Commit | Line | Data |
---|---|---|
1da177e4 LT |
1 | /* |
2 | * Copyright (c) 2000-2002 Silicon Graphics, Inc. All Rights Reserved. | |
3 | * | |
4 | * This program is free software; you can redistribute it and/or modify it | |
5 | * under the terms of version 2 of the GNU General Public License as | |
6 | * published by the Free Software Foundation. | |
7 | * | |
8 | * This program is distributed in the hope that it would be useful, but | |
9 | * WITHOUT ANY WARRANTY; without even the implied warranty of | |
10 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. | |
11 | * | |
12 | * Further, this software is distributed without any warranty that it is | |
13 | * free of the rightful claim of any third person regarding infringement | |
14 | * or the like. Any license provided herein, whether implied or | |
15 | * otherwise, applies only to this software file. Patent licenses, if | |
16 | * any, provided herein do not apply to combinations of this program with | |
17 | * other software, or any other product whatsoever. | |
18 | * | |
19 | * You should have received a copy of the GNU General Public License along | |
20 | * with this program; if not, write the Free Software Foundation, Inc., 59 | |
21 | * Temple Place - Suite 330, Boston MA 02111-1307, USA. | |
22 | * | |
23 | * Contact information: Silicon Graphics, Inc., 1600 Amphitheatre Pkwy, | |
24 | * Mountain View, CA 94043, or: | |
25 | * | |
26 | * http://www.sgi.com | |
27 | * | |
28 | * For further information regarding this notice, see: | |
29 | * | |
30 | * http://oss.sgi.com/projects/GenInfo/SGIGPLNoticeExplan/ | |
31 | */ | |
1da177e4 | 32 | #include "xfs.h" |
a844f451 | 33 | #include "xfs_fs.h" |
1da177e4 | 34 | #include "xfs_types.h" |
a844f451 | 35 | #include "xfs_bit.h" |
1da177e4 | 36 | #include "xfs_log.h" |
a844f451 | 37 | #include "xfs_inum.h" |
1da177e4 LT |
38 | #include "xfs_trans.h" |
39 | #include "xfs_sb.h" | |
a844f451 | 40 | #include "xfs_ag.h" |
1da177e4 LT |
41 | #include "xfs_dir.h" |
42 | #include "xfs_dir2.h" | |
43 | #include "xfs_dmapi.h" | |
44 | #include "xfs_mount.h" | |
1da177e4 | 45 | #include "xfs_bmap_btree.h" |
a844f451 | 46 | #include "xfs_alloc_btree.h" |
1da177e4 | 47 | #include "xfs_ialloc_btree.h" |
1da177e4 LT |
48 | #include "xfs_dir_sf.h" |
49 | #include "xfs_dir2_sf.h" | |
a844f451 | 50 | #include "xfs_attr_sf.h" |
1da177e4 | 51 | #include "xfs_dinode.h" |
1da177e4 | 52 | #include "xfs_inode.h" |
a844f451 | 53 | #include "xfs_inode_item.h" |
1da177e4 | 54 | #include "xfs_bmap.h" |
a844f451 | 55 | #include "xfs_btree.h" |
1da177e4 LT |
56 | #include "xfs_ialloc.h" |
57 | #include "xfs_itable.h" | |
58 | #include "xfs_dfrag.h" | |
59 | #include "xfs_error.h" | |
60 | #include "xfs_mac.h" | |
61 | #include "xfs_rw.h" | |
62 | ||
