xref: /openbsd-src/sys/ufs/ext2fs/ext2fs_inode.c (revision b2ea75c1b17e1a9a339660e7ed45cd24946b230e)
1 /*	$OpenBSD: ext2fs_inode.c,v 1.14 2001/07/16 03:44:22 csapuntz Exp $	*/
2 /*	$NetBSD: ext2fs_inode.c,v 1.1 1997/06/11 09:33:56 bouyer Exp $	*/
3 
4 /*
5  * Copyright (c) 1997 Manuel Bouyer.
6  * Copyright (c) 1982, 1986, 1989, 1993
7  *	The Regents of the University of California.  All rights reserved.
8  *
9  * Redistribution and use in source and binary forms, with or without
10  * modification, are permitted provided that the following conditions
11  * are met:
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer.
14  * 2. Redistributions in binary form must reproduce the above copyright
15  *    notice, this list of conditions and the following disclaimer in the
16  *    documentation and/or other materials provided with the distribution.
17  * 3. All advertising materials mentioning features or use of this software
18  *    must display the following acknowledgement:
19  *	This product includes software developed by the University of
20  *	California, Berkeley and its contributors.
21  * 4. Neither the name of the University nor the names of its contributors
22  *    may be used to endorse or promote products derived from this software
23  *    without specific prior written permission.
24  *
25  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35  * SUCH DAMAGE.
36  *
37  *	@(#)ffs_inode.c	8.8 (Berkeley) 10/19/94
38  * Modified for ext2fs by Manuel Bouyer.
39  */
40 
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/mount.h>
44 #include <sys/proc.h>
45 #include <sys/file.h>
46 #include <sys/buf.h>
47 #include <sys/vnode.h>
48 #include <sys/kernel.h>
49 #include <sys/malloc.h>
50 #include <sys/resourcevar.h>
51 
52 #include <vm/vm.h>
53 
54 #include <ufs/ufs/quota.h>
55 #include <ufs/ufs/inode.h>
56 #include <ufs/ufs/ufsmount.h>
57 #include <ufs/ufs/ufs_extern.h>
58 
59 #include <ufs/ext2fs/ext2fs.h>
60 #include <ufs/ext2fs/ext2fs_extern.h>
61 
62 static int ext2fs_indirtrunc __P((struct inode *, daddr_t, daddr_t,
63     				  daddr_t, int, long *));
64 
65 int
66 ext2fs_init(vfsp)
67 	struct vfsconf *vfsp;
68 {
69 	return (ufs_init(vfsp));
70 }
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 	} */ *ap = v;
82 	register struct vnode *vp = ap->a_vp;
83 	register struct inode *ip = VTOI(vp);
84 	struct timespec ts;
85 	int error = 0;
86 	extern int prtactive;
87 
88 	if (prtactive && vp->v_usecount != 0)
89 		vprint("ext2fs_inactive: pushing active", vp);
90 	/* Get rid of inodes related to stale file handles. */
91 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0)
92 		goto out;
93 
94 	if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
95 		(void) ext2fs_truncate(ip, (off_t)0, 0, NOCRED);
96 		TIMEVAL_TO_TIMESPEC(&time, &ts);
97 		ip->i_e2fs_dtime = ts.tv_sec;
98 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
99 		ext2fs_inode_free(ip, ip->i_number, ip->i_e2fs_mode);
100 	}
101 	if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) {
102 		ext2fs_update(ip, NULL, NULL, 0);
103 	}
104 out:
105 	VOP_UNLOCK(vp, 0, ap->a_p);
106 	/*
107 	 * If we are done with the inode, reclaim it
108 	 * so that it can be reused immediately.
109 	 */
110 	if (ip->i_e2fs_dtime != 0)
111 		vrecycle(vp, (struct simplelock *)0, ap->a_p);
112 	return (error);
113 }
114 
115 
116 /*
117  * Update the access, modified, and inode change times as specified by the
118  * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
119  * used to specify that the inode needs to be updated but that the times have
120  * already been set. The access and modified times are taken from the second
121  * and third parameters; the inode change time is always taken from the current
122  * time. If waitfor is set, then wait for the disk write of the inode to
123  * complete.
