xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_inode.c (revision 4b896b232495b7a9b8b94a1cf1e21873296d53b8)
1 /*	$NetBSD: ext2fs_inode.c,v 1.40 2004/03/22 19:23:08 bouyer 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 AUTHOR ``AS IS'' AND ANY EXPRESS OR
53  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
54  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
55  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
56  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
57  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
58  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
59  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
60  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
61  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
62  *
63  *	@(#)ffs_inode.c	8.8 (Berkeley) 10/19/94
64  * Modified for ext2fs by Manuel Bouyer.
65  */
66 
67 #include <sys/cdefs.h>
68 __KERNEL_RCSID(0, "$NetBSD: ext2fs_inode.c,v 1.40 2004/03/22 19:23:08 bouyer Exp $");
69 
70 #include <sys/param.h>
71 #include <sys/systm.h>
72 #include <sys/mount.h>
73 #include <sys/proc.h>
74 #include <sys/file.h>
75 #include <sys/buf.h>
76 #include <sys/vnode.h>
77 #include <sys/kernel.h>
78 #include <sys/malloc.h>
79 #include <sys/trace.h>
80 #include <sys/resourcevar.h>
81 
82 #include <ufs/ufs/inode.h>
83 #include <ufs/ufs/ufsmount.h>
84 #include <ufs/ufs/ufs_extern.h>
85 
86 #include <ufs/ext2fs/ext2fs.h>
87 #include <ufs/ext2fs/ext2fs_extern.h>
88 
89 extern int prtactive;
90 
91 static int ext2fs_indirtrunc __P((struct inode *, daddr_t, daddr_t,
92 				  daddr_t, int, long *));
93 
94 /*
95  * Last reference to an inode.  If necessary, write or delete it.
96  */
97 int
98 ext2fs_inactive(v)
99 	void *v;
100 {
101 	struct vop_inactive_args /* {
102 		struct vnode *a_vp;
103 		struct proc *a_p;
104 	} */ *ap = v;
105 	struct vnode *vp = ap->a_vp;
106 	struct inode *ip = VTOI(vp);
107 	struct mount *mp;
108 	struct proc *p = ap->a_p;
109 	struct timespec ts;
110 	int error = 0;
111 
112 	if (prtactive && vp->v_usecount != 0)
113 		vprint("ext2fs_inactive: pushing active", vp);
114 	/* Get rid of inodes related to stale file handles. */
115 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0)
116 		goto out;
117 
118 	error = 0;
119 	if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
120 		vn_start_write(vp, &mp, V_WAIT | V_LOWER);
121 		if (ip->i_e2fs_size != 0) {
122 			error = VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, NULL);
123 		}
124 		TIMEVAL_TO_TIMESPEC(&time, &ts);
125 		ip->i_e2fs_dtime = ts.tv_sec;
126 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
127 		VOP_VFREE(vp, ip->i_number, ip->i_e2fs_mode);
128 		vn_finished_write(mp, V_LOWER);
129 	}
130 	if (ip->i_flag &
131 	    (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFIED | IN_ACCESSED)) {
132 		vn_start_write(vp, &mp, V_WAIT | V_LOWER);
133 		VOP_UPDATE(vp, NULL, NULL, 0);
134 		vn_finished_write(mp, V_LOWER);
135 	}
136 out:
137 	VOP_UNLOCK(vp, 0);
138 	/*
139 	 * If we are done with the inode, reclaim it
140 	 * so that it can be reused immediately.
141 	 */
142 	if (ip->i_e2fs_dtime != 0)
143 		vrecycle(vp, NULL, p);
144 	return (error);
145 }
146 
147 
148 /*
149  * Update the access, modified, and inode change times as specified by the
150  * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
151  * used to specify that the inode needs to be updated but that the times have
152  * already been set. The access and modified times are taken from the second
153  * and third parameters; the inode change time is always taken from the current
154  * time. If UPDATE_WAIT or UPDATE_DIROP is set, then wait for the disk
155  * write of the inode to complete.