63 | /* | |
64 | * Syssgi interface for swapext | |
65 | */ | |
66 | int | |
67 | xfs_swapext( | |
d0cfb373 | 68 | xfs_swapext_t __user *sxu) |
1da177e4 | 69 | { |
d0cfb373 | 70 | xfs_swapext_t *sxp; |
1da177e4 LT |
71 | xfs_inode_t *ip=NULL, *tip=NULL, *ips[2]; |
72 | xfs_trans_t *tp; | |
73 | xfs_mount_t *mp; | |
74 | xfs_bstat_t *sbp; | |
75 | struct file *fp = NULL, *tfp = NULL; | |
76 | vnode_t *vp, *tvp; | |
77 | bhv_desc_t *bdp, *tbdp; | |
78 | vn_bhv_head_t *bhp, *tbhp; | |
d0cfb373 | 79 | static uint lock_flags = XFS_ILOCK_EXCL | XFS_IOLOCK_EXCL; |
1da177e4 LT |
80 | int ilf_fields, tilf_fields; |
81 | int error = 0; | |
d0cfb373 | 82 | xfs_ifork_t *tempifp, *ifp, *tifp; |
1da177e4 LT |
83 | __uint64_t tmp; |
84 | int aforkblks = 0; | |
85 | int taforkblks = 0; | |
d0cfb373 | 86 | char locked = 0; |
1da177e4 | 87 | |
d0cfb373 ES |
88 | sxp = kmem_alloc(sizeof(xfs_swapext_t), KM_MAYFAIL); |
89 | tempifp = kmem_alloc(sizeof(xfs_ifork_t), KM_MAYFAIL); | |
90 | if (!sxp || !tempifp) { | |
91 | error = XFS_ERROR(ENOMEM); | |
92 | goto error0; | |
93 | } | |
94 | ||
95 | if (copy_from_user(sxp, sxu, sizeof(xfs_swapext_t))) { | |
96 | error = XFS_ERROR(EFAULT); | |
97 | goto error0; | |
98 | } | |
1da177e4 LT |
99 | |
100 | /* Pull information for the target fd */ | |
d0cfb373 | 101 | if (((fp = fget((int)sxp->sx_fdtarget)) == NULL) || |
1da177e4 LT |
102 | ((vp = LINVFS_GET_VP(fp->f_dentry->d_inode)) == NULL)) { |
103 | error = XFS_ERROR(EINVAL); | |
104 | goto error0; | |
105 | } | |
106 | ||
107 | bhp = VN_BHV_HEAD(vp); | |
108 | bdp = vn_bhv_lookup(bhp, &xfs_vnodeops); | |
109 | if (bdp == NULL) { | |
110 | error = XFS_ERROR(EBADF); | |
111 | goto error0; | |
112 | } else { | |
113 | ip = XFS_BHVTOI(bdp); | |
114 | } | |
115 | ||
d0cfb373 | 116 | if (((tfp = fget((int)sxp->sx_fdtmp)) == NULL) || |
1da177e4 LT |
117 | ((tvp = LINVFS_GET_VP(tfp->f_dentry->d_inode)) == NULL)) { |
118 | error = XFS_ERROR(EINVAL); | |
119 | goto error0; | |
120 | } | |
121 | ||
122 | tbhp = VN_BHV_HEAD(tvp); | |
123 | tbdp = vn_bhv_lookup(tbhp, &xfs_vnodeops); | |
124 | if (tbdp == NULL) { | |
125 | error = XFS_ERROR(EBADF); | |
126 | goto error0; | |
127 | } else { | |
128 | tip = XFS_BHVTOI(tbdp); | |
129 | } | |
130 | ||
131 | if (ip->i_mount != tip->i_mount) { | |
132 | error = XFS_ERROR(EINVAL); | |
133 | goto error0; | |
134 | } | |
135 | ||
136 | if (ip->i_ino == tip->i_ino) { | |
137 | error = XFS_ERROR(EINVAL); | |
138 | goto error0; | |
139 | } | |
140 | ||
141 | mp = ip->i_mount; | |
142 | ||
d0cfb373 | 143 | sbp = &sxp->sx_stat; |
1da177e4 LT |
144 | |
145 | if (XFS_FORCED_SHUTDOWN(mp)) { | |
146 | error = XFS_ERROR(EIO); | |
147 | goto error0; | |
148 | } | |
149 | ||
150 | locked = 1; | |
151 | ||
152 | /* Lock in i_ino order */ | |
