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"
49 #include <sys/capability.h>
53 * Syssgi interface for swapext
57 xfs_swapext_t __user *sxu)
60 xfs_inode_t *ip=NULL, *tip=NULL;
62 xfs_vnode_t *vp = NULL, *tvp = NULL;
63 struct vnode *bvp, *btvp;
66 sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL);
68 error = XFS_ERROR(ENOMEM);
77 if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) {
78 error = XFS_ERROR(EFAULT);
82 /* Pull information for the target fd */
83 if (fgetvp(td, (int)sxp->sx_fdtarget, CAP_READ | CAP_WRITE, &bvp)
85 error = XFS_ERROR(EINVAL);
92 error = XFS_ERROR(EBADF);
96 if (fgetvp(td, (int)sxp->sx_fdtmp, CAP_READ | CAP_WRITE, &btvp) != 0) {
97 error = XFS_ERROR(EINVAL);
101 tvp = VPTOXFSVP(btvp);
104 error = XFS_ERROR(EBADF);
108 if (ip->i_mount != tip->i_mount) {
109 error = XFS_ERROR(EINVAL);
113 if (ip->i_ino == tip->i_ino) {
114 error = XFS_ERROR(EINVAL);
120 if (XFS_FORCED_SHUTDOWN(mp)) {
121 error = XFS_ERROR(EIO);
125 error = XFS_SWAP_EXTENTS(mp, &ip->i_iocore, &tip->i_iocore, sxp);
136 kmem_free(sxp, sizeof(xfs_swapext_t));
150 xfs_bstat_t *sbp = &sxp->sx_stat;
151 xfs_vnode_t *vp, *tvp;
152 xfs_ifork_t *tempifp, *ifp, *tifp;
153 int ilf_fields, tilf_fields;
154 static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL;
163 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
165 error = XFS_ERROR(ENOMEM);
173 /* Lock in i_ino order */
174 if (ip->i_ino < tip->i_ino) {
182 xfs_lock_inodes(ips, 2, 0, lock_flags);
185 /* Check permissions */
186 error = xfs_iaccess(ip, VWRITE, NULL);
190 error = xfs_iaccess(tip, VWRITE, NULL);
194 /* Verify that both files have the same format */
195 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
196 error = XFS_ERROR(EINVAL);
200 /* Verify both files are either real-time or non-realtime */
201 if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) !=
202 (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) {
203 error = XFS_ERROR(EINVAL);
207 /* Should never get a local format */
208 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
209 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) {
210 error = XFS_ERROR(EINVAL);
214 if (VN_CACHED(tvp) != 0) {
215 xfs_inval_cached_trace(&tip->i_iocore, 0, -1, 0, -1);
216 XVOP_FLUSHINVAL_PAGES(tvp, 0, -1, FI_REMAPF_LOCKED);
219 /* Verify O_DIRECT for ftmp */
220 if (VN_CACHED(tvp) != 0) {
221 error = XFS_ERROR(EINVAL);
225 /* Verify all data are being swapped */
226 if (sxp->sx_offset != 0 ||
227 sxp->sx_length != ip->i_d.di_size ||
228 sxp->sx_length != tip->i_d.di_size) {
229 error = XFS_ERROR(EFAULT);
234 * If the target has extended attributes, the tmp file
235 * must also in order to ensure the correct data fork
238 if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) {
239 error = XFS_ERROR(EINVAL);
244 * Compare the current change & modify times with that
245 * passed in. If they differ, we abort this swap.
246 * This is the mechanism used to ensure the calling
247 * process that the file was not changed out from
250 if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) ||
251 (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) ||
252 (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) ||
253 (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) {
254 error = XFS_ERROR(EBUSY);
258 /* We need to fail if the file is memory mapped. Once we have tossed
259 * all existing pages, the page fault will have no option
260 * but to go to the filesystem for pages. By making the page fault call
261 * VOP_READ (or write in the case of autogrow) they block on the iolock
262 * until we have switched the extents.
265 error = XFS_ERROR(EBUSY);
269 xfs_iunlock(ip, XFS_ILOCK_EXCL);
270 xfs_iunlock(tip, XFS_ILOCK_EXCL);
273 * There is a race condition here since we gave up the
274 * ilock. However, the data fork will not change since
275 * we have the iolock (locked for truncation too) so we
276 * are safe. We don't really care if non-io related
280 XVOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF);
282 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
283 if ((error = xfs_trans_reserve(tp, 0,
284 XFS_ICHANGE_LOG_RES(mp), 0,
286 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
287 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
288 xfs_trans_cancel(tp, 0);
292 xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL);
295 * Count the number of extended attribute blocks
297 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
298 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
299 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
301 xfs_trans_cancel(tp, 0);
305 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
306 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
307 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
310 xfs_trans_cancel(tp, 0);
316 * Swap the data forks of the inodes
320 *tempifp = *ifp; /* struct copy */
321 *ifp = *tifp; /* struct copy */
322 *tifp = *tempifp; /* struct copy */
325 * Fix the on-disk inode values
327 tmp = (__uint64_t)ip->i_d.di_nblocks;
328 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
329 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
331 tmp = (__uint64_t) ip->i_d.di_nextents;
332 ip->i_d.di_nextents = tip->i_d.di_nextents;
333 tip->i_d.di_nextents = tmp;
335 tmp = (__uint64_t) ip->i_d.di_format;
336 ip->i_d.di_format = tip->i_d.di_format;
337 tip->i_d.di_format = tmp;
339 ilf_fields = XFS_ILOG_CORE;
341 switch(ip->i_d.di_format) {
342 case XFS_DINODE_FMT_EXTENTS:
343 /* If the extents fit in the inode, fix the
344 * pointer. Otherwise it's already NULL or
345 * pointing to the extent.
347 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
348 ifp->if_u1.if_extents =
349 ifp->if_u2.if_inline_ext;
351 ilf_fields |= XFS_ILOG_DEXT;
353 case XFS_DINODE_FMT_BTREE:
354 ilf_fields |= XFS_ILOG_DBROOT;
358 tilf_fields = XFS_ILOG_CORE;
360 switch(tip->i_d.di_format) {
361 case XFS_DINODE_FMT_EXTENTS:
362 /* If the extents fit in the inode, fix the
363 * pointer. Otherwise it's already NULL or
364 * pointing to the extent.
366 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
367 tifp->if_u1.if_extents =
368 tifp->if_u2.if_inline_ext;
370 tilf_fields |= XFS_ILOG_DEXT;
372 case XFS_DINODE_FMT_BTREE:
373 tilf_fields |= XFS_ILOG_DBROOT;
377 #ifdef XXXKAN /* Not necessary, vnodes are vrefed already by fgetvp */
379 * Increment vnode ref counts since xfs_trans_commit &
380 * xfs_trans_cancel will both unlock the inodes and
381 * decrement the associated ref counts.
387 xfs_trans_ijoin(tp, ip, lock_flags);
388 xfs_trans_ijoin(tp, tip, lock_flags);
390 xfs_trans_log_inode(tp, ip, ilf_fields);
391 xfs_trans_log_inode(tp, tip, tilf_fields);
394 * If this is a synchronous mount, make sure that the
395 * transaction goes to disk before returning to the user.
397 if (mp->m_flags & XFS_MOUNT_WSYNC) {
398 xfs_trans_set_sync(tp);
401 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL);
407 xfs_iunlock(ip, lock_flags);
408 xfs_iunlock(tip, lock_flags);
411 kmem_free(tempifp, sizeof(xfs_ifork_t));