156  */
157 int
158 ext2fs_update(v)
159 	void *v;
160 {
161 	struct vop_update_args /* {
162 		struct vnode *a_vp;
163 		struct timespec *a_access;
164 		struct timespec *a_modify;
165 		int a_flags;
166 	} */ *ap = v;
167 	struct m_ext2fs *fs;
168 	struct buf *bp;
169 	struct inode *ip;
170 	int error;
171 	struct timespec ts;
172 	caddr_t cp;
173 	int flags;
174 
175 	if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
176 		return (0);
177 	ip = VTOI(ap->a_vp);
178 	TIMEVAL_TO_TIMESPEC(&time, &ts);
179 	EXT2FS_ITIMES(ip,
180 	    ap->a_access ? ap->a_access : &ts,
181 	    ap->a_modify ? ap->a_modify : &ts, &ts);
182 	flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED);
183 	if (flags == 0)
184 		return (0);
185 	fs = ip->i_e2fs;
186 
187 	error = bread(ip->i_devvp,
188 			  fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
189 			  (int)fs->e2fs_bsize, NOCRED, &bp);
190 	if (error) {
191 		brelse(bp);
192 		return (error);
193 	}
194 	ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED);
195 	cp = (caddr_t)bp->b_data +
196 	    (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
197 	e2fs_isave(ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp);
198 	if ((ap->a_flags & (UPDATE_WAIT|UPDATE_DIROP)) != 0 &&
199 	    (flags & IN_MODIFIED) != 0 &&
200 	    (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0)
201 		return (bwrite(bp));
202 	else {
203 		bdwrite(bp);
204 		return (0);
205 	}
206 }
207 
208 #define	SINGLE	0	/* index of single indirect block */
209 #define	DOUBLE	1	/* index of double indirect block */
210 #define	TRIPLE	2	/* index of triple indirect block */
211 /*
212  * Truncate the inode oip to at most length size, freeing the
213  * disk blocks.
214  */
215 int
216 ext2fs_truncate(v)
217 	void *v;
218 {
219 	struct vop_truncate_args /* {
220 		struct vnode *a_vp;
221 		off_t a_length;
222 		int a_flags;
223 		struct ucred *a_cred;
224 		struct proc *a_p;
225 	} */ *ap = v;
226 	struct vnode *ovp = ap->a_vp;
227 	daddr_t lastblock;
228 	struct inode *oip;
229 	daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR];
230 	/* XXX ondisk32 */
231 	int32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
232 	off_t length = ap->a_length;
233 	struct m_ext2fs *fs;
234 	int offset, size, level;
235 	long count, nblocks, blocksreleased = 0;
236 	int i;
237 	int error, allerror = 0;
238 	off_t osize;
239 
240 	if (length < 0)
241 		return (EINVAL);
242 
243 	oip = VTOI(ovp);
244 	if (ovp->v_type == VLNK &&
245 		(oip->i_e2fs_size < ovp->v_mount->mnt_maxsymlinklen ||
246 		 (ovp->v_mount->mnt_maxsymlinklen == 0 &&
247 		  oip->i_e2fs_nblock == 0))) {
248 #ifdef DIAGNOSTIC
249 		if (length != 0)
250 			panic("ext2fs_truncate: partial truncate of symlink");
251 #endif
252 		memset((char *)&oip->i_din.e2fs_din->e2di_shortlink, 0,
253 			(u_int)oip->i_e2fs_size);
254 		oip->i_e2fs_size = 0;
255 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
256 		return (VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT));
257 	}
258 	if (oip->i_e2fs_size == length) {
259 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
260 		return (VOP_UPDATE(ovp, NULL, NULL, 0));
261 	}
262 	fs = oip->i_e2fs;
263 	osize = oip->i_e2fs_size;
264 	/*
265 	 * Lengthen the size of the file. We must ensure that the
266 	 * last byte of the file is allocated. Since the smallest
267 	 * value of osize is 0, length will be at least 1.