153 | if (ip->i_ino < tip->i_ino) { | |
154 | ips[0] = ip; | |
155 | ips[1] = tip; | |
156 | } else { | |
157 | ips[0] = tip; | |
158 | ips[1] = ip; | |
159 | } | |
d0cfb373 | 160 | |
1da177e4 LT |
161 | xfs_lock_inodes(ips, 2, 0, lock_flags); |
162 | ||
163 | /* Check permissions */ | |
164 | error = xfs_iaccess(ip, S_IWUSR, NULL); | |
165 | if (error) | |
166 | goto error0; | |
167 | ||
168 | error = xfs_iaccess(tip, S_IWUSR, NULL); | |
169 | if (error) | |
170 | goto error0; | |
171 | ||
172 | /* Verify that both files have the same format */ | |
173 | if ((ip->i_d.di_mode & S_IFMT) != (tip->i_d.di_mode & S_IFMT)) { | |
174 | error = XFS_ERROR(EINVAL); | |
175 | goto error0; | |
176 | } | |
177 | ||
178 | /* Verify both files are either real-time or non-realtime */ | |
179 | if ((ip->i_d.di_flags & XFS_DIFLAG_REALTIME) != | |
180 | (tip->i_d.di_flags & XFS_DIFLAG_REALTIME)) { | |
181 | error = XFS_ERROR(EINVAL); | |
182 | goto error0; | |
183 | } | |
184 | ||
185 | /* Should never get a local format */ | |
186 | if (ip->i_d.di_format == XFS_DINODE_FMT_LOCAL || | |
187 | tip->i_d.di_format == XFS_DINODE_FMT_LOCAL) { | |
188 | error = XFS_ERROR(EINVAL); | |
189 | goto error0; | |
190 | } | |
191 | ||
bd5a876a CH |
192 | if (VN_CACHED(tvp) != 0) { |
193 | xfs_inval_cached_trace(&tip->i_iocore, 0, -1, 0, -1); | |
194 | VOP_FLUSHINVAL_PAGES(tvp, 0, -1, FI_REMAPF_LOCKED); | |
195 | } | |
1da177e4 LT |
196 | |
197 | /* Verify O_DIRECT for ftmp */ | |
198 | if (VN_CACHED(tvp) != 0) { | |
199 | error = XFS_ERROR(EINVAL); | |
200 | goto error0; | |
201 | } | |
202 | ||
203 | /* Verify all data are being swapped */ | |
d0cfb373 ES |
204 | if (sxp->sx_offset != 0 || |
205 | sxp->sx_length != ip->i_d.di_size || | |
206 | sxp->sx_length != tip->i_d.di_size) { | |
1da177e4 LT |
207 | error = XFS_ERROR(EFAULT); |
208 | goto error0; | |
209 | } | |
210 | ||
211 | /* | |
212 | * If the target has extended attributes, the tmp file | |
213 | * must also in order to ensure the correct data fork | |
214 | * format. | |
215 | */ | |
216 | if ( XFS_IFORK_Q(ip) != XFS_IFORK_Q(tip) ) { | |
217 | error = XFS_ERROR(EINVAL); | |
218 | goto error0; | |
219 | } | |
220 | ||
221 | /* | |
222 | * Compare the current change & modify times with that | |
223 | * passed in. If they differ, we abort this swap. | |
224 | * This is the mechanism used to ensure the calling | |
225 | * process that the file was not changed out from | |
226 | * under it. | |
227 | */ | |
228 | if ((sbp->bs_ctime.tv_sec != ip->i_d.di_ctime.t_sec) || | |
229 | (sbp->bs_ctime.tv_nsec != ip->i_d.di_ctime.t_nsec) || | |
230 | (sbp->bs_mtime.tv_sec != ip->i_d.di_mtime.t_sec) || | |
231 | (sbp->bs_mtime.tv_nsec != ip->i_d.di_mtime.t_nsec)) { | |
232 | error = XFS_ERROR(EBUSY); | |
233 | goto error0; | |
