]>
Commit | Line | Data |
---|---|---|
1da177e4 | 1 | /* |
3e57ecf6 | 2 | * Copyright (c) 2000-2006 Silicon Graphics, Inc. |
7b718769 | 3 | * All Rights Reserved. |
1da177e4 | 4 | * |
7b718769 NS |
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 | |
1da177e4 LT |
7 | * published by the Free Software Foundation. |
8 | * | |
7b718769 NS |
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. | |
1da177e4 | 13 | * |
7b718769 NS |
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 | |
1da177e4 | 17 | */ |
1da177e4 | 18 | #include "xfs.h" |
a844f451 | 19 | #include "xfs_fs.h" |
1da177e4 | 20 | #include "xfs_types.h" |
a844f451 | 21 | #include "xfs_bit.h" |
1da177e4 | 22 | #include "xfs_log.h" |
a844f451 | 23 | #include "xfs_inum.h" |
1da177e4 LT |
24 | #include "xfs_trans.h" |
25 | #include "xfs_sb.h" | |
a844f451 | 26 | #include "xfs_ag.h" |
1da177e4 LT |
27 | #include "xfs_dir2.h" |
28 | #include "xfs_dmapi.h" | |
29 | #include "xfs_mount.h" | |
1da177e4 | 30 | #include "xfs_bmap_btree.h" |
a844f451 | 31 | #include "xfs_alloc_btree.h" |
1da177e4 | 32 | #include "xfs_ialloc_btree.h" |
1da177e4 | 33 | #include "xfs_dir2_sf.h" |
a844f451 | 34 | #include "xfs_attr_sf.h" |
1da177e4 | 35 | #include "xfs_dinode.h" |
1da177e4 | 36 | #include "xfs_inode.h" |
a844f451 | 37 | #include "xfs_inode_item.h" |
1da177e4 | 38 | #include "xfs_bmap.h" |
a844f451 | 39 | #include "xfs_btree.h" |
1da177e4 LT |
40 | #include "xfs_ialloc.h" |
41 | #include "xfs_itable.h" | |
42 | #include "xfs_dfrag.h" | |
43 | #include "xfs_error.h" | |
1da177e4 | 44 | #include "xfs_rw.h" |
739bfb2a | 45 | #include "xfs_vnodeops.h" |
1da177e4 LT |
46 | |
47 | /* | |
48 | * Syssgi interface for swapext | |
49 | */ | |
50 | int | |
51 | xfs_swapext( | |
743bb465 | 52 | xfs_swapext_t *sxp) |
1da177e4 | 53 | { |
35fec8df CH |
54 | xfs_inode_t *ip, *tip; |
55 | struct file *file, *target_file; | |
1da177e4 | 56 | int error = 0; |
1da177e4 | 57 | |
1da177e4 | 58 | /* Pull information for the target fd */ |
35fec8df CH |
59 | file = fget((int)sxp->sx_fdtarget); |
60 | if (!file) { | |
1da177e4 | 61 | error = XFS_ERROR(EINVAL); |
ac12b4e2 | 62 | goto out; |
1da177e4 LT |
63 | } |
64 | ||
f6aa7f21 CH |
65 | if (!(file->f_mode & FMODE_WRITE) || (file->f_flags & O_APPEND)) { |
66 | error = XFS_ERROR(EBADF); | |
67 | goto out_put_file; | |
68 | } | |
69 | ||
35fec8df CH |
70 | target_file = fget((int)sxp->sx_fdtmp); |
71 | if (!target_file) { | |
1da177e4 | 72 | error = XFS_ERROR(EINVAL); |
35fec8df | 73 | goto out_put_file; |
1da177e4 LT |
74 | } |
75 | ||
f6aa7f21 CH |
76 | if (!(target_file->f_mode & FMODE_WRITE) || |
77 | (target_file->f_flags & O_APPEND)) { | |
78 | error = XFS_ERROR(EBADF); | |
79 | goto out_put_target_file; | |
80 | } | |
81 | ||
7c8f7af6 CH |
82 | if (IS_SWAPFILE(file->f_path.dentry->d_inode) || |
83 | IS_SWAPFILE(target_file->f_path.dentry->d_inode)) { | |
84 | error = XFS_ERROR(EINVAL); | |
85 | goto out_put_target_file; | |
86 | } | |
87 | ||
35fec8df CH |
88 | ip = XFS_I(file->f_path.dentry->d_inode); |
89 | tip = XFS_I(target_file->f_path.dentry->d_inode); | |
1da177e4 LT |
90 | |
91 | if (ip->i_mount != tip->i_mount) { | |
35fec8df CH |
92 | error = XFS_ERROR(EINVAL); |
93 | goto out_put_target_file; | |
1da177e4 LT |
94 | } |
95 | ||
96 | if (ip->i_ino == tip->i_ino) { | |
