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 LT |
44 | #include "xfs_rw.h" |
45 | ||
46 | /* | |
47 | * Syssgi interface for swapext | |
48 | */ | |
49 | int | |
50 | xfs_swapext( | |
d0cfb373 | 51 | xfs_swapext_t __user *sxu) |
1da177e4 | 52 | { |
d0cfb373 | 53 | xfs_swapext_t *sxp; |
3e57ecf6 | 54 | xfs_inode_t *ip=NULL, *tip=NULL; |
1da177e4 | 55 | xfs_mount_t *mp; |
1da177e4 | 56 | struct file *fp = NULL, *tfp = NULL; |
67fcaa73 | 57 | bhv_vnode_t *vp, *tvp; |
1da177e4 | 58 | int error = 0; |
1da177e4 | 59 | |
d0cfb373 | 60 | sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL); |
3e57ecf6 | 61 | if (!sxp) { |
d0cfb373 ES |
62 | error = XFS_ERROR(ENOMEM); |
63 | goto error0; | |
64 | } | |
65 | ||
66 | if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) { | |
67 | error = XFS_ERROR(EFAULT); | |
68 | goto error0; | |
69 | } | |
1da177e4 LT |
70 | |
71 | /* Pull information for the target fd */ | |
d0cfb373 | 72 | if (((fp = fget((int)sxp->sx_fdtarget)) == NULL) || |
e678fb0d | 73 | ((vp = vn_from_inode(fp->f_path.dentry->d_inode)) == NULL)) { |
1da177e4 LT |
74 | error = XFS_ERROR(EINVAL); |
75 | goto error0; | |
76 | } | |
77 | ||
75e17b3c CH |
78 | ip = xfs_vtoi(vp); |
79 | if (ip == NULL) { | |
1da177e4 LT |
80 | error = XFS_ERROR(EBADF); |
81 | goto error0; | |
1da177e4 LT |
82 | } |
83 | ||
d0cfb373 | 84 | if (((tfp = fget((int)sxp->sx_fdtmp)) == NULL) || |
e678fb0d | 85 | ((tvp = vn_from_inode(tfp->f_path.dentry->d_inode)) == NULL)) { |
1da177e4 LT |
86 | error = XFS_ERROR(EINVAL); |
87 | goto error0; | |
88 | } | |
89 | ||
75e17b3c CH |
90 | tip = xfs_vtoi(tvp); |
91 | if (tip == NULL) { | |
1da177e4 LT |
92 | error = XFS_ERROR(EBADF); |
93 | goto error0; | |
1da177e4 LT |
94 | } |
95 | ||
96 | if (ip->i_mount != tip->i_mount) { | |
97 | error = XFS_ERROR(EINVAL); | |
98 | goto error0; | |
99 | } | |
100 | ||
101 | if (ip->i_ino == tip->i_ino) { | |
102 | error = XFS_ERROR(EINVAL); | |
103 | goto error0; | |
104 | } | |
105 | ||
106 | mp = ip->i_mount; | |
107 | ||
1da177e4 LT |
108 | if (XFS_FORCED_SHUTDOWN(mp)) { |
109 | error = XFS_ERROR(EIO); | |
110 | goto error0; | |
111 | } | |
112 | ||
3e57ecf6 OW |
113 | error = XFS_SWAP_EXTENTS(mp, &ip->i_iocore, &tip->i_iocore, sxp); |
114 | ||
115 | error0: | |
116 | if (fp != NULL) | |
117 | fput(fp); | |
118 | if (tfp != NULL) | |
119 | fput(tfp); | |
120 | ||
121 | if (sxp != NULL) | |
122 | kmem_free(sxp, sizeof(xfs_swapext_t)); | |
123 | ||
124 | return error; | |
125 | } | |
126 | ||
127 | int | |
128 | xfs_swap_extents( | |
129 | xfs_inode_t *ip, | |
130 | xfs_inode_t *tip, | |
131 | xfs_swapext_t *sxp) | |
132 | { | |
133 | xfs_mount_t *mp; | |
134 | xfs_inode_t *ips[2]; | |
135 | xfs_trans_t *tp; | |
136 | xfs_bstat_t *sbp = &sxp->sx_stat; | |
67fcaa73 | 137 | bhv_vnode_t *vp, *tvp; |
3e57ecf6 OW |
138 | xfs_ifork_t *tempifp, *ifp, *tifp; |
139 | int ilf_fields, tilf_fields; | |
140 | static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL; | |
141 | int error = 0; | |
142 | int aforkblks = 0; | |
143 | int taforkblks = 0; | |
144 | __uint64_t tmp; | |
145 | char locked = 0; | |
146 | ||
147 | mp = ip->i_mount; | |
148 | ||
149 | tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL); | |
150 | if (!tempifp) { | |
151 | error = XFS_ERROR(ENOMEM); | |
152 | goto error0; | |
153 | } | |
154 | ||
155 | sbp = &sxp->sx_stat; | |
156 | vp = XFS_ITOV(ip); | |
157 | tvp = XFS_ITOV(tip); | |
1da177e4 LT |
158 | |
159 | /* Lock in i_ino order */ | |
160 | if (ip->i_ino < tip->i_ino) { | |
161 | ips[0] = ip; | |
162 | ips[1] = tip; | |
163 | } else { | |
164 | ips[0] = tip; | |
165 | ips[1] = ip; | |
166 | } | |
d0cfb373 | 167 | |
1da177e4 | 168 | xfs_lock_inodes(ips, 2, 0, lock_flags); |
3e57ecf6 | 169 | locked = 1; |
1da177e4 LT |
170 | |
171 | /* Check permissions */ | |
172 | error = xfs_iaccess(ip, S_IWUSR, NULL); | |
173 | if (error) | |
174 | goto error0; | |
175 | ||
176 | error = xfs_iaccess(tip, S_IWUSR, NULL); | |
