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_mount.h"
29 #include "xfs_bmap_btree.h"
30 #include "xfs_alloc_btree.h"
31 #include "xfs_ialloc_btree.h"
32 #include "xfs_dir2_sf.h"
33 #include "xfs_attr_sf.h"
34 #include "xfs_dinode.h"
35 #include "xfs_inode.h"
36 #include "xfs_inode_item.h"
38 #include "xfs_btree.h"
39 #include "xfs_ialloc.h"
40 #include "xfs_itable.h"
41 #include "xfs_dfrag.h"
42 #include "xfs_error.h"
44 #include "xfs_vnodeops.h"
45 #include "xfs_trace.h"
48 static int xfs_swap_extents(
49 xfs_inode_t *ip, /* target inode */
50 xfs_inode_t *tip, /* tmp inode */
54 * ioctl interface for swapext
60 xfs_inode_t *ip, *tip;
61 struct file *file, *tmp_file;
64 /* Pull information for the target fd */
65 file = fget((int)sxp->sx_fdtarget);
67 error = XFS_ERROR(EINVAL);
71 if (!(file->f_mode & FMODE_WRITE) ||
72 !(file->f_mode & FMODE_READ) ||
73 (file->f_flags & O_APPEND)) {
74 error = XFS_ERROR(EBADF);
78 tmp_file = fget((int)sxp->sx_fdtmp);
80 error = XFS_ERROR(EINVAL);
84 if (!(tmp_file->f_mode & FMODE_WRITE) ||
85 !(tmp_file->f_mode & FMODE_READ) ||
86 (tmp_file->f_flags & O_APPEND)) {
87 error = XFS_ERROR(EBADF);
88 goto out_put_tmp_file;
91 if (IS_SWAPFILE(file->f_path.dentry->d_inode) ||
92 IS_SWAPFILE(tmp_file->f_path.dentry->d_inode)) {
93 error = XFS_ERROR(EINVAL);
94 goto out_put_tmp_file;
97 ip = XFS_I(file->f_path.dentry->d_inode);
98 tip = XFS_I(tmp_file->f_path.dentry->d_inode);
100 if (ip->i_mount != tip->i_mount) {
101 error = XFS_ERROR(EINVAL);
102 goto out_put_tmp_file;
105 if (ip->i_ino == tip->i_ino) {
106 error = XFS_ERROR(EINVAL);
107 goto out_put_tmp_file;
110 if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
111 error = XFS_ERROR(EIO);
112 goto out_put_tmp_file;
115 error = xfs_swap_extents(ip, tip, sxp);
126 * We need to check that the format of the data fork in the temporary inode is
127 * valid for the target inode before doing the swap. This is not a problem with
128 * attr1 because of the fixed fork offset, but attr2 has a dynamically sized
129 * data fork depending on the space the attribute fork is taking so we can get
130 * invalid formats on the target inode.
132 * E.g. target has space for 7 extents in extent format, temp inode only has
133 * space for 6. If we defragment down to 7 extents, then the tmp format is a
134 * btree, but when swapped it needs to be in extent format. Hence we can't just
135 * blindly swap data forks on attr2 filesystems.
137 * Note that we check the swap in both directions so that we don't end up with
138 * a corrupt temporary inode, either.
140 * Note that fixing the way xfs_fsr sets up the attribute fork in the source
141 * inode will prevent this situation from occurring, so all we do here is
142 * reject and log the attempt. basically we are putting the responsibility on
143 * userspace to get this right.
146 xfs_swap_extents_check_format(
147 xfs_inode_t *ip, /* target inode */
148 xfs_inode_t *tip) /* tmp inode */
151 /* Should never get a local format */
152 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
153 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL)
157 * if the target inode has less extents that then temporary inode then
158 * why did userspace call us?
160 if (ip->i_d.di_nextents < tip->i_d.di_nextents)
164 * if the target inode is in extent form and the temp inode is in btree
165 * form then we will end up with the target inode in the wrong format
166 * as we already know there are less extents in the temp inode.
