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( | |
d0cfb373 | 52 | xfs_swapext_t __user *sxu) |
1da177e4 | 53 | { |
d0cfb373 | 54 | xfs_swapext_t *sxp; |
35fec8df CH |
55 | xfs_inode_t *ip, *tip; |
56 | struct file *file, *target_file; | |
1da177e4 | 57 | int error = 0; |
1da177e4 | 58 | |
d0cfb373 | 59 | sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL); |
3e57ecf6 | 60 | if (!sxp) { |
d0cfb373 | 61 | error = XFS_ERROR(ENOMEM); |
35fec8df | 62 | goto out; |
d0cfb373 ES |
63 | } |
64 | ||
65 | if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) { | |
66 | error = XFS_ERROR(EFAULT); | |
35fec8df | 67 | goto out_free_sxp; |
d0cfb373 | 68 | } |
1da177e4 LT |
69 | |
70 | /* Pull information for the target fd */ | |
35fec8df CH |
71 | file = fget((int)sxp->sx_fdtarget); |
72 | if (!file) { | |
1da177e4 | 73 | error = XFS_ERROR(EINVAL); |
35fec8df | 74 | goto out_free_sxp; |
1da177e4 LT |
75 | } |
76 | ||
35fec8df CH |
77 | target_file = fget((int)sxp->sx_fdtmp); |
78 | if (!target_file) { | |
1da177e4 | 79 | error = XFS_ERROR(EINVAL); |
35fec8df | 80 | goto out_put_file; |
1da177e4 LT |
81 | } |
82 | ||
35fec8df CH |
83 | ip = XFS_I(file->f_path.dentry->d_inode); |
84 | tip = XFS_I(target_file->f_path.dentry->d_inode); | |
1da177e4 LT |
85 | |
86 | if (ip->i_mount != tip->i_mount) { | |
35fec8df CH |
87 | error = XFS_ERROR(EINVAL); |
88 | goto out_put_target_file; | |
1da177e4 LT |
89 | } |
90 | ||
91 | if (ip->i_ino == tip->i_ino) { | |
35fec8df CH |
92 | error = XFS_ERROR(EINVAL); |
93 | goto out_put_target_file; | |
1da177e4 LT |
94 | } |
95 | ||
35fec8df CH |
96 | if (XFS_FORCED_SHUTDOWN(ip->i_mount)) { |
97 | error = XFS_ERROR(EIO); | |
98 | goto out_put_target_file; | |
1da177e4 LT |
99 | } |
100 | ||
541d7d3c | 101 | error = xfs_swap_extents(ip, tip, sxp); |
3e57ecf6 | 102 | |
35fec8df CH |
103 | out_put_target_file: |
104 | fput(target_file); | |
105 | out_put_file: | |
106 | fput(file); | |
107 | out_free_sxp: | |
108 | kmem_free(sxp, sizeof(xfs_swapext_t)); | |
109 | out: | |
3e57ecf6 OW |
110 | return error; |
111 | } | |
112 | ||
113 | int | |
114 | xfs_swap_extents( | |
115 | xfs_inode_t *ip, | |
116 | xfs_inode_t *tip, | |
117 | xfs_swapext_t *sxp) | |
118 | { | |
119 | xfs_mount_t *mp; | |
120 | xfs_inode_t *ips[2]; | |
121 | xfs_trans_t *tp; | |
122 | xfs_bstat_t *sbp = &sxp->sx_stat; | |
67fcaa73 | 123 | bhv_vnode_t *vp, *tvp; |
3e57ecf6 OW |
124 | xfs_ifork_t *tempifp, *ifp, *tifp; |
125 | int ilf_fields, tilf_fields; | |
126 | static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL; | |
127 | int error = 0; | |
128 | int aforkblks = 0; | |
129 | int taforkblks = 0; | |
130 | __uint64_t tmp; | |
131 | char locked = 0; | |
132 | ||
133 | mp = ip->i_mount; | |
134 | ||
135 | tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL); | |
136 | if (!tempifp) { | |
137 | error = XFS_ERROR(ENOMEM); | |
138 | goto error0; | |
139 | } | |
140 | ||
141 | sbp = &sxp->sx_stat; | |
142 | vp = XFS_ITOV(ip); | |
143 | tvp = XFS_ITOV(tip); | |
1da177e4 LT |
144 | |
145 | /* Lock in i_ino order */ | |
146 | if (ip->i_ino < tip->i_ino) { | |
147 | ips[0] = ip; | |
148 | ips[1] = tip; | |
149 | } else { | |
150 | ips[0] = tip; | |
151 | ips[1] = ip; | |
152 | } | |
d0cfb373 | 153 | |
1da177e4 | 154 | xfs_lock_inodes(ips, 2, 0, lock_flags); |
3e57ecf6 | 155 | locked = 1; |
1da177e4 LT |
156 | |
157 | /* Check permissions */ | |
158 | error = xfs_iaccess(ip, S_IWUSR, NULL); | |
159 | if (error) | |
160 | goto error0; | |
161 | ||
162 | error = xfs_iaccess(tip, S_IWUSR, NULL); | |
163 | if (error) | |
164 | goto error0; | |
165 | ||
166 | /* Verify that both files have the same format */ | |
167 | if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) { | |
168 | error = XFS_ERROR(EINVAL); | |
169 | goto error0; | |
170 | } | |
171 | ||