124  */
125 int
126 ext2fs_update(struct inode *ip, struct timespec *atime, struct timespec *mtime,
127     int waitfor)
128 {
129 	struct m_ext2fs *fs;
130 	struct buf *bp;
131 	int error;
132 	struct timespec ts;
133 
134 	if (ITOV(ip)->v_mount->mnt_flag & MNT_RDONLY)
135 		return (0);
136 	TIMEVAL_TO_TIMESPEC(&time, &ts);
137 	EXT2FS_ITIMES(ip,
138 	    atime ? atime : &ts,
139 	    mtime ? mtime : &ts);
140 	if ((ip->i_flag & IN_MODIFIED) == 0)
141 		return (0);
142 	ip->i_flag &= ~IN_MODIFIED;
143 	fs = ip->i_e2fs;
144 	error = bread(ip->i_devvp,
145 			  fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
146 			  (int)fs->e2fs_bsize, NOCRED, &bp);
147 	if (error) {
148 		brelse(bp);
149 		return (error);
150 	}
151 	bcopy(&ip->i_din.e2fs_din,
152 		((struct ext2fs_dinode *)bp->b_data + ino_to_fsbo(fs, ip->i_number)),
153 		sizeof(struct ext2fs_dinode));
154 	if (waitfor)
155 		return (bwrite(bp));
156 	else {
157 		bdwrite(bp);
158 		return (0);
159 	}
160 }
161 
162 #define	SINGLE	0	/* index of single indirect block */
163 #define	DOUBLE	1	/* index of double indirect block */
164 #define	TRIPLE	2	/* index of triple indirect block */
165 /*
166  * Truncate the inode oip to at most length size, freeing the
167  * disk blocks.
168  */
169 int
170 ext2fs_truncate(struct inode *oip, off_t length, int flags, struct ucred *cred)
171 {
172 	struct vnode *ovp = ITOV(oip);
173 	daddr_t lastblock;
174 	daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
175 	daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
176 	register struct m_ext2fs *fs;
177 	struct buf *bp;
178 	int offset, size, level;
179 	long count, nblocks, vflags, blocksreleased = 0;
180 	register int i;
181 	int aflags, error, allerror;
182 	off_t osize;
183 
184 	if (length < 0)
185 		return (EINVAL);
186 
187 	if (ovp->v_type != VREG &&
188 	    ovp->v_type != VDIR &&
189 	    ovp->v_type != VLNK)
190 		return (0);
191 
192 	if (ovp->v_type == VLNK &&
193 		(oip->i_e2fs_size < ovp->v_mount->mnt_maxsymlinklen ||
194 		 (ovp->v_mount->mnt_maxsymlinklen == 0 &&
195 		  oip->i_e2fs_nblock == 0))) {
196 #ifdef DIAGNOSTIC
197 		if (length != 0)
198 			panic("ext2fs_truncate: partial truncate of symlink");
199 #endif
200 		bzero((char *)&oip->i_din.e2fs_din.e2di_shortlink,
201 			(u_int)oip->i_e2fs_size);
202 		oip->i_e2fs_size = 0;
203 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
204 		return (ext2fs_update(oip, NULL, NULL, 1));
205 	}
206 	if (oip->i_e2fs_size == length) {
207 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
208 		return (ext2fs_update(oip, NULL, NULL, 0));
209 	}
210 	fs = oip->i_e2fs;
211 	osize = oip->i_e2fs_size;
212 	/*
213 	 * Lengthen the size of the file. We must ensure that the
214 	 * last byte of the file is allocated. Since the smallest
215 	 * value of osize is 0, length will be at least 1.
216 	 */
217 	if (osize < length) {
218 #if 0 /* XXX */
219 		if (length > fs->fs_maxfilesize)
220 			return (EFBIG);
221 #endif
222 		offset = blkoff(fs, length - 1);
223 		lbn = lblkno(fs, length - 1);
224 		aflags = B_CLRBUF;
225 		if (flags & IO_SYNC)
226 			aflags |= B_SYNC;
227 		error = ext2fs_buf_alloc(oip, lbn, offset + 1, cred, &bp,
228 		    aflags);
229 		if (error)
230 			return (error);
231 		oip->i_e2fs_size = length;
232 		uvm_vnp_setsize(ovp, length);
233 		uvm_vnp_uncache(ovp);
234 		if (aflags & B_SYNC)
235 			bwrite(bp);
236 		else
237 			bawrite(bp);
238 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
239 		return (ext2fs_update(oip, NULL, NULL, 1));
240 	}
241 	/*
242 	 * Shorten the size of the file. If the file is not being
243 	 * truncated to a block boundry, the contents of the
244 	 * partial block following the end of the file must be
245 	 * zero'ed in case it ever become accessable again because
246 	 * of subsequent file growth.
