blob: 654dc6f05bac7781f288f63fbcb1a5a970b06faa [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
Olaf Weber3e57ecf2006-06-09 14:48:12 +10002 * Copyright (c) 2000-2006 Silicon Graphics, Inc.
Nathan Scott7b718762005-11-02 14:58:39 +11003 * All Rights Reserved.
Linus Torvalds1da177e2005-04-16 15:20:36 -07004 *
Nathan Scott7b718762005-11-02 14:58:39 +11005 * This program is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU General Public License as
Linus Torvalds1da177e2005-04-16 15:20:36 -07007 * published by the Free Software Foundation.
8 *
Nathan Scott7b718762005-11-02 14:58:39 +11009 * 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.
Linus Torvalds1da177e2005-04-16 15:20:36 -070013 *
Nathan Scott7b718762005-11-02 14:58:39 +110014 * 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
Linus Torvalds1da177e2005-04-16 15:20:36 -070017 */
Linus Torvalds1da177e2005-04-16 15:20:36 -070018#include "xfs.h"
Nathan Scotta844f452005-11-02 14:38:42 +110019#include "xfs_fs.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070020#include "xfs_types.h"
Nathan Scotta844f452005-11-02 14:38:42 +110021#include "xfs_bit.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070022#include "xfs_log.h"
Nathan Scotta844f452005-11-02 14:38:42 +110023#include "xfs_inum.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070024#include "xfs_trans.h"
25#include "xfs_sb.h"
Nathan Scotta844f452005-11-02 14:38:42 +110026#include "xfs_ag.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070027#include "xfs_mount.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070028#include "xfs_bmap_btree.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070029#include "xfs_dinode.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070030#include "xfs_inode.h"
Nathan Scotta844f452005-11-02 14:38:42 +110031#include "xfs_inode_item.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070032#include "xfs_bmap.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070033#include "xfs_itable.h"
34#include "xfs_dfrag.h"
35#include "xfs_error.h"
Christoph Hellwig739bfb22007-08-29 10:58:01 +100036#include "xfs_vnodeops.h"
Christoph Hellwig0b1b2132009-12-14 23:14:59 +000037#include "xfs_trace.h"
Linus Torvalds1da177e2005-04-16 15:20:36 -070038
Dave Chinner6bded0f2010-01-14 01:33:56 +000039
40static int xfs_swap_extents(
41 xfs_inode_t *ip, /* target inode */
42 xfs_inode_t *tip, /* tmp inode */
43 xfs_swapext_t *sxp);
44
Linus Torvalds1da177e2005-04-16 15:20:36 -070045/*
Dave Chinner6bded0f2010-01-14 01:33:56 +000046 * ioctl interface for swapext
Linus Torvalds1da177e2005-04-16 15:20:36 -070047 */
48int
49xfs_swapext(
sandeen@sandeen.net743bb4652008-11-25 21:20:06 -060050 xfs_swapext_t *sxp)
Linus Torvalds1da177e2005-04-16 15:20:36 -070051{
Christoph Hellwig35fec8d2008-02-05 12:13:07 +110052 xfs_inode_t *ip, *tip;
Dave Chinner6bded0f2010-01-14 01:33:56 +000053 struct file *file, *tmp_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -070054 int error = 0;
Linus Torvalds1da177e2005-04-16 15:20:36 -070055
Linus Torvalds1da177e2005-04-16 15:20:36 -070056 /* Pull information for the target fd */
Christoph Hellwig35fec8d2008-02-05 12:13:07 +110057 file = fget((int)sxp->sx_fdtarget);
58 if (!file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070059 error = XFS_ERROR(EINVAL);
