ext2fs_inode.c revision 1.12 1 /* $NetBSD: ext2fs_inode.c,v 1.12 1999/03/05 21:09:49 mycroft Exp $ */
2
3 /*
4 * Copyright (c) 1997 Manuel Bouyer.
5 * Copyright (c) 1982, 1986, 1989, 1993
6 * The Regents of the University of California. All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. All advertising materials mentioning features or use of this software
17 * must display the following acknowledgement:
18 * This product includes software developed by the University of
19 * California, Berkeley and its contributors.
20 * 4. Neither the name of the University nor the names of its contributors
21 * may be used to endorse or promote products derived from this software
22 * without specific prior written permission.
23 *
24 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
25 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
28 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
29 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
30 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
31 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
32 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
33 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
34 * SUCH DAMAGE.
35 *
36 * @(#)ffs_inode.c 8.8 (Berkeley) 10/19/94
37 * Modified for ext2fs by Manuel Bouyer.
38 */
39
40 #if defined(_KERNEL) && !defined(_LKM)
41 #include "opt_uvm.h"
42 #endif
43
44 #include <sys/param.h>
45 #include <sys/systm.h>
46 #include <sys/mount.h>
47 #include <sys/proc.h>
48 #include <sys/file.h>
49 #include <sys/buf.h>
50 #include <sys/vnode.h>
51 #include <sys/kernel.h>
52 #include <sys/malloc.h>
53 #include <sys/trace.h>
54 #include <sys/resourcevar.h>
55
56 #include <vm/vm.h>
57 #if defined(UVM)
58 #include <uvm/uvm_extern.h>
59 #endif
60
61 #include <ufs/ufs/quota.h>
62 #include <ufs/ufs/inode.h>
63 #include <ufs/ufs/ufsmount.h>
64 #include <ufs/ufs/ufs_extern.h>
65
66 #include <ufs/ext2fs/ext2fs.h>
67 #include <ufs/ext2fs/ext2fs_extern.h>
68
69 static int ext2fs_indirtrunc __P((struct inode *, ufs_daddr_t, ufs_daddr_t,
70 ufs_daddr_t, int, long *));
71
72 /*
73 * Last reference to an inode. If necessary, write or delete it.
74 */
75 int
76 ext2fs_inactive(v)
77 void *v;
78 {
79 struct vop_inactive_args /* {
80 struct vnode *a_vp;
81 struct proc *a_p;
82 } */ *ap = v;
83 struct vnode *vp = ap->a_vp;
84 struct inode *ip = VTOI(vp);
85 struct proc *p = ap->a_p;
86 struct timespec ts;
87 int error = 0;
88 extern int prtactive;
89
90 if (prtactive && vp->v_usecount != 0)
91 vprint("ext2fs_inactive: pushing active", vp);
92 /* Get rid of inodes related to stale file handles. */
93 if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0)
94 goto out;
95
96 error = 0;
97 if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
98 error = VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, NULL);
99 TIMEVAL_TO_TIMESPEC(&time, &ts);
100 ip->i_e2fs_dtime = ts.tv_sec;
101 ip->i_flag |= IN_CHANGE | IN_UPDATE;
102 VOP_VFREE(vp, ip->i_number, ip->i_e2fs_mode);
103 }
104 if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
105 VOP_UPDATE(vp, NULL, NULL, 0);
106 out:
107 VOP_UNLOCK(vp, 0);
108 /*
109 * If we are done with the inode, reclaim it
110 * so that it can be reused immediately.
111 */
112 if (ip->i_e2fs_dtime != 0)
113 vrecycle(vp, (struct simplelock *)0, p);
114 return (error);
115 }
116
117
118 /*
119 * Update the access, modified, and inode change times as specified by the
120 * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
121 * used to specify that the inode needs to be updated but that the times have
122 * already been set. The access and modified times are taken from the second
123 * and third parameters; the inode change time is always taken from the current
124 * time. If waitfor is set, then wait for the disk write of the inode to
125 * complete.
