ext2fs_inode.c revision 1.37 1 /* $NetBSD: ext2fs_inode.c,v 1.37 2003/10/15 11:29:01 hannken Exp $ */
2
3 /*
4 * Copyright (c) 1982, 1986, 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 *
31 * @(#)ffs_inode.c 8.8 (Berkeley) 10/19/94
32 * Modified for ext2fs by Manuel Bouyer.
33 */
34
35 /*
36 * Copyright (c) 1997 Manuel Bouyer.
37 *
38 * Redistribution and use in source and binary forms, with or without
39 * modification, are permitted provided that the following conditions
40 * are met:
41 * 1. Redistributions of source code must retain the above copyright
42 * notice, this list of conditions and the following disclaimer.
43 * 2. Redistributions in binary form must reproduce the above copyright
44 * notice, this list of conditions and the following disclaimer in the
45 * documentation and/or other materials provided with the distribution.
46 * 3. All advertising materials mentioning features or use of this software
47 * must display the following acknowledgement:
48 * This product includes software developed by Manuel Bouyer.
49 * 4. The name of the author may not be used to endorse or promote products
50 * derived from this software without specific prior written permission.
51 *
52 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
53 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
54 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
55 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
56 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
57 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
58 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
59 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
60 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
61 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
62 * SUCH DAMAGE.
63 *
64 * @(#)ffs_inode.c 8.8 (Berkeley) 10/19/94
65 * Modified for ext2fs by Manuel Bouyer.
66 */
67
68 #include <sys/cdefs.h>
69 __KERNEL_RCSID(0, "$NetBSD: ext2fs_inode.c,v 1.37 2003/10/15 11:29:01 hannken Exp $");
70
71 #include <sys/param.h>
72 #include <sys/systm.h>
73 #include <sys/mount.h>
74 #include <sys/proc.h>
75 #include <sys/file.h>
76 #include <sys/buf.h>
77 #include <sys/vnode.h>
78 #include <sys/kernel.h>
79 #include <sys/malloc.h>
80 #include <sys/trace.h>
81 #include <sys/resourcevar.h>
82
83 #include <ufs/ufs/inode.h>
84 #include <ufs/ufs/ufsmount.h>
85 #include <ufs/ufs/ufs_extern.h>
86
87 #include <ufs/ext2fs/ext2fs.h>
88 #include <ufs/ext2fs/ext2fs_extern.h>
89
90 extern int prtactive;
91
92 static int ext2fs_indirtrunc __P((struct inode *, daddr_t, daddr_t,
93 daddr_t, int, long *));
94
95 /*
96 * Last reference to an inode. If necessary, write or delete it.
97 */
98 int
99 ext2fs_inactive(v)
100 void *v;
101 {
102 struct vop_inactive_args /* {
103 struct vnode *a_vp;
104 struct proc *a_p;
105 } */ *ap = v;
106 struct vnode *vp = ap->a_vp;
107 struct inode *ip = VTOI(vp);
108 struct mount *mp;
109 struct proc *p = ap->a_p;
110 struct timespec ts;
111 int error = 0;
112
113 if (prtactive && vp->v_usecount != 0)
114 vprint("ext2fs_inactive: pushing active", vp);
115 /* Get rid of inodes related to stale file handles. */
116 if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0)
117 goto out;
118
119 error = 0;
120 if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
121 error = vn_start_write(vp, &mp, V_WAIT | V_LOWER);
122 if (error)
123 return (error);
124 if (ip->i_e2fs_size != 0) {
125 error = VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, NULL);
126 }
127 TIMEVAL_TO_TIMESPEC(&time, &ts);
128 ip->i_e2fs_dtime = ts.tv_sec;
129 ip->i_flag |= IN_CHANGE | IN_UPDATE;
130 VOP_VFREE(vp, ip->i_number, ip->i_e2fs_mode);
131 vn_finished_write(mp, V_LOWER);
132 }
133 if (ip->i_flag &
134 (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFIED | IN_ACCESSED)) {
135 error = vn_start_write(vp, &mp, V_WAIT | V_LOWER);
136 if (error)
137 return (error);
138 VOP_UPDATE(vp, NULL, NULL, 0);
139 vn_finished_write(mp, V_LOWER);
140 }
141 out:
142 VOP_UNLOCK(vp, 0);
143 /*
144 * If we are done with the inode, reclaim it
145 * so that it can be reused immediately.
