xref: /openbsd-src/sys/ufs/ext2fs/ext2fs_inode.c (revision 47911bd667ac77dc523b8a13ef40b012dbffa741)
1 /*	$OpenBSD: ext2fs_inode.c,v 1.21 2002/03/14 01:27:14 millert Exp $	*/
2 /*	$NetBSD: ext2fs_inode.c,v 1.24 2001/06/19 12:59:18 wiz 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 <uvm/uvm_extern.h>
53 
54 #include <ufs/ufs/extattr.h>
55 #include <ufs/ufs/quota.h>
56 #include <ufs/ufs/inode.h>
57 #include <ufs/ufs/ufsmount.h>
58 #include <ufs/ufs/ufs_extern.h>
59 
60 #include <ufs/ext2fs/ext2fs.h>
61 #include <ufs/ext2fs/ext2fs_extern.h>
62 
63 static int ext2fs_indirtrunc(struct inode *, ufs_daddr_t, ufs_daddr_t,
64 				ufs_daddr_t, int, long *);
65 
66 /*
67  * Last reference to an inode.  If necessary, write or delete it.
68  */
69 int
70 ext2fs_inactive(v)
71 	void *v;
72 {
73 	struct vop_inactive_args /* {
74 		struct vnode *a_vp;
75 		struct proc *a_p;
76 	} */ *ap = v;
77 	struct vnode *vp = ap->a_vp;
78 	struct inode *ip = VTOI(vp);
79 	struct proc *p = ap->a_p;
80 	struct timespec ts;
81 	int error = 0;
82 	extern int prtactive;
83 
84 	if (prtactive && vp->v_usecount != 0)
85 		vprint("ext2fs_inactive: pushing active", vp);
86 	/* Get rid of inodes related to stale file handles. */
87 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0)
88 		goto out;
89 
90 	error = 0;
91 	if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
92 		if (ip->i_e2fs_size != 0) {
93 			error = ext2fs_truncate(ip, (off_t)0, 0, NOCRED);
94 		}
95 		TIMEVAL_TO_TIMESPEC(&time, &ts);
96 		ip->i_e2fs_dtime = ts.tv_sec;
97 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
98 		ext2fs_inode_free(ip, ip->i_number, ip->i_e2fs_mode);
99 	}
100 	if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) {
101 		ext2fs_update(ip, NULL, NULL, 0);
102 	}
103 out:
104 	VOP_UNLOCK(vp, 0, p);
105 	/*
106 	 * If we are done with the inode, reclaim it
107 	 * so that it can be reused immediately.
108 	 */
109 	if (ip->i_e2fs_dtime != 0)
110 		vrecycle(vp, NULL, p);
111 	return (error);
112 }
113 
114 
115 /*
116  * Update the access, modified, and inode change times as specified by the
117  * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
118  * used to specify that the inode needs to be updated but that the times have
119  * already been set. The access and modified times are taken from the second
120  * and third parameters; the inode change time is always taken from the current
121  * time. If waitfor is set, then wait for the disk write of the inode to
122  * complete.
123  */
124 int
125 ext2fs_update(struct inode *ip, struct timespec *atime, struct timespec *mtime,
126     int waitfor)
127 {
128 	struct m_ext2fs *fs;
129 	struct buf *bp;
130 	int error;
131 	struct timespec ts;
132 	caddr_t cp;
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 	ip->i_flag &= ~(IN_MODIFIED);
152 	cp = (caddr_t)bp->b_data +
153 	    (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE);
154 	e2fs_isave(&ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp);
155 	if (waitfor)
156 		return (bwrite(bp));
157 	else {
158 		bdwrite(bp);
159 		return (0);
160 	}
161 }
162 
163 #define	SINGLE	0	/* index of single indirect block */
164 #define	DOUBLE	1	/* index of double indirect block */
165 #define	TRIPLE	2	/* index of triple indirect block */
166 /*
167  * Truncate the inode oip to at most length size, freeing the
168  * disk blocks.