172 | /* Verify both files are either real-time or non-realtime */ | |
71ddabb9 | 173 | if (XFS_IS_REALTIME_INODE(ip) != XFS_IS_REALTIME_INODE(tip)) { |
1da177e4 LT |
174 | error = XFS_ERROR(EINVAL); |
175 | goto error0; | |
176 | } | |
177 | ||
178 | /* Should never get a local format */ | |
179 | if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL || | |
180 | tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) { | |
181 | error = XFS_ERROR(EINVAL); | |
182 | goto error0; | |
183 | } | |
184 | ||
bd5a876a | 185 | if (VN_CACHED(tvp) != 0) { |
613d7043 | 186 | xfs_inval_cached_trace(tip, 0, -1, 0, -1); |
739bfb2a CH |
187 | error = xfs_flushinval_pages(tip, 0, -1, |
188 | FI_REMAPF_LOCKED); | |
d3cf2094 LM |
189 | if (error) |
190 | goto error0; | |
bd5a876a | 191 | } |
1da177e4 LT |
192 | |
193 | /* Verify O_DIRECT for ftmp */ | |
194 | if (VN_CACHED(tvp) != 0) { | |
195 | error = XFS_ERROR(EINVAL); | |
196 | goto error0; | |
197 | } | |
198 | ||
199 | /* Verify all data are being swapped */ | |
d0cfb373 ES |
200 | if (sxp->sx_offset != 0 || |
201 | sxp->sx_length != ip->i_d.di_size || | |
202 | sxp->sx_length != tip->i_d.di_size) { | |
1da177e4 LT |
203 | error = XFS_ERROR(EFAULT); |
204 | goto error0; | |
205 | } | |
206 | ||
207 | /* | |
208 | * If the target has extended attributes, the tmp file | |
209 | * must also in order to ensure the correct data fork | |
210 | * format. | |
211 | */ | |
212 | if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) { | |
213 | error = XFS_ERROR(EINVAL); | |
214 | goto error0; | |
215 | } | |
216 | ||
217 | /* | |
218 | * Compare the current change & modify times with that | |
219 | * passed in. If they differ, we abort this swap. | |
220 | * This is the mechanism used to ensure the calling | |
221 | * process that the file was not changed out from | |
222 | * under it. | |
223 | */ | |
224 | if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) || | |
225 | (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) || | |
226 | (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) || | |
227 | (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) { | |
228 | error = XFS_ERROR(EBUSY); | |
229 | goto error0; | |
230 | } | |
231 | ||
232 | /* We need to fail if the file is memory mapped. Once we have tossed | |
233 | * all existing pages, the page fault will have no option | |
234 | * but to go to the filesystem for pages. By making the page fault call | |
67fcaa73 | 235 | * vop_read (or write in the case of autogrow) they block on the iolock |
1da177e4 LT |
236 | * until we have switched the extents. |
237 | */ | |
238 | if (VN_MAPPED(vp)) { | |
239 | error = XFS_ERROR(EBUSY); | |
240 | goto error0; | |
241 | } | |
242 | ||
243 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
244 | xfs_iunlock(tip, XFS_ILOCK_EXCL); | |
245 | ||
246 | /* | |
247 | * There is a race condition here since we gave up the | |
248 | * ilock. However, the data fork will not change since | |
249 | * we have the iolock (locked for truncation too) so we | |
250 | * are safe. We don't really care if non-io related | |
251 | * fields change. | |
252 | */ | |
253 | ||
739bfb2a | 254 | xfs_tosspages(ip, 0, -1, FI_REMAPF); |
1da177e4 LT |
255 | |
256 | tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT); | |
257 | if ((error = xfs_trans_reserve(tp, 0, | |
258 | XFS_ICHANGE_LOG_RES(mp), 0, | |
259 | 0, 0))) { | |
260 | xfs_iunlock(ip, XFS_IOLOCK_EXCL); | |
261 | xfs_iunlock(tip, XFS_IOLOCK_EXCL); | |
262 | xfs_trans_cancel(tp, 0); | |
d0cfb373 ES |
263 | locked = 0; |
264 | goto error0; | |
1da177e4 LT |
265 | } |
266 | xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL); | |
267 | ||
268 | /* | |
269 | * Count the number of extended attribute blocks | |
270 | */ | |
271 | if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) && | |