146 */
147 if (ip->i_e2fs_dtime != 0)
148 vrecycle(vp, NULL, p);
149 return (error);
150 }
151
152
153 /*
154 * Update the access, modified, and inode change times as specified by the
155 * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
156 * used to specify that the inode needs to be updated but that the times have
157 * already been set. The access and modified times are taken from the second
158 * and third parameters; the inode change time is always taken from the current
159 * time. If UPDATE_WAIT or UPDATE_DIROP is set, then wait for the disk
160 * write of the inode to complete.
161 */
162 int
163 ext2fs_update(v)
164 void *v;
165 {
166 struct vop_update_args /* {
167 struct vnode *a_vp;
168 struct timespec *a_access;
169 struct timespec *a_modify;
170 int a_flags;
171 } */ *ap = v;
172 struct m_ext2fs *fs;
173 struct buf *bp;
174 struct inode *ip;
175 int error;
176 struct timespec ts;
177 caddr_t cp;
178 int flags;
179
180 if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
181 return (0);
182 ip = VTOI(ap->a_vp);
183 TIMEVAL_TO_TIMESPEC(&time, &ts);
184 EXT2FS_ITIMES(ip,
185 ap->a_access ? ap->a_access : &ts,
186 ap->a_modify ? ap->a_modify : &ts, &ts);
187 flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED);
188 if (flags == 0)
189 return (0);
190 fs = ip->i_e2fs;
191
192 error = bread(ip->i_devvp,
193 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
194 (int)fs->e2fs_bsize, NOCRED, &bp);
195 if (error) {
196 brelse(bp);
197 return (error);
198 }
199 ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED);
200 cp = (caddr_t)bp->b_data +
201 (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
202 e2fs_isave(ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp);
203 if ((ap->a_flags & (UPDATE_WAIT|UPDATE_DIROP)) != 0 &&
204 (flags & IN_MODIFIED) != 0 &&
205 (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0)
206 return (bwrite(bp));
207 else {
208 bdwrite(bp);
209 return (0);
210 }
211 }
212
213 #define SINGLE 0 /* index of single indirect block */
214 #define DOUBLE 1 /* index of double indirect block */
215 #define TRIPLE 2 /* index of triple indirect block */
216 /*
217 * Truncate the inode oip to at most length size, freeing the
218 * disk blocks.
219 */
220 int
221 ext2fs_truncate(v)
222 void *v;
223 {
224 struct vop_truncate_args /* {
225 struct vnode *a_vp;
226 off_t a_length;
227 int a_flags;
228 struct ucred *a_cred;
229 struct proc *a_p;
230 } */ *ap = v;
231 struct vnode *ovp = ap->a_vp;
232 daddr_t lastblock;
233 struct inode *oip;
234 daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR];
235 /* XXX ondisk32 */
236 int32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
237 off_t length = ap->a_length;
238 struct m_ext2fs *fs;
239 int offset, size, level;
240 long count, nblocks, blocksreleased = 0;
241 int i;
242 int error, allerror = 0;
243 off_t osize;
244
245 if (length < 0)
246 return (EINVAL);
247
248 oip = VTOI(ovp);
249 if (ovp->v_type == VLNK &&
250 (oip->i_e2fs_size < ovp->v_mount->mnt_maxsymlinklen ||
251 (ovp->v_mount->mnt_maxsymlinklen == 0 &&
252 oip->i_e2fs_nblock == 0))) {
253 #ifdef DIAGNOSTIC
254 if (length != 0)
255 panic("ext2fs_truncate: partial truncate of symlink");
256 #endif
257 memset((char *)&oip->i_din.e2fs_din->e2di_shortlink, 0,
258 (u_int)oip->i_e2fs_size);
259 oip->i_e2fs_size = 0;
260 oip->i_flag |= IN_CHANGE | IN_UPDATE;
261 return (VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT));
262 }
263 if (oip->i_e2fs_size == length) {
264 oip->i_flag |= IN_CHANGE | IN_UPDATE;
265 return (VOP_UPDATE(ovp, NULL, NULL, 0));
266 }
267 fs = oip->i_e2fs;
268 osize = oip->i_e2fs_size;
269 /*
270 * Lengthen the size of the file. We must ensure that the
271 * last byte of the file is allocated. Since the smallest
272 * value of osize is 0, length will be at least 1.