177 | if (error) | |
178 | goto error0; | |
179 | ||
180 | /* Verify that both files have the same format */ | |
181 | if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) { | |
182 | error = XFS_ERROR(EINVAL); | |
183 | goto error0; | |
184 | } | |
185 | ||
186 | /* Verify both files are either real-time or non-realtime */ | |
187 | if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) != | |
188 | (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) { | |
189 | error = XFS_ERROR(EINVAL); | |
190 | goto error0; | |
191 | } | |
192 | ||
193 | /* Should never get a local format */ | |
194 | if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL || | |
195 | tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) { | |
196 | error = XFS_ERROR(EINVAL); | |
197 | goto error0; | |
198 | } | |
199 | ||
bd5a876a CH |
200 | if (VN_CACHED(tvp) != 0) { |
201 | xfs_inval_cached_trace(&tip->i_iocore, 0, -1, 0, -1); | |
67fcaa73 | 202 | bhv_vop_flushinval_pages(tvp, 0, -1, FI_REMAPF_LOCKED); |
bd5a876a | 203 | } |
1da177e4 LT |
204 | |
205 | /* Verify O_DIRECT for ftmp */ | |
206 | if (VN_CACHED(tvp) != 0) { | |
207 | error = XFS_ERROR(EINVAL); | |
208 | goto error0; | |
209 | } | |
210 | ||
211 | /* Verify all data are being swapped */ | |
d0cfb373 ES |
212 | if (sxp->sx_offset != 0 || |
213 | sxp->sx_length != ip->i_d.di_size || | |
214 | sxp->sx_length != tip->i_d.di_size) { | |
1da177e4 LT |
215 | error = XFS_ERROR(EFAULT); |
216 | goto error0; | |
217 | } | |
218 | ||
219 | /* | |
220 | * If the target has extended attributes, the tmp file | |
221 | * must also in order to ensure the correct data fork | |
222 | * format. | |
223 | */ | |
224 | if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) { | |
225 | error = XFS_ERROR(EINVAL); | |
226 | goto error0; | |
227 | } | |
228 | ||
229 | /* | |
230 | * Compare the current change & modify times with that | |
231 | * passed in. If they differ, we abort this swap. | |
232 | * This is the mechanism used to ensure the calling | |
233 | * process that the file was not changed out from | |
234 | * under it. | |
235 | */ | |
236 | if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) || | |
237 | (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) || | |
238 | (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) || | |
239 | (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) { | |
240 | error = XFS_ERROR(EBUSY); | |
241 | goto error0; | |
242 | } | |
243 | ||
244 | /* We need to fail if the file is memory mapped. Once we have tossed | |
245 | * all existing pages, the page fault will have no option | |
246 | * but to go to the filesystem for pages. By making the page fault call | |
67fcaa73 | 247 | * vop_read (or write in the case of autogrow) they block on the iolock |
1da177e4 LT |
248 | * until we have switched the extents. |
249 | */ | |
250 | if (VN_MAPPED(vp)) { | |
251 | error = XFS_ERROR(EBUSY); | |
252 | goto error0; | |
253 | } | |
254 | ||
255 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
256 | xfs_iunlock(tip, XFS_ILOCK_EXCL); | |
257 | ||
258 | /* | |
259 | * There is a race condition here since we gave up the | |
260 | * ilock. However, the data fork will not change since | |
261 | * we have the iolock (locked for truncation too) so we | |
262 | * are safe. We don't really care if non-io related | |
263 | * fields change. | |
264 | */ | |
265 | ||
67fcaa73 | 266 | bhv_vop_toss_pages(vp, 0, -1, FI_REMAPF); |
1da177e4 LT |
267 | |
268 | tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT); | |
269 | if ((error = xfs_trans_reserve(tp, 0, | |
270 | XFS_ICHANGE_LOG_RES(mp), 0, | |
271 | 0, 0))) { | |
272 | xfs_iunlock(ip, XFS_IOLOCK_EXCL); | |
273 | xfs_iunlock(tip, XFS_IOLOCK_EXCL); | |
274 | xfs_trans_cancel(tp, 0); | |
d0cfb373 ES |
275 | locked = 0; |
276 | goto error0; | |
1da177e4 LT |
277 | } |
278 | xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL); | |
279 | ||
280 | /* | |