169  */
170 int
171 ext2fs_truncate(struct inode *oip, off_t length, int flags, struct ucred *cred)
172 {
173 	struct vnode *ovp = ITOV(oip);
174 	ufs_daddr_t lastblock;
175 	ufs_daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
176 	ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
177 	struct m_ext2fs *fs;
178 	struct buf *bp;
179 	int offset, size, level;
180 	long count, nblocks, vflags, blocksreleased = 0;
181 	int i;
182 	int aflags, error, allerror;
183 	off_t osize;
184 
185 	if (length < 0)
186 		return (EINVAL);
187 
188 	if (ovp->v_type != VREG &&
189 	    ovp->v_type != VDIR &&
190 	    ovp->v_type != VLNK)
191 		return (0);
192 
193 	if (ovp->v_type == VLNK &&
194 		(oip->i_e2fs_size < ovp->v_mount->mnt_maxsymlinklen ||
195 		 (ovp->v_mount->mnt_maxsymlinklen == 0 &&
196 		  oip->i_e2fs_nblock == 0))) {
197 #ifdef DIAGNOSTIC
198 		if (length != 0)
199 			panic("ext2fs_truncate: partial truncate of symlink");
200 #endif
201 		bzero((char *)&oip->i_din.e2fs_din.e2di_shortlink,
202 			(u_int)oip->i_e2fs_size);
203 		oip->i_e2fs_size = 0;
204 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
205 		return (ext2fs_update(oip, NULL, NULL, 1));
206 	}
207 	if (oip->i_e2fs_size == length) {
208 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
209 		return (ext2fs_update(oip, NULL, NULL, 0));
210 	}
211 	fs = oip->i_e2fs;
212 	osize = oip->i_e2fs_size;
213 	/*
214 	 * Lengthen the size of the file. We must ensure that the
215 	 * last byte of the file is allocated. Since the smallest
216 	 * value of osize is 0, length will be at least 1.
217 	 */
218 	if (osize < length) {
219 #if 0 /* XXX */
220 		if (length > fs->fs_maxfilesize)
221 			return (EFBIG);
222 #endif
223 		offset = blkoff(fs, length - 1);
224 		lbn = lblkno(fs, length - 1);
225 		aflags = B_CLRBUF;
226 		if (flags & IO_SYNC)
227 			aflags |= B_SYNC;
228 		error = ext2fs_buf_alloc(oip, lbn, offset + 1, cred, &bp,
229 		    aflags);
230 		if (error)
231 			return (error);
232 		oip->i_e2fs_size = length;
233 		uvm_vnp_setsize(ovp, length);
234 		uvm_vnp_uncache(ovp);
235 		if (aflags & B_SYNC)
236 			bwrite(bp);
237 		else
238 			bawrite(bp);
239 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
240 		return (ext2fs_update(oip, NULL, NULL, 1));
241 	}
242 	/*
243 	 * Shorten the size of the file. If the file is not being
244 	 * truncated to a block boundry, the contents of the
245 	 * partial block following the end of the file must be
246 	 * zero'ed in case it ever become accessible again because
247 	 * of subsequent file growth.
248 	 */
249 	offset = blkoff(fs, length);
250 	if (offset == 0) {
251 		oip->i_e2fs_size = length;
252 	} else {
253 		lbn = lblkno(fs, length);
254 		aflags = B_CLRBUF;
255 		if (flags & IO_SYNC)
256 			aflags |= B_SYNC;
257 		error = ext2fs_buf_alloc(oip, lbn, offset, cred, &bp,
258 		    aflags);
259 		if (error)
260 			return (error);
261 		oip->i_e2fs_size = length;
262 		size = fs->e2fs_bsize;
263 		uvm_vnp_setsize(ovp, length);
264 		uvm_vnp_uncache(ovp);
265 		bzero((char *)bp->b_data + offset, (u_int)(size - offset));
266 		allocbuf(bp, size);
267 		if (aflags & B_SYNC)
268 			bwrite(bp);
269 		else
270 			bawrite(bp);
271 	}
272 	/*
273 	 * Calculate index into inode's block list of
274 	 * last direct and indirect blocks (if any)
275 	 * which we want to keep.  Lastblock is -1 when
276 	 * the file is truncated to 0.