273 */
274 if (osize < length) {
275 #if 0 /* XXX */
276 if (length > fs->fs_maxfilesize)
277 return (EFBIG);
278 #endif
279 ufs_balloc_range(ovp, length - 1, 1, ap->a_cred,
280 ap->a_flags & IO_SYNC ? B_SYNC : 0);
281 oip->i_flag |= IN_CHANGE | IN_UPDATE;
282 return (VOP_UPDATE(ovp, NULL, NULL, 1));
283 }
284 /*
285 * Shorten the size of the file. If the file is not being
286 * truncated to a block boundry, the contents of the
287 * partial block following the end of the file must be
288 * zero'ed in case it ever become accessible again because
289 * of subsequent file growth.
290 */
291 offset = blkoff(fs, length);
292 if (offset != 0) {
293 size = fs->e2fs_bsize;
294
295 /* XXXUBC we should handle more than just VREG */
296 uvm_vnp_zerorange(ovp, length, size - offset);
297 }
298 oip->i_e2fs_size = length;
299 uvm_vnp_setsize(ovp, length);
300
301 /*
302 * Calculate index into inode's block list of
303 * last direct and indirect blocks (if any)
304 * which we want to keep. Lastblock is -1 when
305 * the file is truncated to 0.
306 */
307 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
308 lastiblock[SINGLE] = lastblock - NDADDR;
309 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
310 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
311 nblocks = btodb(fs->e2fs_bsize);
312 /*
313 * Update file and block pointers on disk before we start freeing
314 * blocks. If we crash before free'ing blocks below, the blocks
315 * will be returned to the free list. lastiblock values are also
316 * normalized to -1 for calls to ext2fs_indirtrunc below.
317 */
318 memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
319 for (level = TRIPLE; level >= SINGLE; level--)
320 if (lastiblock[level] < 0) {
321 oip->i_e2fs_blocks[NDADDR + level] = 0;
322 lastiblock[level] = -1;
323 }
324 for (i = NDADDR - 1; i > lastblock; i--)
325 oip->i_e2fs_blocks[i] = 0;
326 oip->i_flag |= IN_CHANGE | IN_UPDATE;
327 error = VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT);
328 if (error && !allerror)
329 allerror = error;
330
331 /*
332 * Having written the new inode to disk, save its new configuration
333 * and put back the old block pointers long enough to process them.
334 * Note that we save the new block configuration so we can check it
335 * when we are done.
336 */
337
338 memcpy((caddr_t)newblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof newblks);
339 memcpy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)oldblks, sizeof oldblks);
340 oip->i_e2fs_size = osize;
341 error = vtruncbuf(ovp, lastblock + 1, 0, 0);
342 if (error && !allerror)
343 allerror = error;
344
345 /*
346 * Indirect blocks first.
347 */
348 indir_lbn[SINGLE] = -NDADDR;
349 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1;
350 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
351 for (level = TRIPLE; level >= SINGLE; level--) {
352 /* XXX ondisk32 */
353 bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]);
354 if (bn != 0) {
355 error = ext2fs_indirtrunc(oip, indir_lbn[level],
356 fsbtodb(fs, bn), lastiblock[level], level, &count);
357 if (error)
358 allerror = error;
359 blocksreleased += count;
360 if (lastiblock[level] < 0) {
361 oip->i_e2fs_blocks[NDADDR + level] = 0;
362 ext2fs_blkfree(oip, bn);
363 blocksreleased += nblocks;
364 }
365 }
366 if (lastiblock[level] >= 0)
367 goto done;
368 }
369
370 /*
371 * All whole direct blocks or frags.
372 */
373 for (i = NDADDR - 1; i > lastblock; i--) {
374 /* XXX ondisk32 */
375 bn = fs2h32(oip->i_e2fs_blocks[i]);
376 if (bn == 0)
377 continue;
378 oip->i_e2fs_blocks[i] = 0;
379 ext2fs_blkfree(oip, bn);
380 blocksreleased += btodb(fs->e2fs_bsize);
381 }
382
383 done:
384 #ifdef DIAGNOSTIC
385 for (level = SINGLE; level <= TRIPLE; level++)
386 if (newblks[NDADDR + level] !=
387 oip->i_e2fs_blocks[NDADDR + level])
388 panic("ext2fs_truncate1");
389 for (i = 0; i < NDADDR; i++)
390 if (newblks[i] != oip->i_e2fs_blocks[i])
391 panic("ext2fs_truncate2");
392 if (length == 0 &&
393 (!LIST_EMPTY(&ovp->v_cleanblkhd) ||
394 !LIST_EMPTY(&ovp->v_dirtyblkhd)))
395 panic("ext2fs_truncate3");
396 #endif /* DIAGNOSTIC */
397 /*
398 * Put back the real size.
