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"
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;
59 struct file *fp = NULL, *tfp = NULL;
60 bhv_vnode_t *vp, *tvp;
63 sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
65 error = XFS_ERROR(ENOMEM);
69 if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
70 error = XFS_ERROR(EFAULT);
74 /* Pull information for the target fd */
75 if (((fp = fget((int)sxp->sx_fdtarget)) == NULL) ||
76 ((vp = vn_from_inode(fp->f_dentry->d_inode)) == NULL)) {
77 error = XFS_ERROR(EINVAL);
83 error = XFS_ERROR(EBADF);
87 if (((tfp = fget((int)sxp->sx_fdtmp)) == NULL) ||
88 ((tvp = vn_from_inode(tfp->f_dentry->d_inode)) == NULL)) {
89 error = XFS_ERROR(EINVAL);
95 error = XFS_ERROR(EBADF);
99 if (ip->i_mount != tip->i_mount) {
100 error = XFS_ERROR(EINVAL);
104 if (ip->i_ino == tip->i_ino) {
105 error = XFS_ERROR(EINVAL);
111 if (XFS_FORCED_SHUTDOWN(mp)) {
112 error = XFS_ERROR(EIO);
116 error = XFS_SWAP_EXTENTS(mp, &ip->i_iocore, &tip->i_iocore, sxp);
125 kmem_free(sxp, sizeof(xfs_swapext_t));
139 xfs_bstat_t *sbp = &sxp->sx_stat;
140 bhv_vnode_t *vp, *tvp;
141 xfs_ifork_t *tempifp, *ifp, *tifp;
142 int ilf_fields, tilf_fields;
143 static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
152 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
154 error = XFS_ERROR(ENOMEM);
162 /* Lock in i_ino order */
163 if (ip->i_ino < tip->i_ino) {
171 xfs_lock_inodes(ips, 2, 0, lock_flags);
174 /* Check permissions */
175 error = xfs_iaccess(ip, S_IWUSR, NULL);
179 error = xfs_iaccess(tip, S_IWUSR, NULL);
183 /* Verify that both files have the same format */
184 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
185 error = XFS_ERROR(EINVAL);
189 /* Verify both files are either real-time or non-realtime */
190 if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
191 (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
192 error = XFS_ERROR(EINVAL);
196 /* Should never get a local format */
197 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
198 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
199 error = XFS_ERROR(EINVAL);
203 if (VN_CACHED(tvp) != 0) {
204 xfs_inval_cached_trace(&tip->i_iocore, 0, -1, 0, -1);
205 bhv_vop_flushinval_pages(tvp, 0, -1, FI_REMAPF_LOCKED);
208 /* Verify O_DIRECT for ftmp */
209 if (VN_CACHED(tvp) != 0) {
210 error = XFS_ERROR(EINVAL);
214 /* Verify all data are being swapped */
215 if (sxp->sx_offset != 0 ||
216 sxp->sx_length != ip->i_d.di_size ||
217 sxp->sx_length != tip->i_d.di_size) {
218 error = XFS_ERROR(EFAULT);
223 * If the target has extended attributes, the tmp file
224 * must also in order to ensure the correct data fork
227 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
228 error = XFS_ERROR(EINVAL);
233 * Compare the current change & modify times with that
234 * passed in. If they differ, we abort this swap.
235 * This is the mechanism used to ensure the calling
236 * process that the file was not changed out from
239 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
240 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
241 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
242 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
243 error = XFS_ERROR(EBUSY);
247 /* We need to fail if the file is memory mapped. Once we have tossed
248 * all existing pages, the page fault will have no option
249 * but to go to the filesystem for pages. By making the page fault call
250 * vop_read (or write in the case of autogrow) they block on the iolock
251 * until we have switched the extents.
254 error = XFS_ERROR(EBUSY);
258 xfs_iunlock(ip, XFS_ILOCK_EXCL);
259 xfs_iunlock(tip, XFS_ILOCK_EXCL);
262 * There is a race condition here since we gave up the
263 * ilock. However, the data fork will not change since
264 * we have the iolock (locked for truncation too) so we
265 * are safe. We don't really care if non-io related
269 bhv_vop_toss_pages(vp, 0, -1, FI_REMAPF);
271 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
272 if ((error = xfs_trans_reserve(tp, 0,
273 XFS_ICHANGE_LOG_RES(mp), 0,
275 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
276 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
277 xfs_trans_cancel(tp, 0);
281 xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
284 * Count the number of extended attribute blocks
286 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
287 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
288 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
290 xfs_trans_cancel(tp, 0);
294 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
295 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
296 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
299 xfs_trans_cancel(tp, 0);
305 * Swap the data forks of the inodes
309 *tempifp = *ifp; /* struct copy */
310 *ifp = *tifp; /* struct copy */
311 *tifp = *tempifp; /* struct copy */
314 * Fix the on-disk inode values
316 tmp = (__uint64_t)ip->i_d.di_nblocks;
317 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
318 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
320 tmp = (__uint64_t) ip->i_d.di_nextents;
321 ip->i_d.di_nextents = tip->i_d.di_nextents;
322 tip->i_d.di_nextents = tmp;
324 tmp = (__uint64_t) ip->i_d.di_format;
325 ip->i_d.di_format = tip->i_d.di_format;
326 tip->i_d.di_format = tmp;
328 ilf_fields = XFS_ILOG_CORE;
330 switch(ip->i_d.di_format) {
331 case XFS_DINODE_FMT_EXTENTS:
332 /* If the extents fit in the inode, fix the
333 * pointer. Otherwise it's already NULL or
334 * pointing to the extent.
336 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
337 ifp->if_u1.if_extents =
338 ifp->if_u2.if_inline_ext;
340 ilf_fields |= XFS_ILOG_DEXT;
342 case XFS_DINODE_FMT_BTREE:
343 ilf_fields |= XFS_ILOG_DBROOT;
347 tilf_fields = XFS_ILOG_CORE;
349 switch(tip->i_d.di_format) {
350 case XFS_DINODE_FMT_EXTENTS:
351 /* If the extents fit in the inode, fix the
352 * pointer. Otherwise it's already NULL or
353 * pointing to the extent.
355 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
356 tifp->if_u1.if_extents =
357 tifp->if_u2.if_inline_ext;
359 tilf_fields |= XFS_ILOG_DEXT;
361 case XFS_DINODE_FMT_BTREE:
362 tilf_fields |= XFS_ILOG_DBROOT;
367 * Increment vnode ref counts since xfs_trans_commit &
368 * xfs_trans_cancel will both unlock the inodes and
369 * decrement the associated ref counts.
374 xfs_trans_ijoin(tp, ip, lock_flags);
375 xfs_trans_ijoin(tp, tip, lock_flags);
377 xfs_trans_log_inode(tp, ip, ilf_fields);
378 xfs_trans_log_inode(tp, tip, tilf_fields);
381 * If this is a synchronous mount, make sure that the
382 * transaction goes to disk before returning to the user.
384 if (mp->m_flags & XFS_MOUNT_WSYNC) {
385 xfs_trans_set_sync(tp);
388 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
393 xfs_iunlock(ip, lock_flags);
394 xfs_iunlock(tip, lock_flags);
397 kmem_free(tempifp, sizeof(xfs_ifork_t));