xref: /netbsd-src/sys/ufs/ext2fs/ext2fs_inode.c (revision d20841bb642898112fe68f0ad3f7b26dddf56f07)
1 /*	$NetBSD: ext2fs_inode.c,v 1.39 2004/01/25 18:06:49 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 1.39 2004/01/25 18:06:49 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 		vn_start_write(vp, &mp, V_WAIT | V_LOWER);
122 		if (ip->i_e2fs_size != 0) {
123 			error = VOP_TRUNCATE(vp, (off_t)0, 0, NOCRED, NULL);
124 		}
125 		TIMEVAL_TO_TIMESPEC(&time, &ts);
126 		ip->i_e2fs_dtime = ts.tv_sec;
127 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
128 		VOP_VFREE(vp, ip->i_number, ip->i_e2fs_mode);
129 		vn_finished_write(mp, V_LOWER);
130 	}
131 	if (ip->i_flag &
132 	    (IN_ACCESS | IN_CHANGE | IN_UPDATE | IN_MODIFIED | IN_ACCESSED)) {
133 		vn_start_write(vp, &mp, V_WAIT | V_LOWER);
134 		VOP_UPDATE(vp, NULL, NULL, 0);
135 		vn_finished_write(mp, V_LOWER);
136 	}
137 out:
138 	VOP_UNLOCK(vp, 0);
139 	/*
140 	 * If we are done with the inode, reclaim it
141 	 * so that it can be reused immediately.
142 	 */
143 	if (ip->i_e2fs_dtime != 0)
144 		vrecycle(vp, NULL, p);
145 	return (error);
146 }
147 
148 
149 /*
150  * Update the access, modified, and inode change times as specified by the
151  * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
152  * used to specify that the inode needs to be updated but that the times have
153  * already been set. The access and modified times are taken from the second
154  * and third parameters; the inode change time is always taken from the current
155  * time. If UPDATE_WAIT or UPDATE_DIROP is set, then wait for the disk
156  * write of the inode to complete.
157  */
158 int
159 ext2fs_update(v)
160 	void *v;
161 {
162 	struct vop_update_args /* {
163 		struct vnode *a_vp;
164 		struct timespec *a_access;
165 		struct timespec *a_modify;
166 		int a_flags;
167 	} */ *ap = v;
168 	struct m_ext2fs *fs;
169 	struct buf *bp;
170 	struct inode *ip;
171 	int error;
172 	struct timespec ts;
173 	caddr_t cp;
174 	int flags;
175 
176 	if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
177 		return (0);
178 	ip = VTOI(ap->a_vp);
179 	TIMEVAL_TO_TIMESPEC(&time, &ts);
180 	EXT2FS_ITIMES(ip,
181 	    ap->a_access ? ap->a_access : &ts,
182 	    ap->a_modify ? ap->a_modify : &ts, &ts);
183 	flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED);
184 	if (flags == 0)
185 		return (0);
186 	fs = ip->i_e2fs;
187 
188 	error = bread(ip->i_devvp,
189 			  fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
190 			  (int)fs->e2fs_bsize, NOCRED, &bp);
191 	if (error) {
192 		brelse(bp);
193 		return (error);
194 	}
195 	ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED);
196 	cp = (caddr_t)bp->b_data +
197 	    (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
198 	e2fs_isave(ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp);
199 	if ((ap->a_flags & (UPDATE_WAIT|UPDATE_DIROP)) != 0 &&
200 	    (flags & IN_MODIFIED) != 0 &&
201 	    (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0)
202 		return (bwrite(bp));
203 	else {
204 		bdwrite(bp);
205 		return (0);
206 	}
207 }
208 
209 #define	SINGLE	0	/* index of single indirect block */
210 #define	DOUBLE	1	/* index of double indirect block */
211 #define	TRIPLE	2	/* index of triple indirect block */
212 /*
213  * Truncate the inode oip to at most length size, freeing the
214  * disk blocks.
215  */
216 int
217 ext2fs_truncate(v)
218 	void *v;
219 {
220 	struct vop_truncate_args /* {
221 		struct vnode *a_vp;
222 		off_t a_length;
223 		int a_flags;
224 		struct ucred *a_cred;
225 		struct proc *a_p;
226 	} */ *ap = v;
227 	struct vnode *ovp = ap->a_vp;
228 	daddr_t lastblock;
229 	struct inode *oip;
230 	daddr_t bn, lastiblock[NIADDR], indir_lbn[NIADDR];
231 	/* XXX ondisk32 */
232 	int32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
233 	off_t length = ap->a_length;
234 	struct m_ext2fs *fs;
235 	int offset, size, level;
236 	long count, nblocks, blocksreleased = 0;
237 	int i;
238 	int error, allerror = 0;
239 	off_t osize;
240 
241 	if (length < 0)
242 		return (EINVAL);
243 
244 	oip = VTOI(ovp);
245 	if (ovp->v_type == VLNK &&
246 		(oip->i_e2fs_size < ovp->v_mount->mnt_maxsymlinklen ||
247 		 (ovp->v_mount->mnt_maxsymlinklen == 0 &&
248 		  oip->i_e2fs_nblock == 0))) {
249 #ifdef DIAGNOSTIC
250 		if (length != 0)
251 			panic("ext2fs_truncate: partial truncate of symlink");
252 #endif
253 		memset((char *)&oip->i_din.e2fs_din->e2di_shortlink, 0,
254 			(u_int)oip->i_e2fs_size);
255 		oip->i_e2fs_size = 0;
256 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
257 		return (VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT));
258 	}
259 	if (oip->i_e2fs_size == length) {
260 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
261 		return (VOP_UPDATE(ovp, NULL, NULL, 0));
262 	}
263 	fs = oip->i_e2fs;
264 	osize = oip->i_e2fs_size;
265 	/*
266 	 * Lengthen the size of the file. We must ensure that the
267 	 * last byte of the file is allocated. Since the smallest
268 	 * value of osize is 0, length will be at least 1.