268 	 */
269 	if (osize < length) {
270 #if 0 /* XXX */
271 		if (length > fs->fs_maxfilesize)
272 			return (EFBIG);
273 #endif
274 		ufs_balloc_range(ovp, length - 1, 1, ap->a_cred,
275 		    ap->a_flags & IO_SYNC ? B_SYNC : 0);
276 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
277 		return (VOP_UPDATE(ovp, NULL, NULL, 1));
278 	}
279 	/*
280 	 * Shorten the size of the file. If the file is not being
281 	 * truncated to a block boundry, the contents of the
282 	 * partial block following the end of the file must be
283 	 * zero'ed in case it ever become accessible again because
284 	 * of subsequent file growth.
285 	 */
286 	offset = blkoff(fs, length);
287 	if (offset != 0) {
288 		size = fs->e2fs_bsize;
289 
290 		/* XXXUBC we should handle more than just VREG */
291 		uvm_vnp_zerorange(ovp, length, size - offset);
292 	}
293 	oip->i_e2fs_size = length;
294 	uvm_vnp_setsize(ovp, length);
295 
296 	/*
297 	 * Calculate index into inode's block list of
298 	 * last direct and indirect blocks (if any)
299 	 * which we want to keep.  Lastblock is -1 when
300 	 * the file is truncated to 0.
301 	 */
302 	lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
303 	lastiblock[SINGLE] = lastblock - NDADDR;
304 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
305 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
306 	nblocks = btodb(fs->e2fs_bsize);
307 	/*
308 	 * Update file and block pointers on disk before we start freeing
309 	 * blocks.  If we crash before free'ing blocks below, the blocks
310 	 * will be returned to the free list.  lastiblock values are also
311 	 * normalized to -1 for calls to ext2fs_indirtrunc below.
312 	 */
313 	memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
314 	for (level = TRIPLE; level >= SINGLE; level--)
315 		if (lastiblock[level] < 0) {
316 			oip->i_e2fs_blocks[NDADDR + level] = 0;
317 			lastiblock[level] = -1;
318 		}
319 	for (i = NDADDR - 1; i > lastblock; i--)
320 		oip->i_e2fs_blocks[i] = 0;
321 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
322 	error = VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT);
323 	if (error && !allerror)
324 		allerror = error;
325 
326 	/*
327 	 * Having written the new inode to disk, save its new configuration
328 	 * and put back the old block pointers long enough to process them.
329 	 * Note that we save the new block configuration so we can check it
330 	 * when we are done.
331 	 */
332 
333 	memcpy((caddr_t)newblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof newblks);
334 	memcpy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)oldblks, sizeof oldblks);
335 	oip->i_e2fs_size = osize;
336 	error = vtruncbuf(ovp, lastblock + 1, 0, 0);
337 	if (error && !allerror)
338 		allerror = error;
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 		/* XXX ondisk32 */
348 		bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]);
349 		if (bn != 0) {
350 			error = ext2fs_indirtrunc(oip, indir_lbn[level],
351 			    fsbtodb(fs, bn), lastiblock[level], level, &count);
352 			if (error)
353 				allerror = error;
354 			blocksreleased += count;
355 			if (lastiblock[level] < 0) {
356 				oip->i_e2fs_blocks[NDADDR + level] = 0;
357 				ext2fs_blkfree(oip, bn);
358 				blocksreleased += nblocks;
359 			}
360 		}
361 		if (lastiblock[level] >= 0)
362 			goto done;
363 	}
364 
365 	/*
366 	 * All whole direct blocks or frags.