126 */
127 int
128 ext2fs_update(v)
129 void *v;
130 {
131 struct vop_update_args /* {
132 struct vnode *a_vp;
133 struct timespec *a_access;
134 struct timespec *a_modify;
135 int a_waitfor;
136 } */ *ap = v;
137 register struct m_ext2fs *fs;
138 struct buf *bp;
139 struct inode *ip;
140 int error;
141 struct timespec ts;
142 caddr_t cp;
143
144 if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
145 return (0);
146 ip = VTOI(ap->a_vp);
147 TIMEVAL_TO_TIMESPEC(&time, &ts);
148 EXT2FS_ITIMES(ip,
149 ap->a_access ? ap->a_access : &ts,
150 ap->a_modify ? ap->a_modify : &ts, &ts);
151 if ((ip->i_flag & IN_MODIFIED) == 0)
152 return (0);
153 ip->i_flag &= ~IN_MODIFIED;
154 fs = ip->i_e2fs;
155 error = bread(ip->i_devvp,
156 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
157 (int)fs->e2fs_bsize, NOCRED, &bp);
158 if (error) {
159 brelse(bp);
160 return (error);
161 }
162 cp = (caddr_t)bp->b_data +
163 (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
164 e2fs_isave(&ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp);
165 if (ap->a_waitfor && (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0)
166 return (bwrite(bp));
167 else {
168 bdwrite(bp);
169 return (0);
170 }
171 }
172
173 #define SINGLE 0 /* index of single indirect block */
174 #define DOUBLE 1 /* index of double indirect block */
175 #define TRIPLE 2 /* index of triple indirect block */
176 /*
177 * Truncate the inode oip to at most length size, freeing the
178 * disk blocks.
179 */
180 int
181 ext2fs_truncate(v)
182 void *v;
183 {
184 struct vop_truncate_args /* {
185 struct vnode *a_vp;
186 off_t a_length;
187 int a_flags;
188 struct ucred *a_cred;
189 struct proc *a_p;
190 } */ *ap = v;
191 register struct vnode *ovp = ap->a_vp;
192 register ufs_daddr_t lastblock;
193 register struct inode *oip;
194 ufs_daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
195 ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
196 off_t length = ap->a_length;
197 register struct m_ext2fs *fs;
198 struct buf *bp;
199 int offset, size, level;
200 long count, nblocks, vflags, blocksreleased = 0;
201 register int i;
202 int aflags, error, allerror;
203 off_t osize;
204
205 if (length < 0)
206 return (EINVAL);
207
208 oip = VTOI(ovp);
209 if (ovp->v_type == VLNK &&
210 (oip->i_e2fs_size < ovp->v_mount->mnt_maxsymlinklen ||
211 (ovp->v_mount->mnt_maxsymlinklen == 0 &&
212 oip->i_e2fs_nblock == 0))) {
213 #ifdef DIAGNOSTIC
214 if (length != 0)
215 panic("ext2fs_truncate: partial truncate of symlink");
216 #endif
217 memset((char *)&oip->i_din.e2fs_din.e2di_shortlink, 0,
218 (u_int)oip->i_e2fs_size);
219 oip->i_e2fs_size = 0;
220 oip->i_flag |= IN_CHANGE | IN_UPDATE;
221 return (VOP_UPDATE(ovp, NULL, NULL, 1));
222 }
223 if (oip->i_e2fs_size == length) {
224 oip->i_flag |= IN_CHANGE | IN_UPDATE;
225 return (VOP_UPDATE(ovp, NULL, NULL, 0));
226 }
227 fs = oip->i_e2fs;
228 osize = oip->i_e2fs_size;
229 /*
230 * Lengthen the size of the file. We must ensure that the
231 * last byte of the file is allocated. Since the smallest
232 * value of osize is 0, length will be at least 1.
233 */
234 if (osize < length) {
235 #if 0 /* XXX */
236 if (length > fs->fs_maxfilesize)
237 return (EFBIG);
238 #endif
239 offset = blkoff(fs, length - 1);
240 lbn = lblkno(fs, length - 1);
241 aflags = B_CLRBUF;
242 if (ap->a_flags & IO_SYNC)
243 aflags |= B_SYNC;
244 error = ext2fs_balloc(oip, lbn, offset + 1, ap->a_cred, &bp,
245 aflags);
246 if (error)
247 return (error);
248 oip->i_e2fs_size = length;
249 #if defined(UVM)
250 uvm_vnp_setsize(ovp, length);
251 (void) uvm_vnp_uncache(ovp);
252 #else
253 vnode_pager_setsize(ovp, length);
254 (void) vnode_pager_uncache(ovp);
255 #endif
256 if (aflags & B_SYNC)
257 bwrite(bp);
258 else
259 bawrite(bp);
260 oip->i_flag |= IN_CHANGE | IN_UPDATE;
261 return (VOP_UPDATE(ovp, NULL, NULL, 1));
262 }
263 /*
264 * Shorten the size of the file. If the file is not being
265 * truncated to a block boundry, the contents of the
266 * partial block following the end of the file must be
267 * zero'ed in case it ever become accessable again because
268 * of subsequent file growth.