269 	 */
270 	if (osize < length) {
271 #if 0 /* XXX */
272 		if (length > fs->fs_maxfilesize)
273 			return (EFBIG);
274 #endif
275 		ufs_balloc_range(ovp, length - 1, 1, ap->a_cred,
276 		    ap->a_flags & IO_SYNC ? B_SYNC : 0);
277 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
278 		return (VOP_UPDATE(ovp, NULL, NULL, 1));
279 	}
280 	/*
281 	 * Shorten the size of the file. If the file is not being
282 	 * truncated to a block boundry, the contents of the
283 	 * partial block following the end of the file must be
284 	 * zero'ed in case it ever become accessible again because
285 	 * of subsequent file growth.
286 	 */
287 	offset = blkoff(fs, length);
288 	if (offset != 0) {
289 		size = fs->e2fs_bsize;
290 
291 		/* XXXUBC we should handle more than just VREG */
292 		uvm_vnp_zerorange(ovp, length, size - offset);
293 	}
294 	oip->i_e2fs_size = length;
295 	uvm_vnp_setsize(ovp, length);
296 
297 	/*
298 	 * Calculate index into inode's block list of
299 	 * last direct and indirect blocks (if any)
300 	 * which we want to keep.  Lastblock is -1 when
301 	 * the file is truncated to 0.
302 	 */
303 	lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
304 	lastiblock[SINGLE] = lastblock - NDADDR;
305 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
306 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
307 	nblocks = btodb(fs->e2fs_bsize);
308 	/*
309 	 * Update file and block pointers on disk before we start freeing
310 	 * blocks.  If we crash before free'ing blocks below, the blocks
311 	 * will be returned to the free list.  lastiblock values are also
312 	 * normalized to -1 for calls to ext2fs_indirtrunc below.
313 	 */
314 	memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
315 	for (level = TRIPLE; level >= SINGLE; level--)
316 		if (lastiblock[level] < 0) {
317 			oip->i_e2fs_blocks[NDADDR + level] = 0;
318 			lastiblock[level] = -1;
319 		}
320 	for (i = NDADDR - 1; i > lastblock; i--)
321 		oip->i_e2fs_blocks[i] = 0;
322 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
323 	error = VOP_UPDATE(ovp, NULL, NULL, UPDATE_WAIT);
324 	if (error && !allerror)
325 		allerror = error;
326 
327 	/*
328 	 * Having written the new inode to disk, save its new configuration
329 	 * and put back the old block pointers long enough to process them.
330 	 * Note that we save the new block configuration so we can check it
331 	 * when we are done.
332 	 */
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 	error = vtruncbuf(ovp, lastblock + 1, 0, 0);
338 	if (error && !allerror)
339 		allerror = error;
340 
341 	/*
342 	 * Indirect blocks first.
343 	 */
344 	indir_lbn[SINGLE] = -NDADDR;
345 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1;
346 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
347 	for (level = TRIPLE; level >= SINGLE; level--) {
348 		/* XXX ondisk32 */
349 		bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]);
350 		if (bn != 0) {
351 			error = ext2fs_indirtrunc(oip, indir_lbn[level],
352 			    fsbtodb(fs, bn), lastiblock[level], level, &count);
353 			if (error)
354 				allerror = error;
355 			blocksreleased += count;
356 			if (lastiblock[level] < 0) {
357 				oip->i_e2fs_blocks[NDADDR + level] = 0;
358 				ext2fs_blkfree(oip, bn);
359 				blocksreleased += nblocks;
360 			}
361 		}
362 		if (lastiblock[level] >= 0)
363 			goto done;
364 	}
365 
366 	/*
367 	 * All whole direct blocks or frags.