Eric Sandeenac12b4e2009-01-25 20:53:00 -060060 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -070061 }
62
Dan Rosenberg18171762010-06-24 12:07:47 +100063 if (!(file->f_mode & FMODE_WRITE) ||
64 !(file->f_mode & FMODE_READ) ||
65 (file->f_flags & O_APPEND)) {
Christoph Hellwigf6aa7f22008-02-05 12:13:15 +110066 error = XFS_ERROR(EBADF);
67 goto out_put_file;
68 }
69
Dave Chinner6bded0f2010-01-14 01:33:56 +000070 tmp_file = fget((int)sxp->sx_fdtmp);
71 if (!tmp_file) {
Linus Torvalds1da177e2005-04-16 15:20:36 -070072 error = XFS_ERROR(EINVAL);
Christoph Hellwig35fec8d2008-02-05 12:13:07 +110073 goto out_put_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -070074 }
75
Dave Chinner6bded0f2010-01-14 01:33:56 +000076 if (!(tmp_file->f_mode & FMODE_WRITE) ||
Dan Rosenberg18171762010-06-24 12:07:47 +100077 !(tmp_file->f_mode & FMODE_READ) ||
Dave Chinner6bded0f2010-01-14 01:33:56 +000078 (tmp_file->f_flags & O_APPEND)) {
Christoph Hellwigf6aa7f22008-02-05 12:13:15 +110079 error = XFS_ERROR(EBADF);
Dave Chinner6bded0f2010-01-14 01:33:56 +000080 goto out_put_tmp_file;
Christoph Hellwigf6aa7f22008-02-05 12:13:15 +110081 }
82
Christoph Hellwig7c8f7af2009-02-12 19:56:00 +010083 if (IS_SWAPFILE(file->f_path.dentry->d_inode) ||
Dave Chinner6bded0f2010-01-14 01:33:56 +000084 IS_SWAPFILE(tmp_file->f_path.dentry->d_inode)) {
Christoph Hellwig7c8f7af2009-02-12 19:56:00 +010085 error = XFS_ERROR(EINVAL);
Dave Chinner6bded0f2010-01-14 01:33:56 +000086 goto out_put_tmp_file;
Christoph Hellwig7c8f7af2009-02-12 19:56:00 +010087 }
88
Christoph Hellwig35fec8d2008-02-05 12:13:07 +110089 ip = XFS_I(file->f_path.dentry->d_inode);
Dave Chinner6bded0f2010-01-14 01:33:56 +000090 tip = XFS_I(tmp_file->f_path.dentry->d_inode);
Linus Torvalds1da177e2005-04-16 15:20:36 -070091
92 if (ip->i_mount != tip->i_mount) {
Christoph Hellwig35fec8d2008-02-05 12:13:07 +110093 error = XFS_ERROR(EINVAL);
Dave Chinner6bded0f2010-01-14 01:33:56 +000094 goto out_put_tmp_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -070095 }
96
97 if (ip->i_ino == tip->i_ino) {
Christoph Hellwig35fec8d2008-02-05 12:13:07 +110098 error = XFS_ERROR(EINVAL);
Dave Chinner6bded0f2010-01-14 01:33:56 +000099 goto out_put_tmp_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700100 }
101
Christoph Hellwig35fec8d2008-02-05 12:13:07 +1100102 if (XFS_FORCED_SHUTDOWN(ip->i_mount)) {
103 error = XFS_ERROR(EIO);
Dave Chinner6bded0f2010-01-14 01:33:56 +0000104 goto out_put_tmp_file;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700105 }
106
Lachlan McIlroy541d7d32007-10-11 17:34:33 +1000107 error = xfs_swap_extents(ip, tip, sxp);
Olaf Weber3e57ecf2006-06-09 14:48:12 +1000108
Dave Chinner6bded0f2010-01-14 01:33:56 +0000109 out_put_tmp_file:
110 fput(tmp_file);
Christoph Hellwig35fec8d2008-02-05 12:13:07 +1100111 out_put_file:
112 fput(file);
Christoph Hellwig35fec8d2008-02-05 12:13:07 +1100113 out:
Olaf Weber3e57ecf2006-06-09 14:48:12 +1000114 return error;
115}
116
Dave Chinnere09f9862010-01-14 01:33:54 +0000117/*
118 * We need to check that the format of the data fork in the temporary inode is
119 * valid for the target inode before doing the swap. This is not a problem with
120 * attr1 because of the fixed fork offset, but attr2 has a dynamically sized
121 * data fork depending on the space the attribute fork is taking so we can get
122 * invalid formats on the target inode.