247 	 */
248 	offset = blkoff(fs, length);
249 	if (offset == 0) {
250 		oip->i_e2fs_size = length;
251 	} else {
252 		lbn = lblkno(fs, length);
253 		aflags = B_CLRBUF;
254 		if (flags & IO_SYNC)
255 			aflags |= B_SYNC;
256 		error = ext2fs_buf_alloc(oip, lbn, offset, cred, &bp,
257 		    aflags);
258 		if (error)
259 			return (error);
260 		oip->i_e2fs_size = length;
261 		size = fs->e2fs_bsize;
262 		uvm_vnp_setsize(ovp, length);
263 		uvm_vnp_uncache(ovp);
264 		bzero((char *)bp->b_data + offset, (u_int)(size - offset));
265 		allocbuf(bp, size);
266 		if (aflags & B_SYNC)
267 			bwrite(bp);
268 		else
269 			bawrite(bp);
270 	}
271 	/*
272 	 * Calculate index into inode's block list of
273 	 * last direct and indirect blocks (if any)
274 	 * which we want to keep.  Lastblock is -1 when
275 	 * the file is truncated to 0.
276 	 */
277 	lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
278 	lastiblock[SINGLE] = lastblock - NDADDR;
279 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
280 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
281 	nblocks = btodb(fs->e2fs_bsize);
282 	/*
283 	 * Update file and block pointers on disk before we start freeing
284 	 * blocks.  If we crash before free'ing blocks below, the blocks
285 	 * will be returned to the free list.  lastiblock values are also
286 	 * normalized to -1 for calls to ext2fs_indirtrunc below.
287 	 */
288 	bcopy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)oldblks, sizeof oldblks);
289 	for (level = TRIPLE; level >= SINGLE; level--)
290 		if (lastiblock[level] < 0) {
291 			oip->i_e2fs_blocks[NDADDR + level] = 0;
292 			lastiblock[level] = -1;
293 		}
294 	for (i = NDADDR - 1; i > lastblock; i--)
295 		oip->i_e2fs_blocks[i] = 0;
296 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
297 	if ((error = ext2fs_update(oip, NULL, NULL, 1)) != 0)
298 		allerror = error;
299 	/*
300 	 * Having written the new inode to disk, save its new configuration
301 	 * and put back the old block pointers long enough to process them.
302 	 * Note that we save the new block configuration so we can check it
303 	 * when we are done.
304 	 */
305 	bcopy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)newblks, sizeof newblks);
306 	bcopy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
307 	oip->i_e2fs_size = osize;
308 	vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
309 	allerror = vinvalbuf(ovp, vflags, cred, curproc, 0, 0);
310 
311 	/*
312 	 * Indirect blocks first.
313 	 */
314 	indir_lbn[SINGLE] = -NDADDR;
315 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1;
316 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
317 	for (level = TRIPLE; level >= SINGLE; level--) {
318 		bn = oip->i_e2fs_blocks[NDADDR + level];
319 		if (bn != 0) {
320 			error = ext2fs_indirtrunc(oip, indir_lbn[level],
321 				fsbtodb(fs, bn), lastiblock[level], level, &count);
322 			if (error)
323 				allerror = error;
324 			blocksreleased += count;
325 			if (lastiblock[level] < 0) {
326 				oip->i_e2fs_blocks[NDADDR + level] = 0;
327 				ext2fs_blkfree(oip, bn);
328 				blocksreleased += nblocks;
329 			}
330 		}
331 		if (lastiblock[level] >= 0)
332 			goto done;
333 	}
334 
335 	/*
336 	 * All whole direct blocks or frags.
337 	 */
338 	for (i = NDADDR - 1; i > lastblock; i--) {
339 		bn = oip->i_e2fs_blocks[i];
340 		if (bn == 0)
341 			continue;
342 		oip->i_e2fs_blocks[i] = 0;
343 		ext2fs_blkfree(oip, bn);
344 		blocksreleased += btodb(fs->e2fs_bsize);
345 	}
346 	if (lastblock < 0)
347 		goto done;
348 
349 done:
350 #ifdef DIAGNOSTIC
351 	for (level = SINGLE; level <= TRIPLE; level++)
352 		if (newblks[NDADDR + level] != oip->i_e2fs_blocks[NDADDR + level])
353 			panic("itrunc1");
354 	for (i = 0; i < NDADDR; i++)
355 		if (newblks[i] != oip->i_e2fs_blocks[i])
356 			panic("itrunc2");
357 	if (length == 0 &&
358 		(ovp->v_dirtyblkhd.lh_first || ovp->v_cleanblkhd.lh_first))
359 		panic("itrunc3");
360 #endif /* DIAGNOSTIC */
361 	/*
362 	 * Put back the real size.
