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 target_file = fget((int)sxp->sx_fdtmp);
79 error = XFS_ERROR(EINVAL);
83 ip = XFS_I(file->f_path.dentry->d_inode);
84 tip = XFS_I(target_file->f_path.dentry->d_inode);
86 if (ip->i_mount != tip->i_mount) {
87 error = XFS_ERROR(EINVAL);
88 goto out_put_target_file;
91 if (ip->i_ino == tip->i_ino) {
92 error = XFS_ERROR(EINVAL);
93 goto out_put_target_file;
96 if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
97 error = XFS_ERROR(EIO);
98 goto out_put_target_file;
101 error = xfs_swap_extents(ip, tip, sxp);
108 kmem_free(sxp, sizeof(xfs_swapext_t));
122 xfs_bstat_t *sbp = &sxp->sx_stat;
123 bhv_vnode_t *vp, *tvp;
124 xfs_ifork_t *tempifp, *ifp, *tifp;
125 int ilf_fields, tilf_fields;
126 static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
135 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
137 error = XFS_ERROR(ENOMEM);
145 /* Lock in i_ino order */
146 if (ip->i_ino < tip->i_ino) {
154 xfs_lock_inodes(ips, 2, 0, lock_flags);
157 /* Check permissions */
158 error = xfs_iaccess(ip, S_IWUSR, NULL);
162 error = xfs_iaccess(tip, S_IWUSR, NULL);
166 /* Verify that both files have the same format */
167 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
168 error = XFS_ERROR(EINVAL);
172 /* Verify both files are either real-time or non-realtime */
173 if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) {
174 error = XFS_ERROR(EINVAL);
178 /* Should never get a local format */
179 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
180 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
181 error = XFS_ERROR(EINVAL);
185 if (VN_CACHED(tvp) != 0) {
186 xfs_inval_cached_trace(tip, 0, -1, 0, -1);
187 error = xfs_flushinval_pages(tip, 0, -1,
193 /* Verify O_DIRECT for ftmp */
194 if (VN_CACHED(tvp) != 0) {
195 error = XFS_ERROR(EINVAL);
199 /* Verify all data are being swapped */
200 if (sxp->sx_offset != 0 ||
201 sxp->sx_length != ip->i_d.di_size ||
202 sxp->sx_length != tip->i_d.di_size) {
203 error = XFS_ERROR(EFAULT);
208 * If the target has extended attributes, the tmp file
209 * must also in order to ensure the correct data fork
212 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
213 error = XFS_ERROR(EINVAL);
218 * Compare the current change & modify times with that
219 * passed in. If they differ, we abort this swap.
220 * This is the mechanism used to ensure the calling
221 * process that the file was not changed out from
224 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
225 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
226 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
227 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
228 error = XFS_ERROR(EBUSY);
232 /* We need to fail if the file is memory mapped. Once we have tossed
233 * all existing pages, the page fault will have no option
234 * but to go to the filesystem for pages. By making the page fault call
235 * vop_read (or write in the case of autogrow) they block on the iolock
236 * until we have switched the extents.
239 error = XFS_ERROR(EBUSY);
243 xfs_iunlock(ip, XFS_ILOCK_EXCL);
244 xfs_iunlock(tip, XFS_ILOCK_EXCL);
247 * There is a race condition here since we gave up the
248 * ilock. However, the data fork will not change since
249 * we have the iolock (locked for truncation too) so we
250 * are safe. We don't really care if non-io related
254 xfs_tosspages(ip, 0, -1, FI_REMAPF);
256 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
257 if ((error = xfs_trans_reserve(tp, 0,
258 XFS_ICHANGE_LOG_RES(mp), 0,
260 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
261 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
262 xfs_trans_cancel(tp, 0);
266 xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
269 * Count the number of extended attribute blocks
271 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
272 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
273 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
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_trans_cancel(tp, 0);
290 * Swap the data forks of the inodes
294 *tempifp = *ifp; /* struct copy */
295 *ifp = *tifp; /* struct copy */
296 *tifp = *tempifp; /* struct copy */
299 * Fix the on-disk inode values
301 tmp = (__uint64_t)ip->i_d.di_nblocks;
302 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
303 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
305 tmp = (__uint64_t) ip->i_d.di_nextents;
306 ip->i_d.di_nextents = tip->i_d.di_nextents;
307 tip->i_d.di_nextents = tmp;
309 tmp = (__uint64_t) ip->i_d.di_format;
310 ip->i_d.di_format = tip->i_d.di_format;
311 tip->i_d.di_format = tmp;
313 ilf_fields = XFS_ILOG_CORE;
315 switch(ip->i_d.di_format) {
316 case XFS_DINODE_FMT_EXTENTS:
317 /* If the extents fit in the inode, fix the
318 * pointer. Otherwise it's already NULL or
319 * pointing to the extent.
321 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
322 ifp->if_u1.if_extents =
323 ifp->if_u2.if_inline_ext;
325 ilf_fields |= XFS_ILOG_DEXT;
327 case XFS_DINODE_FMT_BTREE:
328 ilf_fields |= XFS_ILOG_DBROOT;
332 tilf_fields = XFS_ILOG_CORE;
334 switch(tip->i_d.di_format) {
335 case XFS_DINODE_FMT_EXTENTS:
336 /* If the extents fit in the inode, fix the
337 * pointer. Otherwise it's already NULL or
338 * pointing to the extent.
340 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
341 tifp->if_u1.if_extents =
342 tifp->if_u2.if_inline_ext;
344 tilf_fields |= XFS_ILOG_DEXT;
346 case XFS_DINODE_FMT_BTREE:
347 tilf_fields |= XFS_ILOG_DBROOT;
352 * Increment vnode ref counts since xfs_trans_commit &
353 * xfs_trans_cancel will both unlock the inodes and
354 * decrement the associated ref counts.
359 xfs_trans_ijoin(tp, ip, lock_flags);
360 xfs_trans_ijoin(tp, tip, lock_flags);
362 xfs_trans_log_inode(tp, ip, ilf_fields);
363 xfs_trans_log_inode(tp, tip, tilf_fields);
366 * If this is a synchronous mount, make sure that the
367 * transaction goes to disk before returning to the user.
369 if (mp->m_flags & XFS_MOUNT_WSYNC) {
370 xfs_trans_set_sync(tp);
373 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
378 xfs_iunlock(ip, lock_flags);
379 xfs_iunlock(tip, lock_flags);
382 kmem_free(tempifp, sizeof(xfs_ifork_t));