123 *
124 * E.g. target has space for 7 extents in extent format, temp inode only has
125 * space for 6. If we defragment down to 7 extents, then the tmp format is a
126 * btree, but when swapped it needs to be in extent format. Hence we can't just
127 * blindly swap data forks on attr2 filesystems.
128 *
129 * Note that we check the swap in both directions so that we don't end up with
130 * a corrupt temporary inode, either.
131 *
132 * Note that fixing the way xfs_fsr sets up the attribute fork in the source
133 * inode will prevent this situation from occurring, so all we do here is
134 * reject and log the attempt. basically we are putting the responsibility on
135 * userspace to get this right.
136 */
137static int
138xfs_swap_extents_check_format(
139 xfs_inode_t *ip, /* target inode */
140 xfs_inode_t *tip) /* tmp inode */
141{
142
143 /* Should never get a local format */
144 if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL ||
145 tip->i_d.di_format == XFS_DINODE_FMT_LOCAL)
146 return EINVAL;
147
148 /*
149 * if the target inode has less extents that then temporary inode then
150 * why did userspace call us?
151 */
152 if (ip->i_d.di_nextents < tip->i_d.di_nextents)
153 return EINVAL;
154
155 /*
156 * if the target inode is in extent form and the temp inode is in btree
157 * form then we will end up with the target inode in the wrong format
158 * as we already know there are less extents in the temp inode.
159 */
160 if (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS &&
161 tip->i_d.di_format == XFS_DINODE_FMT_BTREE)
162 return EINVAL;
163
164 /* Check temp in extent form to max in target */
165 if (tip->i_d.di_format == XFS_DINODE_FMT_EXTENTS &&
166 XFS_IFORK_NEXTENTS(tip, XFS_DATA_FORK) > ip->i_df.if_ext_max)
167 return EINVAL;
168
169 /* Check target in extent form to max in temp */
170 if (ip->i_d.di_format == XFS_DINODE_FMT_EXTENTS &&
171 XFS_IFORK_NEXTENTS(ip, XFS_DATA_FORK) > tip->i_df.if_ext_max)
172 return EINVAL;
173
Dave Chinnerdd77ef92010-04-20 17:00:37 +1000174 /*
175 * If we are in a btree format, check that the temp root block will fit
176 * in the target and that it has enough extents to be in btree format
177 * in the target.
178 *
179 * Note that we have to be careful to allow btree->extent conversions
180 * (a common defrag case) which will occur when the temp inode is in
181 * extent format...
182 */
Dave Chinnere09f9862010-01-14 01:33:54 +0000183 if (tip->i_d.di_format == XFS_DINODE_FMT_BTREE &&
Dave Chinnerdd77ef92010-04-20 17:00:37 +1000184 ((XFS_IFORK_BOFF(ip) &&
185 tip->i_df.if_broot_bytes > XFS_IFORK_BOFF(ip)) ||
186 XFS_IFORK_NEXTENTS(tip, XFS_DATA_FORK) <= ip->i_df.if_ext_max))
Dave Chinnere09f9862010-01-14 01:33:54 +0000187 return EINVAL;
188
Dave Chinnerdd77ef92010-04-20 17:00:37 +1000189 /* Reciprocal target->temp btree format checks */
Dave Chinnere09f9862010-01-14 01:33:54 +0000190 if (ip->i_d.di_format == XFS_DINODE_FMT_BTREE &&
Dave Chinnerdd77ef92010-04-20 17:00:37 +1000191 ((XFS_IFORK_BOFF(tip) &&
192 ip->i_df.if_broot_bytes > XFS_IFORK_BOFF(tip)) ||
193 XFS_IFORK_NEXTENTS(ip, XFS_DATA_FORK) <= tip->i_df.if_ext_max))
Dave Chinnere09f9862010-01-14 01:33:54 +0000194 return EINVAL;
195
196 return 0;
197}
198
Dave Chinner6bded0f2010-01-14 01:33:56 +0000199static int
Olaf Weber3e57ecf2006-06-09 14:48:12 +1000200xfs_swap_extents(
Dave Chinnere09f9862010-01-14 01:33:54 +0000201 xfs_inode_t *ip, /* target inode */
202 xfs_inode_t *tip, /* tmp inode */
Olaf Weber3e57ecf2006-06-09 14:48:12 +1000203 xfs_swapext_t *sxp)
204{
David Sterba45c51b92011-04-13 22:03:28 +0000205 xfs_mount_t *mp = ip->i_mount;
