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
52 xfs_swapext_t __user *sxu)
55 xfs_inode_t *ip, *tip;
56 struct file *file, *target_file;
59 sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
61 error = XFS_ERROR(ENOMEM);
65 if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
66 error = XFS_ERROR(EFAULT);
70 /* Pull information for the target fd */
71 file = fget((int)sxp->sx_fdtarget);
73 error = XFS_ERROR(EINVAL);
77 if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) {
78 error = XFS_ERROR(EBADF);
82 target_file = fget((int)sxp->sx_fdtmp);
84 error = XFS_ERROR(EINVAL);
88 if (!(target_file->f_mode & FMODE_WRITE) ||
89 (target_file->f_flags & O_APPEND)) {
90 error = XFS_ERROR(EBADF);
91 goto out_put_target_file;
94 ip = XFS_I(file->f_path.dentry->d_inode);
95 tip = XFS_I(target_file->f_path.dentry->d_inode);
97 if (ip->i_mount != tip->i_mount) {
98 error = XFS_ERROR(EINVAL);
99 goto out_put_target_file;
102 if (ip->i_ino == tip->i_ino) {
103 error = XFS_ERROR(EINVAL);
104 goto out_put_target_file;
107 if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
108 error = XFS_ERROR(EIO);
109 goto out_put_target_file;
112 error = xfs_swap_extents(ip, tip, sxp);
132 xfs_bstat_t *sbp = &sxp->sx_stat;
133 bhv_vnode_t *vp, *tvp;
134 xfs_ifork_t *tempifp, *ifp, *tifp;
135 int ilf_fields, tilf_fields;
136 static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
145 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
147 error = XFS_ERROR(ENOMEM);
155 xfs_lock_two_inodes(ip, tip, lock_flags);
158 /* Verify that both files have the same format */
159 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
160 error = XFS_ERROR(EINVAL);
164 /* Verify both files are either real-time or non-realtime */
165 if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) {
166 error = XFS_ERROR(EINVAL);
170 /* Should never get a local format */
171 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
172 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
173 error = XFS_ERROR(EINVAL);
177 if (VN_CACHED(tvp) != 0) {
178 xfs_inval_cached_trace(tip, 0, -1, 0, -1);
179 error = xfs_flushinval_pages(tip, 0, -1,
185 /* Verify O_DIRECT for ftmp */
186 if (VN_CACHED(tvp) != 0) {
187 error = XFS_ERROR(EINVAL);
191 /* Verify all data are being swapped */
192 if (sxp->sx_offset != 0 ||
193 sxp->sx_length != ip->i_d.di_size ||
194 sxp->sx_length != tip->i_d.di_size) {
195 error = XFS_ERROR(EFAULT);
200 * If the target has extended attributes, the tmp file
201 * must also in order to ensure the correct data fork
204 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
205 error = XFS_ERROR(EINVAL);
210 * Compare the current change & modify times with that
211 * passed in. If they differ, we abort this swap.
212 * This is the mechanism used to ensure the calling
213 * process that the file was not changed out from
216 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
217 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
218 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
219 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
220 error = XFS_ERROR(EBUSY);
224 /* We need to fail if the file is memory mapped. Once we have tossed
225 * all existing pages, the page fault will have no option
226 * but to go to the filesystem for pages. By making the page fault call
227 * vop_read (or write in the case of autogrow) they block on the iolock
228 * until we have switched the extents.
231 error = XFS_ERROR(EBUSY);
235 xfs_iunlock(ip, XFS_ILOCK_EXCL);
236 xfs_iunlock(tip, XFS_ILOCK_EXCL);
239 * There is a race condition here since we gave up the
240 * ilock. However, the data fork will not change since
241 * we have the iolock (locked for truncation too) so we
242 * are safe. We don't really care if non-io related
246 xfs_tosspages(ip, 0, -1, FI_REMAPF);
248 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
249 if ((error = xfs_trans_reserve(tp, 0,
250 XFS_ICHANGE_LOG_RES(mp), 0,
252 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
253 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
254 xfs_trans_cancel(tp, 0);
258 xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
261 * Count the number of extended attribute blocks
263 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
264 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
265 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
267 xfs_trans_cancel(tp, 0);
271 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
272 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
273 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
276 xfs_trans_cancel(tp, 0);
282 * Swap the data forks of the inodes
286 *tempifp = *ifp; /* struct copy */
287 *ifp = *tifp; /* struct copy */
288 *tifp = *tempifp; /* struct copy */
291 * Fix the on-disk inode values
293 tmp = (__uint64_t)ip->i_d.di_nblocks;
294 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
295 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
297 tmp = (__uint64_t) ip->i_d.di_nextents;
298 ip->i_d.di_nextents = tip->i_d.di_nextents;
299 tip->i_d.di_nextents = tmp;
301 tmp = (__uint64_t) ip->i_d.di_format;
302 ip->i_d.di_format = tip->i_d.di_format;
303 tip->i_d.di_format = tmp;
305 ilf_fields = XFS_ILOG_CORE;
307 switch(ip->i_d.di_format) {
308 case XFS_DINODE_FMT_EXTENTS:
309 /* If the extents fit in the inode, fix the
310 * pointer. Otherwise it's already NULL or
311 * pointing to the extent.
313 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
314 ifp->if_u1.if_extents =
315 ifp->if_u2.if_inline_ext;
317 ilf_fields |= XFS_ILOG_DEXT;
319 case XFS_DINODE_FMT_BTREE:
320 ilf_fields |= XFS_ILOG_DBROOT;
324 tilf_fields = XFS_ILOG_CORE;
326 switch(tip->i_d.di_format) {
327 case XFS_DINODE_FMT_EXTENTS:
328 /* If the extents fit in the inode, fix the
329 * pointer. Otherwise it's already NULL or
330 * pointing to the extent.
332 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
333 tifp->if_u1.if_extents =
334 tifp->if_u2.if_inline_ext;
336 tilf_fields |= XFS_ILOG_DEXT;
338 case XFS_DINODE_FMT_BTREE:
339 tilf_fields |= XFS_ILOG_DBROOT;
345 xfs_trans_ijoin(tp, ip, lock_flags);
348 xfs_trans_ijoin(tp, tip, lock_flags);
350 xfs_trans_log_inode(tp, ip, ilf_fields);
351 xfs_trans_log_inode(tp, tip, tilf_fields);
354 * If this is a synchronous mount, make sure that the
355 * transaction goes to disk before returning to the user.
357 if (mp->m_flags & XFS_MOUNT_WSYNC) {
358 xfs_trans_set_sync(tp);
361 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
366 xfs_iunlock(ip, lock_flags);
367 xfs_iunlock(tip, lock_flags);