272 | (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { | |
273 | error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks); | |
274 | if (error) { | |
1da177e4 | 275 | xfs_trans_cancel(tp, 0); |
d0cfb373 | 276 | goto error0; |
1da177e4 LT |
277 | } |
278 | } | |
279 | if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) && | |
280 | (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { | |
281 | error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK, | |
282 | &taforkblks); | |
283 | if (error) { | |
1da177e4 | 284 | xfs_trans_cancel(tp, 0); |
d0cfb373 | 285 | goto error0; |
1da177e4 LT |
286 | } |
287 | } | |
288 | ||
289 | /* | |
290 | * Swap the data forks of the inodes | |
291 | */ | |
292 | ifp = &ip->i_df; | |
293 | tifp = &tip->i_df; | |
d0cfb373 ES |
294 | *tempifp = *ifp; /* struct copy */ |
295 | *ifp = *tifp; /* struct copy */ | |
296 | *tifp = *tempifp; /* struct copy */ | |
1da177e4 LT |
297 | |
298 | /* | |
299 | * Fix the on-disk inode values | |
300 | */ | |
301 | tmp = (__uint64_t)ip->i_d.di_nblocks; | |
302 | ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks; | |
303 | tip->i_d.di_nblocks = tmp + taforkblks - aforkblks; | |
304 | ||
305 | tmp = (__uint64_t) ip->i_d.di_nextents; | |
306 | ip->i_d.di_nextents = tip->i_d.di_nextents; | |
307 | tip->i_d.di_nextents = tmp; | |
308 | ||
309 | tmp = (__uint64_t) ip->i_d.di_format; | |
310 | ip->i_d.di_format = tip->i_d.di_format; | |
311 | tip->i_d.di_format = tmp; | |
312 | ||
313 | ilf_fields = XFS_ILOG_CORE; | |
314 | ||
315 | switch(ip->i_d.di_format) { | |
316 | case XFS_DINODE_FMT_EXTENTS: | |
317 | /* If the extents fit in the inode, fix the | |
318 | * pointer. Otherwise it's already NULL or | |
319 | * pointing to the extent. | |
320 | */ | |
321 | if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) { | |
322 | ifp->if_u1.if_extents = | |
323 | ifp->if_u2.if_inline_ext; | |
324 | } | |
325 | ilf_fields |= XFS_ILOG_DEXT; | |
326 | break; | |
327 | case XFS_DINODE_FMT_BTREE: | |
328 | ilf_fields |= XFS_ILOG_DBROOT; | |
329 | break; | |
330 | } | |
331 | ||
332 | tilf_fields = XFS_ILOG_CORE; | |
333 | ||
334 | switch(tip->i_d.di_format) { | |
335 | case XFS_DINODE_FMT_EXTENTS: | |
336 | /* If the extents fit in the inode, fix the | |
337 | * pointer. Otherwise it's already NULL or | |
338 | * pointing to the extent. | |
339 | */ | |
340 | if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) { | |
341 | tifp->if_u1.if_extents = | |
342 | tifp->if_u2.if_inline_ext; | |
343 | } | |
344 | tilf_fields |= XFS_ILOG_DEXT; | |
345 | break; | |
346 | case XFS_DINODE_FMT_BTREE: | |
347 | tilf_fields |= XFS_ILOG_DBROOT; | |
348 | break; | |
349 | } | |
350 | ||
351 | /* | |
352 | * Increment vnode ref counts since xfs_trans_commit & | |
353 | * xfs_trans_cancel will both unlock the inodes and | |
354 | * decrement the associated ref counts. | |
355 | */ | |
356 | VN_HOLD(vp); | |
357 | VN_HOLD(tvp); | |
358 | ||
359 | xfs_trans_ijoin(tp, ip, lock_flags); | |
360 | xfs_trans_ijoin(tp, tip, lock_flags); | |
361 | ||
362 | xfs_trans_log_inode(tp, ip, ilf_fields); | |
363 | xfs_trans_log_inode(tp, tip, tilf_fields); | |
364 | ||
365 | /* | |
366 | * If this is a synchronous mount, make sure that the | |
367 | * transaction goes to disk before returning to the user. | |
368 | */ | |
369 | if (mp->m_flags & XFS_MOUNT_WSYNC) { | |
370 | xfs_trans_set_sync(tp); | |
371 | } | |
372 | ||
1c72bf90 | 373 | error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT); |
d0cfb373 | 374 | locked = 0; |
1da177e4 LT |
375 | |
376 | error0: | |
377 | if (locked) { | |
378 | xfs_iunlock(ip, lock_flags); | |
379 | xfs_iunlock(tip, lock_flags); | |
380 | } | |
d0cfb373 ES |
381 | if (tempifp != NULL) |
382 | kmem_free(tempifp, sizeof(xfs_ifork_t)); | |
1da177e4 LT |
383 | return error; |
384 | } |