Olaf Weber3e57ecf2006-06-09 14:48:12 +1000206 xfs_trans_t *tp;
207 xfs_bstat_t *sbp = &sxp->sx_stat;
Olaf Weber3e57ecf2006-06-09 14:48:12 +1000208 xfs_ifork_t *tempifp, *ifp, *tifp;
209 int ilf_fields, tilf_fields;
Olaf Weber3e57ecf2006-06-09 14:48:12 +1000210 int error = 0;
211 int aforkblks = 0;
212 int taforkblks = 0;
213 __uint64_t tmp;
Olaf Weber3e57ecf2006-06-09 14:48:12 +1000214
Olaf Weber3e57ecf2006-06-09 14:48:12 +1000215 tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL);
216 if (!tempifp) {
217 error = XFS_ERROR(ENOMEM);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100218 goto out;
Olaf Weber3e57ecf2006-06-09 14:48:12 +1000219 }
220
David Chinnerf9114eb2008-09-17 16:51:21 +1000221 /*
222 * we have to do two separate lock calls here to keep lockdep
223 * happy. If we try to get all the locks in one call, lock will
224 * report false positives when we drop the ILOCK and regain them
225 * below.
226 */
227 xfs_lock_two_inodes(ip, tip, XFS_IOLOCK_EXCL);
228 xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700229
Linus Torvalds1da177e2005-04-16 15:20:36 -0700230 /* Verify that both files have the same format */
231 if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) {
232 error = XFS_ERROR(EINVAL);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100233 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700234 }
235
236 /* Verify both files are either real-time or non-realtime */
Eric Sandeen71ddabb2007-11-23 16:29:42 +1100237 if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700238 error = XFS_ERROR(EINVAL);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100239 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700240 }
241
Christoph Hellwigdf80c932008-08-13 16:22:09 +1000242 if (VN_CACHED(VFS_I(tip)) != 0) {
Christoph Hellwig739bfb22007-08-29 10:58:01 +1000243 error = xfs_flushinval_pages(tip, 0, -1,
244 FI_REMAPF_LOCKED);
Lachlan McIlroyd3cf2092007-05-08 13:49:27 +1000245 if (error)
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100246 goto out_unlock;
Christoph Hellwigbd5a8762005-06-21 15:47:39 +1000247 }
Linus Torvalds1da177e2005-04-16 15:20:36 -0700248
249 /* Verify O_DIRECT for ftmp */
Christoph Hellwigdf80c932008-08-13 16:22:09 +1000250 if (VN_CACHED(VFS_I(tip)) != 0) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700251 error = XFS_ERROR(EINVAL);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100252 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700253 }
254
255 /* Verify all data are being swapped */
Eric Sandeend0cfb372005-11-02 10:29:04 +1100256 if (sxp->sx_offset != 0 ||
257 sxp->sx_length != ip->i_d.di_size ||
258 sxp->sx_length != tip->i_d.di_size) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700259 error = XFS_ERROR(EFAULT);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100260 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700261 }
262
Dave Chinner3a85cd92010-01-14 01:33:55 +0000263 trace_xfs_swap_extent_before(ip, 0);
264 trace_xfs_swap_extent_before(tip, 1);
265
Dave Chinnere09f9862010-01-14 01:33:54 +0000266 /* check inode formats now that data is flushed */
267 error = xfs_swap_extents_check_format(ip, tip);
268 if (error) {
Dave Chinner53487782011-03-07 10:05:35 +1100269 xfs_notice(mp,
Dave Chinnere09f9862010-01-14 01:33:54 +0000270 "%s: inode 0x%llx format is incompatible for exchanging.",
Dave Chinner53487782011-03-07 10:05:35 +1100271 __func__, ip->i_ino);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100272 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700273 }
274
275 /*
276 * Compare the current change & modify times with that
277 * passed in. If they differ, we abort this swap.