35fec8df CH |
97 | error = XFS_ERROR(EINVAL); |
98 | goto out_put_target_file; | |
1da177e4 LT |
99 | } |
100 | ||
35fec8df CH |
101 | if (XFS_FORCED_SHUTDOWN(ip->i_mount)) { |
102 | error = XFS_ERROR(EIO); | |
103 | goto out_put_target_file; | |
1da177e4 LT |
104 | } |
105 | ||
541d7d3c | 106 | error = xfs_swap_extents(ip, tip, sxp); |
3e57ecf6 | 107 | |
35fec8df CH |
108 | out_put_target_file: |
109 | fput(target_file); | |
110 | out_put_file: | |
111 | fput(file); | |
35fec8df | 112 | out: |
3e57ecf6 OW |
113 | return error; |
114 | } | |
115 | ||
116 | int | |
117 | xfs_swap_extents( | |
118 | xfs_inode_t *ip, | |
119 | xfs_inode_t *tip, | |
120 | xfs_swapext_t *sxp) | |
121 | { | |
122 | xfs_mount_t *mp; | |
3e57ecf6 OW |
123 | xfs_trans_t *tp; |
124 | xfs_bstat_t *sbp = &sxp->sx_stat; | |
3e57ecf6 OW |
125 | xfs_ifork_t *tempifp, *ifp, *tifp; |
126 | int ilf_fields, tilf_fields; | |
3e57ecf6 OW |
127 | int error = 0; |
128 | int aforkblks = 0; | |
129 | int taforkblks = 0; | |
130 | __uint64_t tmp; | |
3e57ecf6 OW |
131 | |
132 | mp = ip->i_mount; | |
133 | ||
134 | tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL); | |
135 | if (!tempifp) { | |
136 | error = XFS_ERROR(ENOMEM); | |
ef8f7fc5 | 137 | goto out; |
3e57ecf6 OW |
138 | } |
139 | ||
140 | sbp = &sxp->sx_stat; | |
1da177e4 | 141 | |
f9114eba DC |
142 | /* |
143 | * we have to do two separate lock calls here to keep lockdep | |
144 | * happy. If we try to get all the locks in one call, lock will | |
145 | * report false positives when we drop the ILOCK and regain them | |
146 | * below. | |
147 | */ | |
148 | xfs_lock_two_inodes(ip, tip, XFS_IOLOCK_EXCL); | |
149 | xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL); | |
1da177e4 | 150 | |
1da177e4 LT |
151 | /* Verify that both files have the same format */ |
152 | if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) { | |
153 | error = XFS_ERROR(EINVAL); | |
ef8f7fc5 | 154 | goto out_unlock; |
1da177e4 LT |
155 | } |
156 | ||
157 | /* Verify both files are either real-time or non-realtime */ | |
71ddabb9 | 158 | if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) { |
1da177e4 | 159 | error = XFS_ERROR(EINVAL); |
ef8f7fc5 | 160 | goto out_unlock; |
1da177e4 LT |
161 | } |
162 | ||
163 | /* Should never get a local format */ | |
164 | if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL || | |
165 | tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) { | |
166 | error = XFS_ERROR(EINVAL); | |
ef8f7fc5 | 167 | goto out_unlock; |
1da177e4 LT |
168 | } |
169 | ||
df80c933 | 170 | if (VN_CACHED(VFS_I(tip)) != 0) { |
613d7043 | 171 | xfs_inval_cached_trace(tip, 0, -1, 0, -1); |
739bfb2a CH |
172 | error = xfs_flushinval_pages(tip, 0, -1, |
173 | FI_REMAPF_LOCKED); | |
d3cf2094 | 174 | if (error) |
ef8f7fc5 | 175 | goto out_unlock; |
bd5a876a | 176 | } |
1da177e4 LT |
177 | |
178 | /* Verify O_DIRECT for ftmp */ | |
df80c933 | 179 | if (VN_CACHED(VFS_I(tip)) != 0) { |
1da177e4 | 180 | error = XFS_ERROR(EINVAL); |
ef8f7fc5 | 181 | goto out_unlock; |
1da177e4 LT |
182 | } |
183 | ||
184 | /* Verify all data are being swapped */ | |
d0cfb373 ES |
185 | if (sxp->sx_offset != 0 || |
186 | sxp->sx_length != ip->i_d.di_size || | |
187 | sxp->sx_length != tip->i_d.di_size) { | |
1da177e4 | 188 | error = XFS_ERROR(EFAULT); |
