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 bhv_desc_t *bdp, *tbdp;
64 vn_bhv_head_t *bhp, *tbhp;
65 static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
66 int ilf_fields, tilf_fields;
68 xfs_ifork_t *tempifp, *ifp, *tifp;
74 sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
75 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
76 if (!sxp || !tempifp) {
77 error = XFS_ERROR(ENOMEM);
81 if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
82 error = XFS_ERROR(EFAULT);
86 /* Pull information for the target fd */
87 if (((fp = fget((int)sxp->sx_fdtarget)) == NULL) ||
88 ((vp = LINVFS_GET_VP(fp->f_dentry->d_inode)) == NULL)) {
89 error = XFS_ERROR(EINVAL);
93 bhp = VN_BHV_HEAD(vp);
94 bdp = vn_bhv_lookup(bhp, &xfs_vnodeops);
96 error = XFS_ERROR(EBADF);
102 if (((tfp = fget((int)sxp->sx_fdtmp)) == NULL) ||
103 ((tvp = LINVFS_GET_VP(tfp->f_dentry->d_inode)) == NULL)) {
104 error = XFS_ERROR(EINVAL);
108 tbhp = VN_BHV_HEAD(tvp);
109 tbdp = vn_bhv_lookup(tbhp, &xfs_vnodeops);
111 error = XFS_ERROR(EBADF);
114 tip = XFS_BHVTOI(tbdp);
117 if (ip->i_mount != tip->i_mount) {
118 error = XFS_ERROR(EINVAL);
122 if (ip->i_ino == tip->i_ino) {
123 error = XFS_ERROR(EINVAL);
131 if (XFS_FORCED_SHUTDOWN(mp)) {
132 error = XFS_ERROR(EIO);
138 /* Lock in i_ino order */
139 if (ip->i_ino < tip->i_ino) {
147 xfs_lock_inodes(ips, 2, 0, lock_flags);
149 /* Check permissions */
150 error = xfs_iaccess(ip, S_IWUSR, NULL);
154 error = xfs_iaccess(tip, S_IWUSR, NULL);
158 /* Verify that both files have the same format */
159 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
160 error = XFS_ERROR(EINVAL);
164 /* Verify both files are either real-time or non-realtime */
165 if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
166 (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
167 error = XFS_ERROR(EINVAL);
171 /* Should never get a local format */
172 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
173 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
174 error = XFS_ERROR(EINVAL);
178 if (VN_CACHED(tvp) != 0) {
179 xfs_inval_cached_trace(&tip->i_iocore, 0, -1, 0, -1);
180 VOP_FLUSHINVAL_PAGES(tvp, 0, -1, FI_REMAPF_LOCKED);
183 /* Verify O_DIRECT for ftmp */
184 if (VN_CACHED(tvp) != 0) {
185 error = XFS_ERROR(EINVAL);
189 /* Verify all data are being swapped */
190 if (sxp->sx_offset != 0 ||
191 sxp->sx_length != ip->i_d.di_size ||
192 sxp->sx_length != tip->i_d.di_size) {
193 error = XFS_ERROR(EFAULT);
198 * If the target has extended attributes, the tmp file
199 * must also in order to ensure the correct data fork
202 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
203 error = XFS_ERROR(EINVAL);
208 * Compare the current change & modify times with that
209 * passed in. If they differ, we abort this swap.
210 * This is the mechanism used to ensure the calling
211 * process that the file was not changed out from
214 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
215 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
216 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
217 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
218 error = XFS_ERROR(EBUSY);
222 /* We need to fail if the file is memory mapped. Once we have tossed
223 * all existing pages, the page fault will have no option
224 * but to go to the filesystem for pages. By making the page fault call
225 * VOP_READ (or write in the case of autogrow) they block on the iolock
226 * until we have switched the extents.
229 error = XFS_ERROR(EBUSY);
233 xfs_iunlock(ip, XFS_ILOCK_EXCL);
234 xfs_iunlock(tip, XFS_ILOCK_EXCL);
237 * There is a race condition here since we gave up the
238 * ilock. However, the data fork will not change since
239 * we have the iolock (locked for truncation too) so we
240 * are safe. We don't really care if non-io related
244 VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
246 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
247 if ((error = xfs_trans_reserve(tp, 0,
248 XFS_ICHANGE_LOG_RES(mp), 0,
250 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
251 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
252 xfs_trans_cancel(tp, 0);
256 xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
259 * Count the number of extended attribute blocks
261 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
262 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
263 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
265 xfs_trans_cancel(tp, 0);
269 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
270 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
271 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
274 xfs_trans_cancel(tp, 0);
280 * Swap the data forks of the inodes
284 *tempifp = *ifp; /* struct copy */
285 *ifp = *tifp; /* struct copy */
286 *tifp = *tempifp; /* struct copy */
289 * Fix the on-disk inode values
291 tmp = (__uint64_t)ip->i_d.di_nblocks;
292 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
293 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
295 tmp = (__uint64_t) ip->i_d.di_nextents;
296 ip->i_d.di_nextents = tip->i_d.di_nextents;
297 tip->i_d.di_nextents = tmp;
299 tmp = (__uint64_t) ip->i_d.di_format;
300 ip->i_d.di_format = tip->i_d.di_format;
301 tip->i_d.di_format = tmp;
303 ilf_fields = XFS_ILOG_CORE;
305 switch(ip->i_d.di_format) {
306 case XFS_DINODE_FMT_EXTENTS:
307 /* If the extents fit in the inode, fix the
308 * pointer. Otherwise it's already NULL or
309 * pointing to the extent.
311 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
312 ifp->if_u1.if_extents =
313 ifp->if_u2.if_inline_ext;
315 ilf_fields |= XFS_ILOG_DEXT;
317 case XFS_DINODE_FMT_BTREE:
318 ilf_fields |= XFS_ILOG_DBROOT;
322 tilf_fields = XFS_ILOG_CORE;
324 switch(tip->i_d.di_format) {
325 case XFS_DINODE_FMT_EXTENTS:
326 /* If the extents fit in the inode, fix the
327 * pointer. Otherwise it's already NULL or
328 * pointing to the extent.
330 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
331 tifp->if_u1.if_extents =
332 tifp->if_u2.if_inline_ext;
334 tilf_fields |= XFS_ILOG_DEXT;
336 case XFS_DINODE_FMT_BTREE:
337 tilf_fields |= XFS_ILOG_DBROOT;
342 * Increment vnode ref counts since xfs_trans_commit &
343 * xfs_trans_cancel will both unlock the inodes and
344 * decrement the associated ref counts.
349 xfs_trans_ijoin(tp, ip, lock_flags);
350 xfs_trans_ijoin(tp, tip, lock_flags);
352 xfs_trans_log_inode(tp, ip, ilf_fields);
353 xfs_trans_log_inode(tp, tip, tilf_fields);
356 * If this is a synchronous mount, make sure that the
357 * transaction goes to disk before returning to the user.
359 if (mp->m_flags & XFS_MOUNT_WSYNC) {
360 xfs_trans_set_sync(tp);
363 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
368 xfs_iunlock(ip, lock_flags);
369 xfs_iunlock(tip, lock_flags);
378 kmem_free(sxp, sizeof(xfs_swapext_t));
380 kmem_free(tempifp, sizeof(xfs_ifork_t));