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"
48 * Syssgi interface for swapext
52 xfs_swapext_t __user *sxu)
55 xfs_inode_t *ip=NULL, *tip=NULL;
57 struct file *fp = NULL, *tfp = NULL;
58 bhv_vnode_t *vp, *tvp;
61 sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
63 error = XFS_ERROR(ENOMEM);
67 if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
68 error = XFS_ERROR(EFAULT);
72 /* Pull information for the target fd */
73 if (((fp = fget((int)sxp->sx_fdtarget)) == NULL) ||
74 ((vp = vn_from_inode(fp->f_dentry->d_inode)) == NULL)) {
75 error = XFS_ERROR(EINVAL);
81 error = XFS_ERROR(EBADF);
85 if (((tfp = fget((int)sxp->sx_fdtmp)) == NULL) ||
86 ((tvp = vn_from_inode(tfp->f_dentry->d_inode)) == NULL)) {
87 error = XFS_ERROR(EINVAL);
93 error = XFS_ERROR(EBADF);
97 if (ip->i_mount != tip->i_mount) {
98 error = XFS_ERROR(EINVAL);
102 if (ip->i_ino == tip->i_ino) {
103 error = XFS_ERROR(EINVAL);
109 if (XFS_FORCED_SHUTDOWN(mp)) {
110 error = XFS_ERROR(EIO);
114 error = XFS_SWAP_EXTENTS(mp, &ip->i_iocore, &tip->i_iocore, sxp);
123 kmem_free(sxp, sizeof(xfs_swapext_t));
137 xfs_bstat_t *sbp = &sxp->sx_stat;
138 bhv_vnode_t *vp, *tvp;
139 xfs_ifork_t *tempifp, *ifp, *tifp;
140 int ilf_fields, tilf_fields;
141 static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
150 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
152 error = XFS_ERROR(ENOMEM);
160 /* Lock in i_ino order */
161 if (ip->i_ino < tip->i_ino) {
169 xfs_lock_inodes(ips, 2, 0, lock_flags);
172 /* Check permissions */
173 error = xfs_iaccess(ip, S_IWUSR, NULL);
177 error = xfs_iaccess(tip, S_IWUSR, NULL);
181 /* Verify that both files have the same format */
182 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
183 error = XFS_ERROR(EINVAL);
187 /* Verify both files are either real-time or non-realtime */
188 if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
189 (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
190 error = XFS_ERROR(EINVAL);
194 /* Should never get a local format */
195 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
196 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
197 error = XFS_ERROR(EINVAL);
201 if (VN_CACHED(tvp) != 0) {
202 xfs_inval_cached_trace(&tip->i_iocore, 0, -1, 0, -1);
203 bhv_vop_flushinval_pages(tvp, 0, -1, FI_REMAPF_LOCKED);
206 /* Verify O_DIRECT for ftmp */
207 if (VN_CACHED(tvp) != 0) {
208 error = XFS_ERROR(EINVAL);
212 /* Verify all data are being swapped */
213 if (sxp->sx_offset != 0 ||
214 sxp->sx_length != ip->i_d.di_size ||
215 sxp->sx_length != tip->i_d.di_size) {
216 error = XFS_ERROR(EFAULT);
221 * If the target has extended attributes, the tmp file
222 * must also in order to ensure the correct data fork
225 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
226 error = XFS_ERROR(EINVAL);
231 * Compare the current change & modify times with that
232 * passed in. If they differ, we abort this swap.
233 * This is the mechanism used to ensure the calling
234 * process that the file was not changed out from
237 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
238 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
239 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
240 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
241 error = XFS_ERROR(EBUSY);
245 /* We need to fail if the file is memory mapped. Once we have tossed
246 * all existing pages, the page fault will have no option
247 * but to go to the filesystem for pages. By making the page fault call
248 * vop_read (or write in the case of autogrow) they block on the iolock
249 * until we have switched the extents.
252 error = XFS_ERROR(EBUSY);
256 xfs_iunlock(ip, XFS_ILOCK_EXCL);
257 xfs_iunlock(tip, XFS_ILOCK_EXCL);
260 * There is a race condition here since we gave up the
261 * ilock. However, the data fork will not change since
262 * we have the iolock (locked for truncation too) so we
263 * are safe. We don't really care if non-io related
267 bhv_vop_toss_pages(vp, 0, -1, FI_REMAPF);
269 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
270 if ((error = xfs_trans_reserve(tp, 0,
271 XFS_ICHANGE_LOG_RES(mp), 0,
273 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
274 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
275 xfs_trans_cancel(tp, 0);
279 xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
282 * Count the number of extended attribute blocks
284 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
285 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
286 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
288 xfs_trans_cancel(tp, 0);
292 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
293 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
294 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
297 xfs_trans_cancel(tp, 0);
303 * Swap the data forks of the inodes
307 *tempifp = *ifp; /* struct copy */
308 *ifp = *tifp; /* struct copy */
309 *tifp = *tempifp; /* struct copy */
312 * Fix the on-disk inode values
314 tmp = (__uint64_t)ip->i_d.di_nblocks;
315 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
316 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
318 tmp = (__uint64_t) ip->i_d.di_nextents;
319 ip->i_d.di_nextents = tip->i_d.di_nextents;
320 tip->i_d.di_nextents = tmp;
322 tmp = (__uint64_t) ip->i_d.di_format;
323 ip->i_d.di_format = tip->i_d.di_format;
324 tip->i_d.di_format = tmp;
326 ilf_fields = XFS_ILOG_CORE;
328 switch(ip->i_d.di_format) {
329 case XFS_DINODE_FMT_EXTENTS:
330 /* If the extents fit in the inode, fix the
331 * pointer. Otherwise it's already NULL or
332 * pointing to the extent.
334 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
335 ifp->if_u1.if_extents =
336 ifp->if_u2.if_inline_ext;
338 ilf_fields |= XFS_ILOG_DEXT;
340 case XFS_DINODE_FMT_BTREE:
341 ilf_fields |= XFS_ILOG_DBROOT;
345 tilf_fields = XFS_ILOG_CORE;
347 switch(tip->i_d.di_format) {
348 case XFS_DINODE_FMT_EXTENTS:
349 /* If the extents fit in the inode, fix the
350 * pointer. Otherwise it's already NULL or
351 * pointing to the extent.
353 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
354 tifp->if_u1.if_extents =
355 tifp->if_u2.if_inline_ext;
357 tilf_fields |= XFS_ILOG_DEXT;
359 case XFS_DINODE_FMT_BTREE:
360 tilf_fields |= XFS_ILOG_DBROOT;
365 * Increment vnode ref counts since xfs_trans_commit &
366 * xfs_trans_cancel will both unlock the inodes and
367 * decrement the associated ref counts.
372 xfs_trans_ijoin(tp, ip, lock_flags);
373 xfs_trans_ijoin(tp, tip, lock_flags);
375 xfs_trans_log_inode(tp, ip, ilf_fields);
376 xfs_trans_log_inode(tp, tip, tilf_fields);
379 * If this is a synchronous mount, make sure that the
380 * transaction goes to disk before returning to the user.
382 if (mp->m_flags & XFS_MOUNT_WSYNC) {
383 xfs_trans_set_sync(tp);
386 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
391 xfs_iunlock(ip, lock_flags);
392 xfs_iunlock(tip, lock_flags);
395 kmem_free(tempifp, sizeof(xfs_ifork_t));