278 * This is the mechanism used to ensure the calling
279 * process that the file was not changed out from
280 * under it.
281 */
Christoph Hellwigf9581b12009-10-06 20:29:26 +0000282 if ((sbp->bs_ctime.tv_sec != VFS_I(ip)->i_ctime.tv_sec) ||
283 (sbp->bs_ctime.tv_nsec != VFS_I(ip)->i_ctime.tv_nsec) ||
284 (sbp->bs_mtime.tv_sec != VFS_I(ip)->i_mtime.tv_sec) ||
285 (sbp->bs_mtime.tv_nsec != VFS_I(ip)->i_mtime.tv_nsec)) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700286 error = XFS_ERROR(EBUSY);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100287 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700288 }
289
290 /* We need to fail if the file is memory mapped. Once we have tossed
291 * all existing pages, the page fault will have no option
292 * but to go to the filesystem for pages. By making the page fault call
Nathan Scott67fcaa72006-06-09 17:00:52 +1000293 * vop_read (or write in the case of autogrow) they block on the iolock
Linus Torvalds1da177e2005-04-16 15:20:36 -0700294 * until we have switched the extents.
295 */
Christoph Hellwigdf80c932008-08-13 16:22:09 +1000296 if (VN_MAPPED(VFS_I(ip))) {
Linus Torvalds1da177e2005-04-16 15:20:36 -0700297 error = XFS_ERROR(EBUSY);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100298 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700299 }
300
301 xfs_iunlock(ip, XFS_ILOCK_EXCL);
302 xfs_iunlock(tip, XFS_ILOCK_EXCL);
303
304 /*
305 * There is a race condition here since we gave up the
306 * ilock. However, the data fork will not change since
307 * we have the iolock (locked for truncation too) so we
308 * are safe. We don't really care if non-io related
309 * fields change.
310 */
311
Christoph Hellwig739bfb22007-08-29 10:58:01 +1000312 xfs_tosspages(ip, 0, -1, FI_REMAPF);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700313
314 tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT);
315 if ((error = xfs_trans_reserve(tp, 0,
316 XFS_ICHANGE_LOG_RES(mp), 0,
317 0, 0))) {
318 xfs_iunlock(ip, XFS_IOLOCK_EXCL);
319 xfs_iunlock(tip, XFS_IOLOCK_EXCL);
320 xfs_trans_cancel(tp, 0);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100321 goto out;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700322 }
Christoph Hellwige1cccd92008-08-13 16:18:07 +1000323 xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700324
325 /*
326 * Count the number of extended attribute blocks
327 */
328 if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) &&
329 (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
330 error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100331 if (error)
332 goto out_trans_cancel;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700333 }
334 if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) &&
335 (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) {
336 error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK,
337 &taforkblks);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100338 if (error)
339 goto out_trans_cancel;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700340 }
341
342 /*
343 * Swap the data forks of the inodes
344 */
345 ifp = &ip->i_df;
346 tifp = &tip->i_df;
Eric Sandeend0cfb372005-11-02 10:29:04 +1100347 *tempifp = *ifp; /* struct copy */
348 *ifp = *tifp; /* struct copy */
349 *tifp = *tempifp; /* struct copy */
Linus Torvalds1da177e2005-04-16 15:20:36 -0700350
351 /*
Dave Chinnere09f9862010-01-14 01:33:54 +0000352 * Fix the in-memory data fork values that are dependent on the fork
353 * offset in the inode. We can't assume they remain the same as attr2
354 * has dynamic fork offsets.