367 	 */
368 	for (i = NDADDR - 1; i > lastblock; i--) {
369 		/* XXX ondisk32 */
370 		bn = fs2h32(oip->i_e2fs_blocks[i]);
371 		if (bn == 0)
372 			continue;
373 		oip->i_e2fs_blocks[i] = 0;
374 		ext2fs_blkfree(oip, bn);
375 		blocksreleased += btodb(fs->e2fs_bsize);
376 	}
377 
378 done:
379 #ifdef DIAGNOSTIC
380 	for (level = SINGLE; level <= TRIPLE; level++)
381 		if (newblks[NDADDR + level] !=
382 		    oip->i_e2fs_blocks[NDADDR + level])
383 			panic("ext2fs_truncate1");
384 	for (i = 0; i < NDADDR; i++)
385 		if (newblks[i] != oip->i_e2fs_blocks[i])
386 			panic("ext2fs_truncate2");
387 	if (length == 0 &&
388 	    (!LIST_EMPTY(&ovp->v_cleanblkhd) ||
389 	     !LIST_EMPTY(&ovp->v_dirtyblkhd)))
390 		panic("ext2fs_truncate3");
391 #endif /* DIAGNOSTIC */
392 	/*
393 	 * Put back the real size.
394 	 */
395 	oip->i_e2fs_size = length;
396 	oip->i_e2fs_nblock -= blocksreleased;
397 	oip->i_flag |= IN_CHANGE;
398 	return (allerror);
399 }
400 
401 /*
402  * Release blocks associated with the inode ip and stored in the indirect
403  * block bn.  Blocks are free'd in LIFO order up to (but not including)
404  * lastbn.  If level is greater than SINGLE, the block is an indirect block
405  * and recursive calls to indirtrunc must be used to cleanse other indirect
406  * blocks.
407  *
408  * NB: triple indirect blocks are untested.
409  */
410 static int
411 ext2fs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
412 	struct inode *ip;
413 	daddr_t lbn, lastbn;
414 	daddr_t dbn;
415 	int level;
416 	long *countp;
417 {
418 	int i;
419 	struct buf *bp;
420 	struct m_ext2fs *fs = ip->i_e2fs;
421 	int32_t *bap;	/* XXX ondisk32 */
422 	struct vnode *vp;
423 	daddr_t nb, nlbn, last;
424 	int32_t *copy = NULL;	/* XXX ondisk32 */
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(vp, bp);
462 		error = biowait(bp);
463 	}
464 	if (error) {
465 		brelse(bp);
466 		*countp = 0;
467 		return (error);
468 	}
469 
470 	bap = (int32_t *)bp->b_data;	/* XXX ondisk32 */
471 	if (lastbn >= 0) {
472 		/* XXX ondisk32 */
473 		MALLOC(copy, int32_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK);
474 		memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize);
475 		memset((caddr_t)&bap[last + 1], 0,
476 			(u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t));
477 		error = bwrite(bp);
478 		if (error)
479 			allerror = error;
480 		bap = copy;
481 	}
482 
483 	/*
484 	 * Recursively free totally unused blocks.
485 	 */
486 	for (i = NINDIR(fs) - 1,
487 		nlbn = lbn + 1 - i * factor; i > last;
488 		i--, nlbn += factor) {
489 		/* XXX ondisk32 */
490 		nb = fs2h32(bap[i]);
491 		if (nb == 0)
492 			continue;
493 		if (level > SINGLE) {
494 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
495 						   (daddr_t)-1, level - 1,
496 						   &blkcount);
497 			if (error)
498 				allerror = error;
499 			blocksreleased += blkcount;
500 		}
501 		ext2fs_blkfree(ip, nb);
502 		blocksreleased += nblocks;
503 	}
504 
505 	/*
506 	 * Recursively free last partial block.
507 	 */
508 	if (level > SINGLE && lastbn >= 0) {
509 		last = lastbn % factor;
510 		/* XXX ondisk32 */
511 		nb = fs2h32(bap[i]);
512 		if (nb != 0) {
513 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
514 						   last, level - 1, &blkcount);
515 			if (error)
516 				allerror = error;
517 			blocksreleased += blkcount;
518 		}
519 	}
520 
521 	if (copy != NULL) {
522 		FREE(copy, M_TEMP);
523 	} else {
524 		bp->b_flags |= B_INVAL;
525 		brelse(bp);
526 	}
527 
528 	*countp = blocksreleased;
529 	return (allerror);
530 }
531