168 if (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS &&
169 tip->i_d.di_format == XFS_DINODE_FMT_BTREE)
172 /* Check temp in extent form to max in target */
173 if (tip->i_d.di_format == XFS_DINODE_FMT_EXTENTS &&
174 XFS_IFORK_NEXTENTS(tip, XFS_DATA_FORK) > ip->i_df.if_ext_max)
177 /* Check target in extent form to max in temp */
178 if (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS &&
179 XFS_IFORK_NEXTENTS(ip, XFS_DATA_FORK) > tip->i_df.if_ext_max)
183 * If we are in a btree format, check that the temp root block will fit
184 * in the target and that it has enough extents to be in btree format
187 * Note that we have to be careful to allow btree->extent conversions
188 * (a common defrag case) which will occur when the temp inode is in
191 if (tip->i_d.di_format == XFS_DINODE_FMT_BTREE &&
192 ((XFS_IFORK_BOFF(ip) &&
193 tip->i_df.if_broot_bytes > XFS_IFORK_BOFF(ip)) ||
194 XFS_IFORK_NEXTENTS(tip, XFS_DATA_FORK) <= ip->i_df.if_ext_max))
197 /* Reciprocal target->temp btree format checks */
198 if (ip->i_d.di_format == XFS_DINODE_FMT_BTREE &&
199 ((XFS_IFORK_BOFF(tip) &&
200 ip->i_df.if_broot_bytes > XFS_IFORK_BOFF(tip)) ||
201 XFS_IFORK_NEXTENTS(ip, XFS_DATA_FORK) <= tip->i_df.if_ext_max))
209 xfs_inode_t *ip, /* target inode */
210 xfs_inode_t *tip, /* tmp inode */
215 xfs_bstat_t *sbp = &sxp->sx_stat;
216 xfs_ifork_t *tempifp, *ifp, *tifp;
217 int ilf_fields, tilf_fields;
225 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
227 error = XFS_ERROR(ENOMEM);
234 * we have to do two separate lock calls here to keep lockdep
235 * happy. If we try to get all the locks in one call, lock will
236 * report false positives when we drop the ILOCK and regain them
239 xfs_lock_two_inodes(ip, tip, XFS_IOLOCK_EXCL);
240 xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
242 /* Verify that both files have the same format */
243 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
244 error = XFS_ERROR(EINVAL);
248 /* Verify both files are either real-time or non-realtime */
249 if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) {
250 error = XFS_ERROR(EINVAL);
254 if (VN_CACHED(VFS_I(tip)) != 0) {
255 error = xfs_flushinval_pages(tip, 0, -1,
261 /* Verify O_DIRECT for ftmp */
262 if (VN_CACHED(VFS_I(tip)) != 0) {
263 error = XFS_ERROR(EINVAL);
267 /* Verify all data are being swapped */
268 if (sxp->sx_offset != 0 ||
269 sxp->sx_length != ip->i_d.di_size ||
270 sxp->sx_length != tip->i_d.di_size) {
271 error = XFS_ERROR(EFAULT);
275 trace_xfs_swap_extent_before(ip, 0);
276 trace_xfs_swap_extent_before(tip, 1);
278 /* check inode formats now that data is flushed */
279 error = xfs_swap_extents_check_format(ip, tip);
281 xfs_fs_cmn_err(CE_NOTE, mp,
282 "%s: inode 0x%llx format is incompatible for exchanging.",
283 __FILE__, ip->i_ino);
288 * Compare the current change & modify times with that
289 * passed in. If they differ, we abort this swap.
290 * This is the mechanism used to ensure the calling
291 * process that the file was not changed out from
294 if ((sbp->bs_ctime.tv_sec != VFS_I(ip)->i_ctime.tv_sec) ||
295 (sbp->bs_ctime.tv_nsec != VFS_I(ip)->i_ctime.tv_nsec) ||
296 (sbp->bs_mtime.tv_sec != VFS_I(ip)->i_mtime.tv_sec) ||
297 (sbp->bs_mtime.tv_nsec != VFS_I(ip)->i_mtime.tv_nsec)) {
298 error = XFS_ERROR(EBUSY);
302 /* We need to fail if the file is memory mapped. Once we have tossed
303 * all existing pages, the page fault will have no option
304 * but to go to the filesystem for pages. By making the page fault call
305 * vop_read (or write in the case of autogrow) they block on the iolock
306 * until we have switched the extents.