269 */
270 offset = blkoff(fs, length);
271 if (offset == 0) {
272 oip->i_e2fs_size = length;
273 } else {
274 lbn = lblkno(fs, length);
275 aflags = B_CLRBUF;
276 if (ap->a_flags & IO_SYNC)
277 aflags |= B_SYNC;
278 error = ext2fs_balloc(oip, lbn, offset, ap->a_cred, &bp, aflags);
279 if (error)
280 return (error);
281 oip->i_e2fs_size = length;
282 size = fs->e2fs_bsize;
283 #if defined(UVM)
284 (void) uvm_vnp_uncache(ovp);
285 #else
286 (void) vnode_pager_uncache(ovp);
287 #endif
288 memset((char *)bp->b_data + offset, 0, (u_int)(size - offset));
289 allocbuf(bp, size);
290 if (aflags & B_SYNC)
291 bwrite(bp);
292 else
293 bawrite(bp);
294 }
295 #if defined(UVM)
296 uvm_vnp_setsize(ovp, length);
297 #else
298 vnode_pager_setsize(ovp, length);
299 #endif
300 /*
301 * Calculate index into inode's block list of
302 * last direct and indirect blocks (if any)
303 * which we want to keep. Lastblock is -1 when
304 * the file is truncated to 0.
305 */
306 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
307 lastiblock[SINGLE] = lastblock - NDADDR;
308 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
309 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
310 nblocks = btodb(fs->e2fs_bsize);
311 /*
312 * Update file and block pointers on disk before we start freeing
313 * blocks. If we crash before free'ing blocks below, the blocks
314 * will be returned to the free list. lastiblock values are also
315 * normalized to -1 for calls to ext2fs_indirtrunc below.
316 */
317 memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
318 for (level = TRIPLE; level >= SINGLE; level--)
319 if (lastiblock[level] < 0) {
320 oip->i_e2fs_blocks[NDADDR + level] = 0;
321 lastiblock[level] = -1;
322 }
323 for (i = NDADDR - 1; i > lastblock; i--)
324 oip->i_e2fs_blocks[i] = 0;
325 oip->i_flag |= IN_CHANGE | IN_UPDATE;
326 if ((error = VOP_UPDATE(ovp, NULL, NULL, 1)) != 0)
327 allerror = error;
328 /*
329 * Having written the new inode to disk, save its new configuration
330 * and put back the old block pointers long enough to process them.
331 * Note that we save the new block configuration so we can check it
332 * when we are done.
333 */
334 memcpy((caddr_t)newblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof newblks);
335 memcpy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)oldblks, sizeof oldblks);
336 oip->i_e2fs_size = osize;
337 vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
338 allerror = vinvalbuf(ovp, vflags, ap->a_cred, ap->a_p, 0, 0);
339
340 /*
341 * Indirect blocks first.
342 */
343 indir_lbn[SINGLE] = -NDADDR;
344 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1;
345 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
346 for (level = TRIPLE; level >= SINGLE; level--) {
347 bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]);
348 if (bn != 0) {
349 error = ext2fs_indirtrunc(oip, indir_lbn[level],
350 fsbtodb(fs, bn), lastiblock[level], level, &count);
351 if (error)
352 allerror = error;
353 blocksreleased += count;
354 if (lastiblock[level] < 0) {
355 oip->i_e2fs_blocks[NDADDR + level] = 0;
356 ext2fs_blkfree(oip, bn);
357 blocksreleased += nblocks;
358 }
359 }
360 if (lastiblock[level] >= 0)
361 goto done;
362 }
363
364 /*
365 * All whole direct blocks or frags.
366 */
367 for (i = NDADDR - 1; i > lastblock; i--) {
368 bn = fs2h32(oip->i_e2fs_blocks[i]);
369 if (bn == 0)
370 continue;
371 oip->i_e2fs_blocks[i] = 0;
372 ext2fs_blkfree(oip, bn);
373 blocksreleased += btodb(fs->e2fs_bsize);
374 }
375 if (lastblock < 0)
376 goto done;
377
378 done:
379 #ifdef DIAGNOSTIC
380 for (level = SINGLE; level <= TRIPLE; level++)
381 if (newblks[NDADDR + level] !=
382 fs2h32(oip->i_e2fs_blocks[NDADDR + level]))
383 panic("itrunc1");
384 for (i = 0; i < NDADDR; i++)
385 if (newblks[i] != fs2h32(oip->i_e2fs_blocks[i]))
386 panic("itrunc2");
387 if (length == 0 &&
388 (ovp->v_dirtyblkhd.lh_first || ovp->v_cleanblkhd.lh_first))
389 panic("itrunc3");
390 #endif /* DIAGNOSTIC */
391 /*
392 * Put back the real size.