277 	 */
278 	lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
279 	lastiblock[SINGLE] = lastblock - NDADDR;
280 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
281 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
282 	nblocks = btodb(fs->e2fs_bsize);
283 	/*
284 	 * Update file and block pointers on disk before we start freeing
285 	 * blocks.  If we crash before free'ing blocks below, the blocks
286 	 * will be returned to the free list.  lastiblock values are also
287 	 * normalized to -1 for calls to ext2fs_indirtrunc below.
288 	 */
289 	memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
290 	for (level = TRIPLE; level >= SINGLE; level--)
291 		if (lastiblock[level] < 0) {
292 			oip->i_e2fs_blocks[NDADDR + level] = 0;
293 			lastiblock[level] = -1;
294 		}
295 	for (i = NDADDR - 1; i > lastblock; i--)
296 		oip->i_e2fs_blocks[i] = 0;
297 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
298 	if ((error = ext2fs_update(oip, NULL, NULL, 1)) != 0)
299 		allerror = error;
300 	/*
301 	 * Having written the new inode to disk, save its new configuration
302 	 * and put back the old block pointers long enough to process them.
303 	 * Note that we save the new block configuration so we can check it
304 	 * when we are done.
305 	 */
306 	bcopy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)newblks, sizeof newblks);
307 	bcopy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
308 	oip->i_e2fs_size = osize;
309 	vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
310 	allerror = vinvalbuf(ovp, vflags, cred, curproc, 0, 0);
311 
312 	/*
313 	 * Indirect blocks first.
314 	 */
315 	indir_lbn[SINGLE] = -NDADDR;
316 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1;
317 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
318 	for (level = TRIPLE; level >= SINGLE; level--) {
319 		bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]);
320 		if (bn != 0) {
321 			error = ext2fs_indirtrunc(oip, indir_lbn[level],
322 			    fsbtodb(fs, bn), lastiblock[level], level, &count);
323 			if (error)
324 				allerror = error;
325 			blocksreleased += count;
326 			if (lastiblock[level] < 0) {
327 				oip->i_e2fs_blocks[NDADDR + level] = 0;
328 				ext2fs_blkfree(oip, bn);
329 				blocksreleased += nblocks;
330 			}
331 		}
332 		if (lastiblock[level] >= 0)
333 			goto done;
334 	}
335 
336 	/*
337 	 * All whole direct blocks or frags.
338 	 */
339 	for (i = NDADDR - 1; i > lastblock; i--) {
340 		bn = fs2h32(oip->i_e2fs_blocks[i]);
341 		if (bn == 0)
342 			continue;
343 		oip->i_e2fs_blocks[i] = 0;
344 		ext2fs_blkfree(oip, bn);
345 		blocksreleased += btodb(fs->e2fs_bsize);
346 	}
347 
348 done:
349 #ifdef DIAGNOSTIC
350 	for (level = SINGLE; level <= TRIPLE; level++)
351 		if (newblks[NDADDR + level] !=
352 		    oip->i_e2fs_blocks[NDADDR + level])
353 			panic("ext2fs_truncate1");
354 	for (i = 0; i < NDADDR; i++)
355 		if (newblks[i] != oip->i_e2fs_blocks[i])
356 			panic("ext2fs_truncate2");
357 	if (length == 0 &&
358 	    (!LIST_EMPTY(&ovp->v_cleanblkhd) ||
359 	     !LIST_EMPTY(&ovp->v_dirtyblkhd)))
360 		panic("ext2fs_truncate3");
361 #endif /* DIAGNOSTIC */
362 	/*
363 	 * Put back the real size.
364 	 */
365 	oip->i_e2fs_size = length;
366 	oip->i_e2fs_nblock -= blocksreleased;
367 	if (oip->i_e2fs_nblock < 0)			/* sanity */
368 		oip->i_e2fs_nblock = 0;
369 	oip->i_flag |= IN_CHANGE;
370 	return (allerror);
371 }
372 
373 /*
374  * Release blocks associated with the inode ip and stored in the indirect
375  * block bn.  Blocks are free'd in LIFO order up to (but not including)
376  * lastbn.  If level is greater than SINGLE, the block is an indirect block
377  * and recursive calls to indirtrunc must be used to cleanse other indirect
378  * blocks.