234 | } | |
235 | ||
236 | /* We need to fail if the file is memory mapped. Once we have tossed | |
237 | * all existing pages, the page fault will have no option | |
238 | * but to go to the filesystem for pages. By making the page fault call | |
239 | * VOP_READ (or write in the case of autogrow) they block on the iolock | |
240 | * until we have switched the extents. | |
241 | */ | |
242 | if (VN_MAPPED(vp)) { | |
243 | error = XFS_ERROR(EBUSY); | |
244 | goto error0; | |
245 | } | |
246 | ||
247 | xfs_iunlock(ip, XFS_ILOCK_EXCL); | |
248 | xfs_iunlock(tip, XFS_ILOCK_EXCL); | |
249 | ||
250 | /* | |
251 | * There is a race condition here since we gave up the | |
252 | * ilock. However, the data fork will not change since | |
253 | * we have the iolock (locked for truncation too) so we | |
254 | * are safe. We don't really care if non-io related | |
255 | * fields change. | |
256 | */ | |
257 | ||
258 | VOP_TOSS_PAGES(vp, 0, -1, FI_REMAPF); | |
259 | ||
260 | tp = xfs_trans_alloc(mp, XFS_TRANS_SWAPEXT); | |
261 | if ((error = xfs_trans_reserve(tp, 0, | |
262 | XFS_ICHANGE_LOG_RES(mp), 0, | |
263 | 0, 0))) { | |
264 | xfs_iunlock(ip, XFS_IOLOCK_EXCL); | |
265 | xfs_iunlock(tip, XFS_IOLOCK_EXCL); | |
266 | xfs_trans_cancel(tp, 0); | |
d0cfb373 ES |
267 | locked = 0; |
268 | goto error0; | |
1da177e4 LT |
269 | } |
270 | xfs_lock_inodes(ips, 2, 0, XFS_ILOCK_EXCL); | |
271 | ||
272 | /* | |
273 | * Count the number of extended attribute blocks | |
274 | */ | |
275 | if ( ((XFS_IFORK_Q(ip) != 0) && (ip->i_d.di_anextents > 0)) && | |
276 | (ip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { | |
277 | error = xfs_bmap_count_blocks(tp, ip, XFS_ATTR_FORK, &aforkblks); | |
278 | if (error) { | |
1da177e4 | 279 | xfs_trans_cancel(tp, 0); |
d0cfb373 | 280 | goto error0; |
1da177e4 LT |
281 | } |
282 | } | |
283 | if ( ((XFS_IFORK_Q(tip) != 0) && (tip->i_d.di_anextents > 0)) && | |
284 | (tip->i_d.di_aformat != XFS_DINODE_FMT_LOCAL)) { | |
285 | error = xfs_bmap_count_blocks(tp, tip, XFS_ATTR_FORK, | |
286 | &taforkblks); | |
287 | if (error) { | |
1da177e4 | 288 | xfs_trans_cancel(tp, 0); |
d0cfb373 | 289 | goto error0; |
1da177e4 LT |
290 | } |
291 | } | |
292 | ||
293 | /* | |
294 | * Swap the data forks of the inodes | |
295 | */ | |
296 | ifp = &ip->i_df; | |
297 | tifp = &tip->i_df; | |
d0cfb373 ES |
298 | *tempifp = *ifp; /* struct copy */ |
299 | *ifp = *tifp; /* struct copy */ | |
300 | *tifp = *tempifp; /* struct copy */ | |
1da177e4 LT |
301 | |
302 | /* | |
303 | * Fix the on-disk inode values | |
304 | */ | |
305 | tmp = (__uint64_t)ip->i_d.di_nblocks; | |
306 | ip->i_d.di_nblocks = tip->i_d.di_nblocks - taforkblks + aforkblks; | |
307 | tip->i_d.di_nblocks = tmp + taforkblks - aforkblks; | |
308 | ||