399 */
400 oip->i_e2fs_size = length;
401 oip->i_e2fs_nblock -= blocksreleased;
402 oip->i_flag |= IN_CHANGE;
403 return (allerror);
404 }
405
406 /*
407 * Release blocks associated with the inode ip and stored in the indirect
408 * block bn. Blocks are free'd in LIFO order up to (but not including)
409 * lastbn. If level is greater than SINGLE, the block is an indirect block
410 * and recursive calls to indirtrunc must be used to cleanse other indirect
411 * blocks.
412 *
413 * NB: triple indirect blocks are untested.
414 */
415 static int
416 ext2fs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
417 struct inode *ip;
418 daddr_t lbn, lastbn;
419 daddr_t dbn;
420 int level;
421 long *countp;
422 {
423 int i;
424 struct buf *bp;
425 struct m_ext2fs *fs = ip->i_e2fs;
426 int32_t *bap; /* XXX ondisk32 */
427 struct vnode *vp;
428 daddr_t nb, nlbn, last;
429 int32_t *copy = NULL; /* XXX ondisk32 */
430 long blkcount, factor;
431 int nblocks, blocksreleased = 0;
432 int error = 0, allerror = 0;
433
434 /*
435 * Calculate index in current block of last
436 * block to be kept. -1 indicates the entire
437 * block so we need not calculate the index.
438 */
439 factor = 1;
440 for (i = SINGLE; i < level; i++)
441 factor *= NINDIR(fs);
442 last = lastbn;
443 if (lastbn > 0)
444 last /= factor;
445 nblocks = btodb(fs->e2fs_bsize);
446 /*
447 * Get buffer of block pointers, zero those entries corresponding
448 * to blocks to be free'd, and update on disk copy first. Since
449 * double(triple) indirect before single(double) indirect, calls
450 * to bmap on these blocks will fail. However, we already have
451 * the on disk address, so we have to set the b_blkno field
452 * explicitly instead of letting bread do everything for us.
453 */
454 vp = ITOV(ip);
455 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0);
456 if (bp->b_flags & (B_DONE | B_DELWRI)) {
457 /* Braces must be here in case trace evaluates to nothing. */
458 trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn);
459 } else {
460 trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn);
461 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */
462 bp->b_flags |= B_READ;
463 if (bp->b_bcount > bp->b_bufsize)
464 panic("ext2fs_indirtrunc: bad buffer size");
465 bp->b_blkno = dbn;
466 VOP_STRATEGY(bp);
467 error = biowait(bp);
468 }
469 if (error) {
470 brelse(bp);
471 *countp = 0;
472 return (error);
473 }
474
475 bap = (int32_t *)bp->b_data; /* XXX ondisk32 */
476 if (lastbn >= 0) {
477 /* XXX ondisk32 */
478 MALLOC(copy, int32_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK);
479 memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize);
480 memset((caddr_t)&bap[last + 1], 0,
481 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t));
482 error = bwrite(bp);
483 if (error)
484 allerror = error;
485 bap = copy;
486 }
487
488 /*
489 * Recursively free totally unused blocks.
490 */
491 for (i = NINDIR(fs) - 1,
492 nlbn = lbn + 1 - i * factor; i > last;
493 i--, nlbn += factor) {
494 /* XXX ondisk32 */
495 nb = fs2h32(bap[i]);
496 if (nb == 0)
497 continue;
498 if (level > SINGLE) {
499 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
500 (daddr_t)-1, level - 1,
501 &blkcount);
502 if (error)
503 allerror = error;
504 blocksreleased += blkcount;
505 }
506 ext2fs_blkfree(ip, nb);
507 blocksreleased += nblocks;
508 }
509
510 /*
511 * Recursively free last partial block.
512 */
513 if (level > SINGLE && lastbn >= 0) {
514 last = lastbn % factor;
515 /* XXX ondisk32 */
516 nb = fs2h32(bap[i]);
517 if (nb != 0) {
518 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
519 last, level - 1, &blkcount);
520 if (error)
521 allerror = error;
522 blocksreleased += blkcount;
523 }
524 }
525
526 if (copy != NULL) {
527 FREE(copy, M_TEMP);
528 } else {
529 bp->b_flags |= B_INVAL;
530 brelse(bp);
531 }
532
533 *countp = blocksreleased;
534 return (allerror);
535 }
536