379  *
380  * NB: triple indirect blocks are untested.
381  */
382 static int
383 ext2fs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
384 	struct inode *ip;
385 	ufs_daddr_t lbn, lastbn;
386 	ufs_daddr_t dbn;
387 	int level;
388 	long *countp;
389 {
390 	int i;
391 	struct buf *bp;
392 	struct m_ext2fs *fs = ip->i_e2fs;
393 	ufs_daddr_t *bap;
394 	struct vnode *vp;
395 	ufs_daddr_t *copy = NULL, nb, nlbn, last;
396 	long blkcount, factor;
397 	int nblocks, blocksreleased = 0;
398 	int error = 0, allerror = 0;
399 
400 	/*
401 	 * Calculate index in current block of last
402 	 * block to be kept.  -1 indicates the entire
403 	 * block so we need not calculate the index.
404 	 */
405 	factor = 1;
406 	for (i = SINGLE; i < level; i++)
407 		factor *= NINDIR(fs);
408 	last = lastbn;
409 	if (lastbn > 0)
410 		last /= factor;
411 	nblocks = btodb(fs->e2fs_bsize);
412 	/*
413 	 * Get buffer of block pointers, zero those entries corresponding
414 	 * to blocks to be free'd, and update on disk copy first.  Since
415 	 * double(triple) indirect before single(double) indirect, calls
416 	 * to bmap on these blocks will fail.  However, we already have
417 	 * the on disk address, so we have to set the b_blkno field
418 	 * explicitly instead of letting bread do everything for us.
419 	 */
420 	vp = ITOV(ip);
421 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0);
422 	if (!(bp->b_flags & (B_DONE | B_DELWRI))) {
423 		curproc->p_stats->p_ru.ru_inblock++;	/* pay for read */
424 		bp->b_flags |= B_READ;
425 		if (bp->b_bcount > bp->b_bufsize)
426 			panic("ext2fs_indirtrunc: bad buffer size");
427 		bp->b_blkno = dbn;
428 		VOP_STRATEGY(bp);
429 		error = biowait(bp);
430 	}
431 	if (error) {
432 		brelse(bp);
433 		*countp = 0;
434 		return (error);
435 	}
436 
437 	bap = (ufs_daddr_t *)bp->b_data;
438 	if (lastbn >= 0) {
439 		MALLOC(copy, ufs_daddr_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK);
440 		memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize);
441 		memset((caddr_t)&bap[last + 1], 0,
442 			(u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t));
443 		error = bwrite(bp);
444 		if (error)
445 			allerror = error;
446 		bap = copy;
447 	}
448 
449 	/*
450 	 * Recursively free totally unused blocks.
451 	 */
452 	for (i = NINDIR(fs) - 1,
453 		nlbn = lbn + 1 - i * factor; i > last;
454 		i--, nlbn += factor) {
455 		nb = fs2h32(bap[i]);
456 		if (nb == 0)
457 			continue;
458 		if (level > SINGLE) {
459 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
460 						   (ufs_daddr_t)-1, level - 1,
461 						   &blkcount);
462 			if (error)
463 				allerror = error;
464 			blocksreleased += blkcount;
465 		}
466 		ext2fs_blkfree(ip, nb);
467 		blocksreleased += nblocks;
468 	}
469 
470 	/*
471 	 * Recursively free last partial block.
472 	 */
473 	if (level > SINGLE && lastbn >= 0) {
474 		last = lastbn % factor;
475 		nb = fs2h32(bap[i]);
476 		if (nb != 0) {
477 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
478 						   last, level - 1, &blkcount);
479 			if (error)
480 				allerror = error;
481 			blocksreleased += blkcount;
482 		}
483 	}
484 
485 	if (copy != NULL) {
486 		FREE(copy, M_TEMP);
487 	} else {
488 		bp->b_flags |= B_INVAL;
489 		brelse(bp);
490 	}
491 
492 	*countp = blocksreleased;
493 	return (allerror);
494 }
495