368 	 */
369 	for (i = NDADDR - 1; i > lastblock; i--) {
370 		/* XXX ondisk32 */
371 		bn = fs2h32(oip->i_e2fs_blocks[i]);
372 		if (bn == 0)
373 			continue;
374 		oip->i_e2fs_blocks[i] = 0;
375 		ext2fs_blkfree(oip, bn);
376 		blocksreleased += btodb(fs->e2fs_bsize);
377 	}
378 
379 done:
380 #ifdef DIAGNOSTIC
381 	for (level = SINGLE; level <= TRIPLE; level++)
382 		if (newblks[NDADDR + level] !=
383 		    oip->i_e2fs_blocks[NDADDR + level])
384 			panic("ext2fs_truncate1");
385 	for (i = 0; i < NDADDR; i++)
386 		if (newblks[i] != oip->i_e2fs_blocks[i])
387 			panic("ext2fs_truncate2");
388 	if (length == 0 &&
389 	    (!LIST_EMPTY(&ovp->v_cleanblkhd) ||
390 	     !LIST_EMPTY(&ovp->v_dirtyblkhd)))
391 		panic("ext2fs_truncate3");
392 #endif /* DIAGNOSTIC */
393 	/*
394 	 * Put back the real size.
395 	 */
396 	oip->i_e2fs_size = length;
397 	oip->i_e2fs_nblock -= blocksreleased;
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 	struct inode *ip;
414 	daddr_t lbn, lastbn;
415 	daddr_t dbn;
416 	int level;
417 	long *countp;
418 {
419 	int i;
420 	struct buf *bp;
421 	struct m_ext2fs *fs = ip->i_e2fs;
422 	int32_t *bap;	/* XXX ondisk32 */
423 	struct vnode *vp;
424 	daddr_t nb, nlbn, last;
425 	int32_t *copy = NULL;	/* XXX ondisk32 */
426 	long blkcount, factor;
427 	int nblocks, blocksreleased = 0;
428 	int error = 0, allerror = 0;
429 
430 	/*
431 	 * Calculate index in current block of last
432 	 * block to be kept.  -1 indicates the entire
433 	 * block so we need not calculate the index.
434 	 */
435 	factor = 1;
436 	for (i = SINGLE; i < level; i++)
437 		factor *= NINDIR(fs);
438 	last = lastbn;
439 	if (lastbn > 0)
440 		last /= factor;
441 	nblocks = btodb(fs->e2fs_bsize);
442 	/*
443 	 * Get buffer of block pointers, zero those entries corresponding
444 	 * to blocks to be free'd, and update on disk copy first.  Since
445 	 * double(triple) indirect before single(double) indirect, calls
446 	 * to bmap on these blocks will fail.  However, we already have
447 	 * the on disk address, so we have to set the b_blkno field
448 	 * explicitly instead of letting bread do everything for us.
449 	 */
450 	vp = ITOV(ip);
451 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0);
452 	if (bp->b_flags & (B_DONE | B_DELWRI)) {
453 		/* Braces must be here in case trace evaluates to nothing. */
454 		trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn);
455 	} else {
456 		trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn);
457 		curproc->p_stats->p_ru.ru_inblock++;	/* pay for read */
458 		bp->b_flags |= B_READ;
459 		if (bp->b_bcount > bp->b_bufsize)
460 			panic("ext2fs_indirtrunc: bad buffer size");
461 		bp->b_blkno = dbn;
462 		VOP_STRATEGY(vp, bp);
463 		error = biowait(bp);
464 	}
465 	if (error) {
466 		brelse(bp);
467 		*countp = 0;
468 		return (error);
469 	}
470 
471 	bap = (int32_t *)bp->b_data;	/* XXX ondisk32 */
472 	if (lastbn >= 0) {
473 		/* XXX ondisk32 */
474 		MALLOC(copy, int32_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK);
475 		memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize);
476 		memset((caddr_t)&bap[last + 1], 0,
477 			(u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t));
478 		error = bwrite(bp);
479 		if (error)
480 			allerror = error;
481 		bap = copy;
482 	}
483 
484 	/*
485 	 * Recursively free totally unused blocks.
486 	 */
487 	for (i = NINDIR(fs) - 1,
488 		nlbn = lbn + 1 - i * factor; i > last;
489 		i--, nlbn += factor) {
490 		/* XXX ondisk32 */
491 		nb = fs2h32(bap[i]);
492 		if (nb == 0)
493 			continue;
494 		if (level > SINGLE) {
495 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
496 						   (daddr_t)-1, level - 1,
497 						   &blkcount);
498 			if (error)
499 				allerror = error;
500 			blocksreleased += blkcount;
501 		}
502 		ext2fs_blkfree(ip, nb);
503 		blocksreleased += nblocks;
504 	}
505 
506 	/*
507 	 * Recursively free last partial block.
508 	 */
509 	if (level > SINGLE && lastbn >= 0) {
510 		last = lastbn % factor;
511 		/* XXX ondisk32 */
512 		nb = fs2h32(bap[i]);
513 		if (nb != 0) {
514 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
515 						   last, level - 1, &blkcount);
516 			if (error)
517 				allerror = error;
518 			blocksreleased += blkcount;
519 		}
520 	}
521 
522 	if (copy != NULL) {
523 		FREE(copy, M_TEMP);
524 	} else {
525 		bp->b_flags |= B_INVAL;
526 		brelse(bp);
527 	}
528 
529 	*countp = blocksreleased;
530 	return (allerror);
531 }
532