ef8f7fc5 | 189 | goto out_unlock; |
1da177e4 LT |
190 | } |
191 | ||
192 | /* | |
193 | * If the target has extended attributes, the tmp file | |
194 | * must also in order to ensure the correct data fork | |
195 | * format. | |
196 | */ | |
197 | if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) { | |
198 | error = XFS_ERROR(EINVAL); | |
ef8f7fc5 | 199 | goto out_unlock; |
1da177e4 LT |
200 | } |
201 | ||
202 | /* | |
203 | * Compare the current change & modify times with that | |
204 | * passed in. If they differ, we abort this swap. | |
205 | * This is the mechanism used to ensure the calling | |
206 | * process that the file was not changed out from | |
207 | * under it. | |
208 | */ | |
209 | if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) || | |
210 | (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) || | |
211 | (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) || | |
212 | (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) { | |
213 | error = XFS_ERROR(EBUSY); | |
ef8f7fc5 | 214 | goto out_unlock; |
1da177e4 LT |
215 | } |
216 | ||
217 | /* We need to fail if the file is memory mapped. Once we have tossed | |
218 | * all existing pages, the page fault will have no option | |
219 | * but to go to the filesystem for pages. By making the page fault call | |
67fcaa73 | 220 | * vop_read (or write in the case of autogrow) they block on the iolock |
1da177e4 LT |
221 | * until we have switched the extents. |
222 | */ | |
df80c933 | 223 | if (VN_MAPPED(VFS_I(ip))) { |
1da177e4 | 224 | error = XFS_ERROR(EBUSY); |
ef8f7fc5 | 225 | goto out_unlock; |
1da177e4 LT |
226 | } |
227 | ||
228 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
229 | xfs_iunlock(tip, XFS_ILOCK_EXCL); | |
230 | ||
231 | /* | |
232 | * There is a race condition here since we gave up the | |
233 | * ilock. However, the data fork will not change since | |
234 | * we have the iolock (locked for truncation too) so we | |
235 | * are safe. We don't really care if non-io related | |
236 | * fields change. | |
237 | */ | |
238 | ||
739bfb2a | 239 | xfs_tosspages(ip, 0, -1, FI_REMAPF); |
1da177e4 LT |
240 | |
241 | tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT); | |
242 | if ((error = xfs_trans_reserve(tp, 0, | |
243 | XFS_ICHANGE_LOG_RES(mp), 0, | |
244 | 0, 0))) { | |
245 | xfs_iunlock(ip, XFS_IOLOCK_EXCL); | |
246 | xfs_iunlock(tip, XFS_IOLOCK_EXCL); | |
247 | xfs_trans_cancel(tp, 0); | |
ef8f7fc5 | 248 | goto out; |
1da177e4 | 249 | } |
e1cccd91 | 250 | xfs_lock_two_inodes(ip, tip, XFS_ILOCK_EXCL); |
1da177e4 LT |
251 | |
252 | /* | |
253 | * Count the number of extended attribute blocks | |
254 | */ | |
255 | if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) && | |
256 | (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { | |
257 | error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks); | |
ef8f7fc5 JJS |
258 | if (error) |
259 | goto out_trans_cancel; | |
1da177e4 LT |
260 | } |
261 | if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) && | |
262 | (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { | |
263 | error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK, | |
264 | &taforkblks); | |
ef8f7fc5 JJS |
265 | if (error) |
266 | goto out_trans_cancel; | |
1da177e4 LT |
267 | } |
268 | ||