309 | tmp = (__uint64_t) ip->i_d.di_nextents; | |
310 | ip->i_d.di_nextents = tip->i_d.di_nextents; | |
311 | tip->i_d.di_nextents = tmp; | |
312 | ||
313 | tmp = (__uint64_t) ip->i_d.di_format; | |
314 | ip->i_d.di_format = tip->i_d.di_format; | |
315 | tip->i_d.di_format = tmp; | |
316 | ||
317 | ilf_fields = XFS_ILOG_CORE; | |
318 | ||
319 | switch(ip->i_d.di_format) { | |
320 | case XFS_DINODE_FMT_EXTENTS: | |
321 | /* If the extents fit in the inode, fix the | |
322 | * pointer. Otherwise it's already NULL or | |
323 | * pointing to the extent. | |
324 | */ | |
325 | if (ip->i_d.di_nextents <= XFS_INLINE_EXTS) { | |
326 | ifp->if_u1.if_extents = | |
327 | ifp->if_u2.if_inline_ext; | |
328 | } | |
329 | ilf_fields |= XFS_ILOG_DEXT; | |
330 | break; | |
331 | case XFS_DINODE_FMT_BTREE: | |
332 | ilf_fields |= XFS_ILOG_DBROOT; | |
333 | break; | |
334 | } | |
335 | ||
336 | tilf_fields = XFS_ILOG_CORE; | |
337 | ||
338 | switch(tip->i_d.di_format) { | |
339 | case XFS_DINODE_FMT_EXTENTS: | |
340 | /* If the extents fit in the inode, fix the | |
341 | * pointer. Otherwise it's already NULL or | |
342 | * pointing to the extent. | |
343 | */ | |
344 | if (tip->i_d.di_nextents <= XFS_INLINE_EXTS) { | |
345 | tifp->if_u1.if_extents = | |
346 | tifp->if_u2.if_inline_ext; | |
347 | } | |
348 | tilf_fields |= XFS_ILOG_DEXT; | |
349 | break; | |
350 | case XFS_DINODE_FMT_BTREE: | |
351 | tilf_fields |= XFS_ILOG_DBROOT; | |
352 | break; | |
353 | } | |
354 | ||
355 | /* | |
356 | * Increment vnode ref counts since xfs_trans_commit & | |
357 | * xfs_trans_cancel will both unlock the inodes and | |
358 | * decrement the associated ref counts. | |
359 | */ | |
360 | VN_HOLD(vp); | |
361 | VN_HOLD(tvp); | |
362 | ||
363 | xfs_trans_ijoin(tp, ip, lock_flags); | |
364 | xfs_trans_ijoin(tp, tip, lock_flags); | |
365 | ||
366 | xfs_trans_log_inode(tp, ip, ilf_fields); | |
367 | xfs_trans_log_inode(tp, tip, tilf_fields); | |
368 | ||
369 | /* | |
370 | * If this is a synchronous mount, make sure that the | |
371 | * transaction goes to disk before returning to the user. | |
372 | */ | |
373 | if (mp->m_flags & XFS_MOUNT_WSYNC) { | |
374 | xfs_trans_set_sync(tp); | |
375 | } | |
376 | ||
377 | error = xfs_trans_commit(tp, XFS_TRANS_SWAPEXT, NULL); | |
d0cfb373 | 378 | locked = 0; |
1da177e4 LT |
379 | |
380 | error0: | |
381 | if (locked) { | |
382 | xfs_iunlock(ip, lock_flags); | |
383 | xfs_iunlock(tip, lock_flags); | |
384 | } | |
385 | ||
d0cfb373 ES |
386 | if (fp != NULL) |
387 | fput(fp); | |
388 | if (tfp != NULL) | |
389 | fput(tfp); | |
390 | ||
391 | if (sxp != NULL) | |
392 | kmem_free(sxp, sizeof(xfs_swapext_t)); | |
393 | if (tempifp != NULL) | |
394 | kmem_free(tempifp, sizeof(xfs_ifork_t)); | |
1da177e4 LT |
395 | |
396 | return error; | |
397 | } |