355 */
356 ifp->if_ext_max = XFS_IFORK_SIZE(ip, XFS_DATA_FORK) /
357 (uint)sizeof(xfs_bmbt_rec_t);
358 tifp->if_ext_max = XFS_IFORK_SIZE(tip, XFS_DATA_FORK) /
359 (uint)sizeof(xfs_bmbt_rec_t);
360
361 /*
Linus Torvalds1da177e2005-04-16 15:20:36 -0700362 * Fix the on-disk inode values
363 */
364 tmp = (__uint64_t)ip->i_d.di_nblocks;
365 ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks;
366 tip->i_d.di_nblocks = tmp + taforkblks - aforkblks;
367
368 tmp = (__uint64_t) ip->i_d.di_nextents;
369 ip->i_d.di_nextents = tip->i_d.di_nextents;
370 tip->i_d.di_nextents = tmp;
371
372 tmp = (__uint64_t) ip->i_d.di_format;
373 ip->i_d.di_format = tip->i_d.di_format;
374 tip->i_d.di_format = tmp;
375
Dave Chinner309c8482010-11-30 15:16:02 +1100376 /*
377 * The extents in the source inode could still contain speculative
378 * preallocation beyond EOF (e.g. the file is open but not modified
379 * while defrag is in progress). In that case, we need to copy over the
380 * number of delalloc blocks the data fork in the source inode is
381 * tracking beyond EOF so that when the fork is truncated away when the
382 * temporary inode is unlinked we don't underrun the i_delayed_blks
383 * counter on that inode.
384 */
385 ASSERT(tip->i_delayed_blks == 0);
386 tip->i_delayed_blks = ip->i_delayed_blks;
387 ip->i_delayed_blks = 0;
388
Linus Torvalds1da177e2005-04-16 15:20:36 -0700389 ilf_fields = XFS_ILOG_CORE;
390
391 switch(ip->i_d.di_format) {
392 case XFS_DINODE_FMT_EXTENTS:
393 /* If the extents fit in the inode, fix the
394 * pointer. Otherwise it's already NULL or
395 * pointing to the extent.
396 */
397 if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) {
398 ifp->if_u1.if_extents =
399 ifp->if_u2.if_inline_ext;
400 }
401 ilf_fields |= XFS_ILOG_DEXT;
402 break;
403 case XFS_DINODE_FMT_BTREE:
404 ilf_fields |= XFS_ILOG_DBROOT;
405 break;
406 }
407
408 tilf_fields = XFS_ILOG_CORE;
409
410 switch(tip->i_d.di_format) {
411 case XFS_DINODE_FMT_EXTENTS:
412 /* If the extents fit in the inode, fix the
413 * pointer. Otherwise it's already NULL or
414 * pointing to the extent.
415 */
416 if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) {
417 tifp->if_u1.if_extents =
418 tifp->if_u2.if_inline_ext;
419 }
420 tilf_fields |= XFS_ILOG_DEXT;
421 break;
422 case XFS_DINODE_FMT_BTREE:
423 tilf_fields |= XFS_ILOG_DBROOT;
424 break;
425 }
426
Linus Torvalds1da177e2005-04-16 15:20:36 -0700427
Christoph Hellwigddc34152011-09-19 15:00:54 +0000428 xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
429 xfs_trans_ijoin(tp, tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700430
431 xfs_trans_log_inode(tp, ip, ilf_fields);
432 xfs_trans_log_inode(tp, tip, tilf_fields);
433
434 /*
435 * If this is a synchronous mount, make sure that the
436 * transaction goes to disk before returning to the user.
437 */
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100438 if (mp->m_flags & XFS_MOUNT_WSYNC)
Linus Torvalds1da177e2005-04-16 15:20:36 -0700439 xfs_trans_set_sync(tp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700440
Christoph Hellwig815cb212011-09-26 09:14:34 +0000441 error = xfs_trans_commit(tp, 0);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700442
Dave Chinner3a85cd92010-01-14 01:33:55 +0000443 trace_xfs_swap_extent_after(ip, 0);
444 trace_xfs_swap_extent_after(tip, 1);
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100445out:
446 kmem_free(tempifp);
Linus Torvalds1da177e2005-04-16 15:20:36 -0700447 return error;
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100448
Felix Blyakher1f239202009-05-07 19:49:45 -0500449out_unlock:
450 xfs_iunlock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
451 xfs_iunlock(tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL);
452 goto out;
453
Josef 'Jeff' Sipekef8f7fc2009-02-04 09:37:43 +0100454out_trans_cancel:
455 xfs_trans_cancel(tp, 0);
456 goto out_unlock;
Linus Torvalds1da177e2005-04-16 15:20:36 -0700457}