data fork will not change since
228 * we have the iolock (locked for truncation too) so we
229 * are safe. We don't really care if non-io related
233 xfs_tosspages(ip, 0, -1, FI_REMAPF);
235 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
236 if ((error = xfs_trans_reserve(tp, 0,
237 XFS_ICHANGE_LOG_RES(mp), 0,
239 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
240 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
241 xfs_trans_cancel(tp, 0);
244 xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
247 * Count the number of extended attribute blocks
249 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
250 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
251 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
253 goto out_trans_cancel;
255 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
256 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
257 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
260 goto out_trans_cancel;
264 * Swap the data forks of the inodes
268 *tempifp = *ifp; /* struct copy */
269 *ifp = *tifp; /* struct copy */
270 *tifp = *tempifp; /* struct copy */
273 * Fix the on-disk inode values
275 tmp = (__uint64_t)ip->i_d.di_nblocks;
276 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
277 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
279 tmp = (__uint64_t) ip->i_d.di_nextents;
280 ip->i_d.di_nextents = tip->i_d.di_nextents;
281 tip->i_d.di_nextents = tmp;
283 tmp = (__uint64_t) ip->i_d.di_format;
284 ip->i_d.di_format = tip->i_d.di_format;
285 tip->i_d.di_format = tmp;
287 ilf_fields = XFS_ILOG_CORE;
289 switch(ip->i_d.di_format) {
290 case XFS_DINODE_FMT_EXTENTS:
291 /* If the extents fit in the inode, fix the
292 * pointer. Otherwise it's already NULL or
293 * pointing to the extent.
295 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
296 ifp->if_u1.if_extents =
297 ifp->if_u2.if_inline_ext;
299 ilf_fields |= XFS_ILOG_DEXT;
301 case XFS_DINODE_FMT_BTREE:
302 ilf_fields |= XFS_ILOG_DBROOT;
306 tilf_fields = XFS_ILOG_CORE;
308 switch(tip->i_d.di_format) {
309 case XFS_DINODE_FMT_EXTENTS:
310 /* If the extents fit in the inode, fix the
311 * pointer. Otherwise it's already NULL or
312 * pointing to the extent.
314 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
315 tifp->if_u1.if_extents =
316 tifp->if_u2.if_inline_ext;
318 tilf_fields |= XFS_ILOG_DEXT;
320 case XFS_DINODE_FMT_BTREE:
321 tilf_fields |= XFS_ILOG_DBROOT;
327 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
330 xfs_trans_ijoin(tp, tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
332 xfs_trans_log_inode(tp, ip, ilf_fields);
333 xfs_trans_log_inode(tp, tip, tilf_fields);
336 * If this is a synchronous mount, make sure that the
337 * transaction goes to disk before returning to the user.
339 if (mp->m_flags & XFS_MOUNT_WSYNC)
340 xfs_trans_set_sync(tp);
342 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
345 xfs_iunlock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
346 xfs_iunlock(tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
352 xfs_trans_cancel(tp, 0);