269 | /* | |
270 | * Swap the data forks of the inodes | |
271 | */ | |
272 | ifp = &ip->i_df; | |
273 | tifp = &tip->i_df; | |
d0cfb373 ES |
274 | *tempifp = *ifp; /* struct copy */ |
275 | *ifp = *tifp; /* struct copy */ | |
276 | *tifp = *tempifp; /* struct copy */ | |
1da177e4 LT |
277 | |
278 | /* | |
279 | * Fix the on-disk inode values | |
280 | */ | |
281 | tmp = (__uint64_t)ip->i_d.di_nblocks; | |
282 | ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks; | |
283 | tip->i_d.di_nblocks = tmp + taforkblks - aforkblks; | |
284 | ||
285 | tmp = (__uint64_t) ip->i_d.di_nextents; | |
286 | ip->i_d.di_nextents = tip->i_d.di_nextents; | |
287 | tip->i_d.di_nextents = tmp; | |
288 | ||
289 | tmp = (__uint64_t) ip->i_d.di_format; | |
290 | ip->i_d.di_format = tip->i_d.di_format; | |
291 | tip->i_d.di_format = tmp; | |
292 | ||
293 | ilf_fields = XFS_ILOG_CORE; | |
294 | ||
295 | switch(ip->i_d.di_format) { | |
296 | case XFS_DINODE_FMT_EXTENTS: | |
297 | /* If the extents fit in the inode, fix the | |
298 | * pointer. Otherwise it's already NULL or | |
299 | * pointing to the extent. | |
300 | */ | |
301 | if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) { | |
302 | ifp->if_u1.if_extents = | |
303 | ifp->if_u2.if_inline_ext; | |
304 | } | |
305 | ilf_fields |= XFS_ILOG_DEXT; | |
306 | break; | |
307 | case XFS_DINODE_FMT_BTREE: | |
308 | ilf_fields |= XFS_ILOG_DBROOT; | |
309 | break; | |
310 | } | |
311 | ||
312 | tilf_fields = XFS_ILOG_CORE; | |
313 | ||
314 | switch(tip->i_d.di_format) { | |
315 | case XFS_DINODE_FMT_EXTENTS: | |
316 | /* If the extents fit in the inode, fix the | |
317 | * pointer. Otherwise it's already NULL or | |
318 | * pointing to the extent. | |
319 | */ | |
320 | if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) { | |
321 | tifp->if_u1.if_extents = | |
322 | tifp->if_u2.if_inline_ext; | |
323 | } | |
324 | tilf_fields |= XFS_ILOG_DEXT; | |
325 | break; | |
326 | case XFS_DINODE_FMT_BTREE: | |
327 | tilf_fields |= XFS_ILOG_DBROOT; | |
328 | break; | |
329 | } | |
330 | ||
1da177e4 | 331 | |
0b1f9177 | 332 | IHOLD(ip); |
ef8f7fc5 | 333 | xfs_trans_ijoin(tp, ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL); |
0b1f9177 CH |
334 | |
335 | IHOLD(tip); | |
ef8f7fc5 | 336 | xfs_trans_ijoin(tp, tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL); |
1da177e4 LT |
337 | |
338 | xfs_trans_log_inode(tp, ip, ilf_fields); | |
339 | xfs_trans_log_inode(tp, tip, tilf_fields); | |
340 | ||
341 | /* | |
342 | * If this is a synchronous mount, make sure that the | |
343 | * transaction goes to disk before returning to the user. | |
344 | */ | |
ef8f7fc5 | 345 | if (mp->m_flags & XFS_MOUNT_WSYNC) |
1da177e4 | 346 | xfs_trans_set_sync(tp); |
1da177e4 | 347 | |
1c72bf90 | 348 | error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT); |
1da177e4 | 349 | |
ef8f7fc5 JJS |
350 | out: |
351 | kmem_free(tempifp); | |
1da177e4 | 352 | return error; |
ef8f7fc5 | 353 | |
1f23920d FB |
354 | out_unlock: |
355 | xfs_iunlock(ip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL); | |
356 | xfs_iunlock(tip, XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL); | |
357 | goto out; | |
358 | ||
ef8f7fc5 JJS |
359 | out_trans_cancel: |
360 | xfs_trans_cancel(tp, 0); | |
361 | goto out_unlock; | |
1da177e4 | 362 | } |