281 | * Count the number of extended attribute blocks | |
282 | */ | |
283 | if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) && | |
284 | (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { | |
285 | error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks); | |
286 | if (error) { | |
1da177e4 | 287 | xfs_trans_cancel(tp, 0); |
d0cfb373 | 288 | goto error0; |
1da177e4 LT |
289 | } |
290 | } | |
291 | if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) && | |
292 | (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { | |
293 | error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK, | |
294 | &taforkblks); | |
295 | if (error) { | |
1da177e4 | 296 | xfs_trans_cancel(tp, 0); |
d0cfb373 | 297 | goto error0; |
1da177e4 LT |
298 | } |
299 | } | |
300 | ||
301 | /* | |
302 | * Swap the data forks of the inodes | |
303 | */ | |
304 | ifp = &ip->i_df; | |
305 | tifp = &tip->i_df; | |
d0cfb373 ES |
306 | *tempifp = *ifp; /* struct copy */ |
307 | *ifp = *tifp; /* struct copy */ | |
308 | *tifp = *tempifp; /* struct copy */ | |
1da177e4 LT |
309 | |
310 | /* | |
311 | * Fix the on-disk inode values | |
312 | */ | |
313 | tmp = (__uint64_t)ip->i_d.di_nblocks; | |
314 | ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks; | |
315 | tip->i_d.di_nblocks = tmp + taforkblks - aforkblks; | |
316 | ||
317 | tmp = (__uint64_t) ip->i_d.di_nextents; | |
318 | ip->i_d.di_nextents = tip->i_d.di_nextents; | |
319 | tip->i_d.di_nextents = tmp; | |
320 | ||
321 | tmp = (__uint64_t) ip->i_d.di_format; | |
322 | ip->i_d.di_format = tip->i_d.di_format; | |
323 | tip->i_d.di_format = tmp; | |
324 | ||
325 | ilf_fields = XFS_ILOG_CORE; | |
326 | ||
327 | switch(ip->i_d.di_format) { | |
328 | case XFS_DINODE_FMT_EXTENTS: | |
329 | /* If the extents fit in the inode, fix the | |
330 | * pointer. Otherwise it's already NULL or | |
331 | * pointing to the extent. | |
332 | */ | |
333 | if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) { | |
334 | ifp->if_u1.if_extents = | |
335 | ifp->if_u2.if_inline_ext; | |
336 | } | |
337 | ilf_fields |= XFS_ILOG_DEXT; | |
338 | break; | |
339 | case XFS_DINODE_FMT_BTREE: | |
340 | ilf_fields |= XFS_ILOG_DBROOT; | |
341 | break; | |
342 | } | |
343 | ||
344 | tilf_fields = XFS_ILOG_CORE; | |
345 | ||
346 | switch(tip->i_d.di_format) { | |
347 | case XFS_DINODE_FMT_EXTENTS: | |
348 | /* If the extents fit in the inode, fix the | |
349 | * pointer. Otherwise it's already NULL or | |
350 | * pointing to the extent. | |
351 | */ | |
352 | if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) { | |
353 | tifp->if_u1.if_extents = | |
354 | tifp->if_u2.if_inline_ext; | |
355 | } | |
356 | tilf_fields |= XFS_ILOG_DEXT; | |
357 | break; | |
358 | case XFS_DINODE_FMT_BTREE: | |
359 | tilf_fields |= XFS_ILOG_DBROOT; | |
360 | break; | |
361 | } | |
362 | ||
363 | /* | |
364 | * Increment vnode ref counts since xfs_trans_commit & | |
365 | * xfs_trans_cancel will both unlock the inodes and | |
366 | * decrement the associated ref counts. | |
367 | */ | |
368 | VN_HOLD(vp); | |
369 | VN_HOLD(tvp); | |
370 | ||
371 | xfs_trans_ijoin(tp, ip, lock_flags); | |
372 | xfs_trans_ijoin(tp, tip, lock_flags); | |
373 | ||
374 | xfs_trans_log_inode(tp, ip, ilf_fields); | |
375 | xfs_trans_log_inode(tp, tip, tilf_fields); | |
376 | ||
377 | /* | |
378 | * If this is a synchronous mount, make sure that the | |
379 | * transaction goes to disk before returning to the user. | |
380 | */ | |
381 | if (mp->m_flags & XFS_MOUNT_WSYNC) { | |
382 | xfs_trans_set_sync(tp); | |
383 | } | |
384 | ||
1c72bf90 | 385 | error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT); |
d0cfb373 | 386 | locked = 0; |
1da177e4 LT |
387 | |
388 | error0: | |
389 | if (locked) { | |
390 | xfs_iunlock(ip, lock_flags); | |
391 | xfs_iunlock(tip, lock_flags); | |
392 | } | |
d0cfb373 ES |
393 | if (tempifp != NULL) |
394 | kmem_free(tempifp, sizeof(xfs_ifork_t)); | |
1da177e4 LT |
395 | return error; |
396 | } |