393 */
394 oip->i_e2fs_size = length;
395 oip->i_e2fs_nblock -= blocksreleased;
396 if (oip->i_e2fs_nblock < 0) /* sanity */
397 oip->i_e2fs_nblock = 0;
398 oip->i_flag |= IN_CHANGE;
399 return (allerror);
400 }
401
402 /*
403 * Release blocks associated with the inode ip and stored in the indirect
404 * block bn. Blocks are free'd in LIFO order up to (but not including)
405 * lastbn. If level is greater than SINGLE, the block is an indirect block
406 * and recursive calls to indirtrunc must be used to cleanse other indirect
407 * blocks.
408 *
409 * NB: triple indirect blocks are untested.
410 */
411 static int
412 ext2fs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
413 register struct inode *ip;
414 ufs_daddr_t lbn, lastbn;
415 ufs_daddr_t dbn;
416 int level;
417 long *countp;
418 {
419 register int i;
420 struct buf *bp;
421 register struct m_ext2fs *fs = ip->i_e2fs;
422 register ufs_daddr_t *bap;
423 struct vnode *vp;
424 ufs_daddr_t *copy = NULL, nb, nlbn, last;
425 long blkcount, factor;
426 int nblocks, blocksreleased = 0;
427 int error = 0, allerror = 0;
428
429 /*
430 * Calculate index in current block of last
431 * block to be kept. -1 indicates the entire
432 * block so we need not calculate the index.
433 */
434 factor = 1;
435 for (i = SINGLE; i < level; i++)
436 factor *= NINDIR(fs);
437 last = lastbn;
438 if (lastbn > 0)
439 last /= factor;
440 nblocks = btodb(fs->e2fs_bsize);
441 /*
442 * Get buffer of block pointers, zero those entries corresponding
443 * to blocks to be free'd, and update on disk copy first. Since
444 * double(triple) indirect before single(double) indirect, calls
445 * to bmap on these blocks will fail. However, we already have
446 * the on disk address, so we have to set the b_blkno field
447 * explicitly instead of letting bread do everything for us.
448 */
449 vp = ITOV(ip);
450 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0);
451 if (bp->b_flags & (B_DONE | B_DELWRI)) {
452 /* Braces must be here in case trace evaluates to nothing. */
453 trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn);
454 } else {
455 trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn);
456 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */
457 bp->b_flags |= B_READ;
458 if (bp->b_bcount > bp->b_bufsize)
459 panic("ext2fs_indirtrunc: bad buffer size");
460 bp->b_blkno = dbn;
461 VOP_STRATEGY(bp);
462 error = biowait(bp);
463 }
464 if (error) {
465 brelse(bp);
466 *countp = 0;
467 return (error);
468 }
469
470 bap = (ufs_daddr_t *)bp->b_data;
471 if (lastbn != -1) {
472 MALLOC(copy, ufs_daddr_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK);
473 memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize);
474 memset((caddr_t)&bap[last + 1], 0,
475 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t));
476 error = bwrite(bp);
477 if (error)
478 allerror = error;
479 bap = copy;
480 }
481
482 /*
483 * Recursively free totally unused blocks.
484 */
485 for (i = NINDIR(fs) - 1,
486 nlbn = lbn + 1 - i * factor; i > last;
487 i--, nlbn += factor) {
488 nb = fs2h32(bap[i]);
489 if (nb == 0)
490 continue;
491 if (level > SINGLE) {
492 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
493 (ufs_daddr_t)-1, level - 1,
494 &blkcount);
495 if (error)
496 allerror = error;
497 blocksreleased += blkcount;
498 }
499 ext2fs_blkfree(ip, nb);
500 blocksreleased += nblocks;
501 }
502
503 /*
504 * Recursively free last partial block.
505 */
506 if (level > SINGLE && lastbn >= 0) {
507 last = lastbn % factor;
508 nb = fs2h32(bap[i]);
509 if (nb != 0) {
510 error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
511 last, level - 1, &blkcount);
512 if (error)
513 allerror = error;
514 blocksreleased += blkcount;
515 }
516 }
517
518 if (copy != NULL) {
519 FREE(copy, M_TEMP);
520 } else {
521 bp->b_flags |= B_INVAL;
522 brelse(bp);
523 }
524
525 *countp = blocksreleased;
526 return (allerror);
527 }
528