2 * Copyright (c) 2000-2002,2005 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"
29 #include "xfs_dmapi.h"
30 #include "xfs_mount.h"
31 #include "xfs_bmap_btree.h"
32 #include "xfs_alloc_btree.h"
33 #include "xfs_ialloc_btree.h"
34 #include "xfs_dir_sf.h"
35 #include "xfs_dir2_sf.h"
36 #include "xfs_attr_sf.h"
37 #include "xfs_dinode.h"
38 #include "xfs_inode.h"
39 #include "xfs_inode_item.h"
41 #include "xfs_btree.h"
42 #include "xfs_ialloc.h"
43 #include "xfs_itable.h"
44 #include "xfs_dfrag.h"
45 #include "xfs_error.h"
50 * Syssgi interface for swapext
54 xfs_swapext_t __user *sxu)
57 xfs_inode_t *ip=NULL, *tip=NULL, *ips[2];
61 struct file *fp = NULL, *tfp = NULL;
63 static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
64 int ilf_fields, tilf_fields;
66 xfs_ifork_t *tempifp, *ifp, *tifp;
72 sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
73 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
74 if (!sxp || !tempifp) {
75 error = XFS_ERROR(ENOMEM);
79 if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
80 error = XFS_ERROR(EFAULT);
84 /* Pull information for the target fd */
85 if (((fp = fget((int)sxp->sx_fdtarget)) == NULL) ||
86 ((vp = LINVFS_GET_VP(fp->f_dentry->d_inode)) == NULL)) {
87 error = XFS_ERROR(EINVAL);
93 error = XFS_ERROR(EBADF);
97 if (((tfp = fget((int)sxp->sx_fdtmp)) == NULL) ||
98 ((tvp = LINVFS_GET_VP(tfp->f_dentry->d_inode)) == NULL)) {
99 error = XFS_ERROR(EINVAL);
105 error = XFS_ERROR(EBADF);
109 if (ip->i_mount != tip->i_mount) {
110 error = XFS_ERROR(EINVAL);
114 if (ip->i_ino == tip->i_ino) {
115 error = XFS_ERROR(EINVAL);
123 if (XFS_FORCED_SHUTDOWN(mp)) {
124 error = XFS_ERROR(EIO);
130 /* Lock in i_ino order */
131 if (ip->i_ino < tip->i_ino) {
139 xfs_lock_inodes(ips, 2, 0, lock_flags);
141 /* Check permissions */
142 error = xfs_iaccess(ip, S_IWUSR, NULL);
146 error = xfs_iaccess(tip, S_IWUSR, NULL);
150 /* Verify that both files have the same format */
151 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
152 error = XFS_ERROR(EINVAL);
156 /* Verify both files are either real-time or non-realtime */
157 if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
158 (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
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(tvp) != 0) {
171 xfs_inval_cached_trace(&tip->i_iocore, 0, -1, 0, -1);
172 VOP_FLUSHINVAL_PAGES(tvp, 0, -1, FI_REMAPF_LOCKED);
175 /* Verify O_DIRECT for ftmp */
176 if (VN_CACHED(tvp) != 0) {
177 error = XFS_ERROR(EINVAL);
181 /* Verify all data are being swapped */
182 if (sxp->sx_offset != 0 ||
183 sxp->sx_length != ip->i_d.di_size ||
184 sxp->sx_length != tip->i_d.di_size) {
185 error = XFS_ERROR(EFAULT);
190 * If the target has extended attributes, the tmp file
191 * must also in order to ensure the correct data fork
194 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
195 error = XFS_ERROR(EINVAL);
200 * Compare the current change & modify times with that
201 * passed in. If they differ, we abort this swap.
202 * This is the mechanism used to ensure the calling
203 * process that the file was not changed out from
206 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
207 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
208 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
209 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
210 error = XFS_ERROR(EBUSY);
214 /* We need to fail if the file is memory mapped. Once we have tossed
215 * all existing pages, the page fault will have no option
216 * but to go to the filesystem for pages. By making the page fault call
217 * VOP_READ (or write in the case of autogrow) they block on the iolock
218 * until we have switched the extents.
221 error = XFS_ERROR(EBUSY);
225 xfs_iunlock(ip, XFS_ILOCK_EXCL);
226 xfs_iunlock(tip, XFS_ILOCK_EXCL);
229 * There is a race condition here since we gave up the
230 * ilock. However, the data fork will not change since
231 * we have the iolock (locked for truncation too) so we
232 * are safe. We don't really care if non-io related
236 VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
238 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
239 if ((error = xfs_trans_reserve(tp, 0,
240 XFS_ICHANGE_LOG_RES(mp), 0,
242 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
243 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
244 xfs_trans_cancel(tp, 0);
248 xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
251 * Count the number of extended attribute blocks
253 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
254 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
255 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
257 xfs_trans_cancel(tp, 0);
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 xfs_trans_cancel(tp, 0);
272 * Swap the data forks of the inodes
276 *tempifp = *ifp; /* struct copy */
277 *ifp = *tifp; /* struct copy */
278 *tifp = *tempifp; /* struct copy */
281 * Fix the on-disk inode values
283 tmp = (__uint64_t)ip->i_d.di_nblocks;
284 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
285 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
287 tmp = (__uint64_t) ip->i_d.di_nextents;
288 ip->i_d.di_nextents = tip->i_d.di_nextents;
289 tip->i_d.di_nextents = tmp;
291 tmp = (__uint64_t) ip->i_d.di_format;
292 ip->i_d.di_format = tip->i_d.di_format;
293 tip->i_d.di_format = tmp;
295 ilf_fields = XFS_ILOG_CORE;
297 switch(ip->i_d.di_format) {
298 case XFS_DINODE_FMT_EXTENTS:
299 /* If the extents fit in the inode, fix the
300 * pointer. Otherwise it's already NULL or
301 * pointing to the extent.
303 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
304 ifp->if_u1.if_extents =
305 ifp->if_u2.if_inline_ext;
307 ilf_fields |= XFS_ILOG_DEXT;
309 case XFS_DINODE_FMT_BTREE:
310 ilf_fields |= XFS_ILOG_DBROOT;
314 tilf_fields = XFS_ILOG_CORE;
316 switch(tip->i_d.di_format) {
317 case XFS_DINODE_FMT_EXTENTS:
318 /* If the extents fit in the inode, fix the
319 * pointer. Otherwise it's already NULL or
320 * pointing to the extent.
322 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
323 tifp->if_u1.if_extents =
324 tifp->if_u2.if_inline_ext;
326 tilf_fields |= XFS_ILOG_DEXT;
328 case XFS_DINODE_FMT_BTREE:
329 tilf_fields |= XFS_ILOG_DBROOT;
334 * Increment vnode ref counts since xfs_trans_commit &
335 * xfs_trans_cancel will both unlock the inodes and
336 * decrement the associated ref counts.
341 xfs_trans_ijoin(tp, ip, lock_flags);
342 xfs_trans_ijoin(tp, tip, lock_flags);
344 xfs_trans_log_inode(tp, ip, ilf_fields);
345 xfs_trans_log_inode(tp, tip, tilf_fields);
348 * If this is a synchronous mount, make sure that the
349 * transaction goes to disk before returning to the user.
351 if (mp->m_flags & XFS_MOUNT_WSYNC) {
352 xfs_trans_set_sync(tp);
355 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
360 xfs_iunlock(ip, lock_flags);
361 xfs_iunlock(tip, lock_flags);
370 kmem_free(sxp, sizeof(xfs_swapext_t));
372 kmem_free(tempifp, sizeof(xfs_ifork_t));