308 if (VN_MAPPED(VFS_I(ip))) {
309 error = XFS_ERROR(EBUSY);
313 xfs_iunlock(ip, XFS_ILOCK_EXCL);
314 xfs_iunlock(tip, XFS_ILOCK_EXCL);
317 * There is a race condition here since we gave up the
318 * ilock. However, the data fork will not change since
319 * we have the iolock (locked for truncation too) so we
320 * are safe. We don't really care if non-io related
324 xfs_tosspages(ip, 0, -1, FI_REMAPF);
326 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
327 if ((error = xfs_trans_reserve(tp, 0,
328 XFS_ICHANGE_LOG_RES(mp), 0,
330 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
331 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
332 xfs_trans_cancel(tp, 0);
335 xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
338 * Count the number of extended attribute blocks
340 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
341 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
342 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
344 goto out_trans_cancel;
346 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
347 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
348 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
351 goto out_trans_cancel;
355 * Swap the data forks of the inodes
359 *tempifp = *ifp; /* struct copy */
360 *ifp = *tifp; /* struct copy */
361 *tifp = *tempifp; /* struct copy */
364 * Fix the in-memory data fork values that are dependent on the fork
365 * offset in the inode. We can't assume they remain the same as attr2
366 * has dynamic fork offsets.
368 ifp->if_ext_max = XFS_IFORK_SIZE(ip, XFS_DATA_FORK) /
369 (uint)sizeof(xfs_bmbt_rec_t);
370 tifp->if_ext_max = XFS_IFORK_SIZE(tip, XFS_DATA_FORK) /
371 (uint)sizeof(xfs_bmbt_rec_t);
374 * Fix the on-disk inode values
376 tmp = (__uint64_t)ip->i_d.di_nblocks;
377 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
378 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
380 tmp = (__uint64_t) ip->i_d.di_nextents;
381 ip->i_d.di_nextents = tip->i_d.di_nextents;
382 tip->i_d.di_nextents = tmp;
384 tmp = (__uint64_t) ip->i_d.di_format;
385 ip->i_d.di_format = tip->i_d.di_format;
386 tip->i_d.di_format = tmp;
388 ilf_fields = XFS_ILOG_CORE;
390 switch(ip->i_d.di_format) {
391 case XFS_DINODE_FMT_EXTENTS:
392 /* If the extents fit in the inode, fix the
393 * pointer. Otherwise it's already NULL or
394 * pointing to the extent.
396 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
397 ifp->if_u1.if_extents =
398 ifp->if_u2.if_inline_ext;
400 ilf_fields |= XFS_ILOG_DEXT;
402 case XFS_DINODE_FMT_BTREE:
403 ilf_fields |= XFS_ILOG_DBROOT;
407 tilf_fields = XFS_ILOG_CORE;
409 switch(tip->i_d.di_format) {
410 case XFS_DINODE_FMT_EXTENTS:
411 /* If the extents fit in the inode, fix the
412 * pointer. Otherwise it's already NULL or
413 * pointing to the extent.
415 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
416 tifp->if_u1.if_extents =
417 tifp->if_u2.if_inline_ext;
419 tilf_fields |= XFS_ILOG_DEXT;
421 case XFS_DINODE_FMT_BTREE:
422 tilf_fields |= XFS_ILOG_DBROOT;
428 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
431 xfs_trans_ijoin(tp, tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
433 xfs_trans_log_inode(tp, ip, ilf_fields);
434 xfs_trans_log_inode(tp, tip, tilf_fields);
437 * If this is a synchronous mount, make sure that the
438 * transaction goes to disk before returning to the user.
440 if (mp->m_flags & XFS_MOUNT_WSYNC)
441 xfs_trans_set_sync(tp);
443 error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT);
445 trace_xfs_swap_extent_after(ip, 0);
446 trace_xfs_swap_extent_after(tip, 1);
452 xfs_iunlock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
453 xfs_iunlock(tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
457 xfs_trans_cancel(tp, 0);