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"
47 * Syssgi interface for swapext
51 xfs_swapext_t __user *sxu)
54 xfs_inode_t *ip=NULL, *tip=NULL;
56 struct file *fp = NULL, *tfp = NULL;
57 bhv_vnode_t *vp, *tvp;
60 sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
62 error = XFS_ERROR(ENOMEM);
66 if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
67 error = XFS_ERROR(EFAULT);
71 /* Pull information for the target fd */
72 if (((fp = fget((int)sxp->sx_fdtarget)) == NULL) ||
73 ((vp = vn_from_inode(fp->f_path.dentry->d_inode)) == NULL)) {
74 error = XFS_ERROR(EINVAL);
80 error = XFS_ERROR(EBADF);
84 if (((tfp = fget((int)sxp->sx_fdtmp)) == NULL) ||
85 ((tvp = vn_from_inode(tfp->f_path.dentry->d_inode)) == NULL)) {
86 error = XFS_ERROR(EINVAL);
92 error = XFS_ERROR(EBADF);
96 if (ip->i_mount != tip->i_mount) {
97 error = XFS_ERROR(EINVAL);
101 if (ip->i_ino == tip->i_ino) {
102 error = XFS_ERROR(EINVAL);
108 if (XFS_FORCED_SHUTDOWN(mp)) {
109 error = XFS_ERROR(EIO);
113 error = XFS_SWAP_EXTENTS(mp, &ip->i_iocore, &tip->i_iocore, sxp);
122 kmem_free(sxp, sizeof(xfs_swapext_t));
136 xfs_bstat_t *sbp = &sxp->sx_stat;
137 bhv_vnode_t *vp, *tvp;
138 xfs_ifork_t *tempifp, *ifp, *tifp;
139 int ilf_fields, tilf_fields;
140 static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
149 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
151 error = XFS_ERROR(ENOMEM);
159 /* Lock in i_ino order */
160 if (ip->i_ino < tip->i_ino) {
168 xfs_lock_inodes(ips, 2, 0, lock_flags);
171 /* Check permissions */
172 error = xfs_iaccess(ip, S_IWUSR, NULL);
176 error = xfs_iaccess(tip, S_IWUSR, NULL);
180 /* Verify that both files have the same format */
181 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
182 error = XFS_ERROR(EINVAL);
186 /* Verify both files are either real-time or non-realtime */
187 if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
188 (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
189 error = XFS_ERROR(EINVAL);
193 /* Should never get a local format */
194 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
195 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
196 error = XFS_ERROR(EINVAL);
200 if (VN_CACHED(tvp) != 0) {
201 xfs_inval_cached_trace(&tip->i_iocore, 0, -1, 0, -1);
202 bhv_vop_flushinval_pages(tvp, 0, -1, FI_REMAPF_LOCKED);
205 /* Verify O_DIRECT for ftmp */
206 if (VN_CACHED(tvp) != 0) {
207 error = XFS_ERROR(EINVAL);
211 /* Verify all data are being swapped */
212 if (sxp->sx_offset != 0 ||
213 sxp->sx_length != ip->i_d.di_size ||
214 sxp->sx_length != tip->i_d.di_size) {
215 error = XFS_ERROR(EFAULT);
220 * If the target has extended attributes, the tmp file
221 * must also in order to ensure the correct data fork
224 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
225 error = XFS_ERROR(EINVAL);
230 * Compare the current change & modify times with that
231 * passed in. If they differ, we abort this swap.
232 * This is the mechanism used to ensure the calling
233 * process that the file was not changed out from
236 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
237 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
238 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
239 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
240 error = XFS_ERROR(EBUSY);
244 /* We need to fail if the file is memory mapped. Once we have tossed
245 * all existing pages, the page fault will have no option
246 * but to go to the filesystem for pages. By making the page fault call
247 * vop_read (or write in the case of autogrow) they block on the iolock
248 * until we have switched the extents.
251 error = XFS_ERROR(EBUSY);
255 xfs_iunlock(ip, XFS_ILOCK_EXCL);
256 xfs_iunlock(tip, XFS_ILOCK_EXCL);
259 * There is a race condition here since we gave up the
260 * ilock. However, the data fork will not change since
261 * we have the iolock (locked for truncation too) so we
262 * are safe. We don't really care if non-io related
266 bhv_vop_toss_pages(vp, 0, -1, FI_REMAPF);
268 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
269 if ((error = xfs_trans_reserve(tp, 0,
270 XFS_ICHANGE_LOG_RES(mp), 0,
272 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
273 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
274 xfs_trans_cancel(tp, 0);
278 xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
281 * Count the number of extended attribute blocks
283 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
284 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
285 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
287 xfs_trans_cancel(tp, 0);
291 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
292 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
293 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
296 xfs_trans_cancel(tp, 0);
302 * Swap the data forks of the inodes
306 *tempifp = *ifp; /* struct copy */
307 *ifp = *tifp; /* struct copy */
308 *tifp = *tempifp; /* struct copy */
311 * Fix the on-disk inode values
313 tmp = (__uint64_t)ip->i_d.di_nblocks;
314 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
315 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
317 tmp = (__uint64_t) ip->i_d.di_nextents;
318 ip->i_d.di_nextents = tip->i_d.di_nextents;
319 tip->i_d.di_nextents = tmp;
321 tmp = (__uint64_t) ip->i_d.di_format;
322 ip->i_d.di_format = tip->i_d.di_format;
323 tip->i_d.di_format = tmp;
325 ilf_fields = XFS_ILOG_CORE;
327 switch(ip->i_d.di_format) {
328 case XFS_DINODE_FMT_EXTENTS:
329 /* If the extents fit in the inode, fix the
330 * pointer. Otherwise it's already NULL or
331 * pointing to the extent.
333 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
334 ifp->if_u1.if_extents =
335 ifp->if_u2.if_inline_ext;
337 ilf_fields |= XFS_ILOG_DEXT;
339 case XFS_DINODE_FMT_BTREE:
340 ilf_fields |= XFS_ILOG_DBROOT;
344 tilf_fields = XFS_ILOG_CORE;
346 switch(tip->i_d.di_format) {
347 case XFS_DINODE_FMT_EXTENTS:
348 /* If the extents fit in the inode, fix the
349 * pointer. Otherwise it's already NULL or
350 * pointing to the extent.
352 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
353 tifp->if_u1.if_extents =
354 tifp->if_u2.if_inline_ext;
356 tilf_fields |= XFS_ILOG_DEXT;
358 case XFS_DINODE_FMT_BTREE:
359 tilf_fields |= XFS_ILOG_DBROOT;
364 * Increment vnode ref counts since xfs_trans_commit &
365 * xfs_trans_cancel will both unlock the inodes and
366 * decrement the associated ref counts.
371 xfs_trans_ijoin(tp, ip, lock_flags);
372 xfs_trans_ijoin(tp, tip, lock_flags);
374 xfs_trans_log_inode(tp, ip, ilf_fields);
375 xfs_trans_log_inode(tp, tip, tilf_fields);
378 * If this is a synchronous mount, make sure that the
379 * transaction goes to disk before returning to the user.
381 if (mp->m_flags & XFS_MOUNT_WSYNC) {
382 xfs_trans_set_sync(tp);
385 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
390 xfs_iunlock(ip, lock_flags);
391 xfs_iunlock(tip, lock_flags);
394 kmem_free(tempifp, sizeof(xfs_ifork_t));