363 	 */
364 	oip->i_e2fs_size = length;
365 	oip->i_e2fs_nblock -= blocksreleased;
366 	if (oip->i_e2fs_nblock < 0)			/* sanity */
367 		oip->i_e2fs_nblock = 0;
368 	oip->i_flag |= IN_CHANGE;
369 	return (allerror);
370 }
371 
372 /*
373  * Release blocks associated with the inode ip and stored in the indirect
374  * block bn.  Blocks are free'd in LIFO order up to (but not including)
375  * lastbn.  If level is greater than SINGLE, the block is an indirect block
376  * and recursive calls to indirtrunc must be used to cleanse other indirect
377  * blocks.
378  *
379  * NB: triple indirect blocks are untested.
380  */
381 static int
382 ext2fs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
383 	register struct inode *ip;
384 	daddr_t lbn, lastbn;
385 	daddr_t dbn;
386 	int level;
387 	long *countp;
388 {
389 	register int i;
390 	struct buf *bp;
391 	register struct m_ext2fs *fs = ip->i_e2fs;
392 	register daddr_t *bap;
393 	struct vnode *vp;
394 	daddr_t *copy = NULL, nb, nlbn, last;
395 	long blkcount, factor;
396 	int nblocks, blocksreleased = 0;
397 	int error = 0, allerror = 0;
398 
399 	/*
400 	 * Calculate index in current block of last
401 	 * block to be kept.  -1 indicates the entire
402 	 * block so we need not calculate the index.
403 	 */
404 	factor = 1;
405 	for (i = SINGLE; i < level; i++)
406 		factor *= NINDIR(fs);
407 	last = lastbn;
408 	if (lastbn > 0)
409 		last /= factor;
410 	nblocks = btodb(fs->e2fs_bsize);
411 	/*
412 	 * Get buffer of block pointers, zero those entries corresponding
413 	 * to blocks to be free'd, and update on disk copy first.  Since
414 	 * double(triple) indirect before single(double) indirect, calls
415 	 * to bmap on these blocks will fail.  However, we already have
416 	 * the on disk address, so we have to set the b_blkno field
417 	 * explicitly instead of letting bread do everything for us.
418 	 */
419 	vp = ITOV(ip);
420 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0);
421 	if (!(bp->b_flags & (B_DONE | B_DELWRI))) {
422 		curproc->p_stats->p_ru.ru_inblock++;	/* pay for read */
423 		bp->b_flags |= B_READ;
424 		if (bp->b_bcount > bp->b_bufsize)
425 			panic("ext2fs_indirtrunc: bad buffer size");
426 		bp->b_blkno = dbn;
427 		VOP_STRATEGY(bp);
428 		error = biowait(bp);
429 	}
430 	if (error) {
431 		brelse(bp);
432 		*countp = 0;
433 		return (error);
434 	}
435 
436 	bap = (daddr_t *)bp->b_data;
437 	if (lastbn != -1) {
438 		MALLOC(copy, daddr_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK);
439 		bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize);
440 		bzero((caddr_t)&bap[last + 1],
441 			(u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t));
442 		error = bwrite(bp);
443 		if (error)
444 			allerror = error;
445 		bap = copy;
446 	}
447 
448 	/*
449 	 * Recursively free totally unused blocks.
450 	 */
451 	for (i = NINDIR(fs) - 1,
452 		nlbn = lbn + 1 - i * factor; i > last;
453 		i--, nlbn += factor) {
454 		nb = bap[i];
455 		if (nb == 0)
456 			continue;
457 		if (level > SINGLE) {
458 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
459 						   (daddr_t)-1, level - 1,
460 						   &blkcount);
461 			if (error)
462 				allerror = error;
463 			blocksreleased += blkcount;
464 		}
465 		ext2fs_blkfree(ip, nb);
466 		blocksreleased += nblocks;
467 	}
468 
469 	/*
470 	 * Recursively free last partial block.
471 	 */
472 	if (level > SINGLE && lastbn >= 0) {
473 		last = lastbn % factor;
474 		nb = bap[i];
475 		if (nb != 0) {
476 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
477 						   last, level - 1, &blkcount);
478 			if (error)
479 				allerror = error;
480 			blocksreleased += blkcount;
481 		}
482 	}
483 
484 	if (copy != NULL) {
485 		FREE(copy, M_TEMP);
486 	} else {
487 		bp->b_flags |= B_INVAL;
488 		brelse(bp);
489 	}
490 
491 	*countp = blocksreleased;
492 	return (allerror);
493 }
494