1 /*	$OpenBSD: ext2fs_inode.c,v 1.29 2005/10/06 17:43:14 pedro 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. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)ffs_inode.c	8.8 (Berkeley) 10/19/94
34  * Modified for ext2fs by Manuel Bouyer.
35  */
36 
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mount.h>
40 #include <sys/proc.h>
41 #include <sys/file.h>
42 #include <sys/buf.h>
43 #include <sys/vnode.h>
44 #include <sys/kernel.h>
45 #include <sys/malloc.h>
46 #include <sys/resourcevar.h>
47 
48 #include <uvm/uvm_extern.h>
49 
50 #include <ufs/ufs/quota.h>
51 #include <ufs/ufs/inode.h>
52 #include <ufs/ufs/ufsmount.h>
53 #include <ufs/ufs/ufs_extern.h>
54 
55 #include <ufs/ext2fs/ext2fs.h>
56 #include <ufs/ext2fs/ext2fs_extern.h>
57 
58 static int ext2fs_indirtrunc(struct inode *, ufs1_daddr_t, ufs1_daddr_t,
59 				ufs1_daddr_t, int, long *);
60 
61 /*
62  * Get the size of an inode.
63  */
64 u_int64_t
ext2fs_size(struct inode * ip)65 ext2fs_size(struct inode *ip)
66 {
67         u_int64_t size = ip->i_e2fs_size;
68 
69         if ((ip->i_e2fs_mode & IFMT) == IFREG)
70                 size |= (u_int64_t)ip->i_e2fs_dacl << 32;
71 
72         return (size);
73 }
74 
75 int
ext2fs_setsize(struct inode * ip,u_int64_t size)76 ext2fs_setsize(struct inode *ip, u_int64_t size)
77 {
78         if ((ip->i_e2fs_mode & IFMT) == IFREG ||
79             ip->i_e2fs_mode == 0) {
80                 ip->i_e2fs_dacl = size >> 32;
81                 if (size >= 0x80000000U) {
82                         struct m_ext2fs *fs = ip->i_e2fs;
83 
84                         if (fs->e2fs.e2fs_rev <= E2FS_REV0) {
85                                 /* Linux automagically upgrades to REV1 here! */
86                                 return (EFBIG);
87                         }
88                         if (!(fs->e2fs.e2fs_features_rocompat
89                             & EXT2F_ROCOMPAT_LARGEFILE)) {
90                                 fs->e2fs.e2fs_features_rocompat |=
91                                     EXT2F_ROCOMPAT_LARGEFILE;
92                                 fs->e2fs_fmod = 1;
93                         }
94                 }
95         } else if (size >= 0x80000000U)
96                 return (EFBIG);
97 
98         ip->i_e2fs_size = size;
99 
100         return (0);
101 }
102 
103 
104 /*
105  * Last reference to an inode.  If necessary, write or delete it.
106  */
107 int
ext2fs_inactive(v)108 ext2fs_inactive(v)
109 	void *v;
110 {
111 	struct vop_inactive_args /* {
112 		struct vnode *a_vp;
113 		struct proc *a_p;
114 	} */ *ap = v;
115 	struct vnode *vp = ap->a_vp;
116 	struct inode *ip = VTOI(vp);
117 	struct proc *p = ap->a_p;
118 	struct timespec ts;
119 	int error = 0;
120 #ifdef DIAGNOSTIC
121 	extern int prtactive;
122 
123 	if (prtactive && vp->v_usecount != 0)
124 		vprint("ext2fs_inactive: pushing active", vp);
125 #endif
126 
127 	/* Get rid of inodes related to stale file handles. */
128 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime)
129 		goto out;
130 
131 	error = 0;
132 	if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
133 		if (ext2fs_size(ip) != 0) {
134 			error = ext2fs_truncate(ip, (off_t)0, 0, NOCRED);
135 		}
136 		TIMEVAL_TO_TIMESPEC(&time, &ts);
137 		ip->i_e2fs_dtime = ts.tv_sec;
138 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
139 		ext2fs_inode_free(ip, ip->i_number, ip->i_e2fs_mode);
140 	}
141 	if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE)) {
142 		ext2fs_update(ip, NULL, NULL, 0);
143 	}
144 out:
145 	VOP_UNLOCK(vp, 0, p);
146 	/*
147 	 * If we are done with the inode, reclaim it
148 	 * so that it can be reused immediately.
149 	 */
150 	if (ip->i_e2fs_dtime != 0)
151 		vrecycle(vp, NULL, p);
152 	return (error);
153 }
154 
155 
156 /*
157  * Update the access, modified, and inode change times as specified by the
158  * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
159  * used to specify that the inode needs to be updated but that the times have
160  * already been set. The access and modified times are taken from the second
161  * and third parameters; the inode change time is always taken from the current
162  * time. If waitfor is set, then wait for the disk write of the inode to
163  * complete.
164  */
165 int
ext2fs_update(struct inode * ip,struct timespec * atime,struct timespec * mtime,int waitfor)166 ext2fs_update(struct inode *ip, struct timespec *atime, struct timespec *mtime,
167     int waitfor)
168 {
169 	struct m_ext2fs *fs;
170 	struct buf *bp;
171 	int error;
172 	struct timespec ts;
173 	caddr_t cp;
174 
175 	if (ITOV(ip)->v_mount->mnt_flag & MNT_RDONLY)
176 		return (0);
177 	TIMEVAL_TO_TIMESPEC(&time, &ts);
178 	EXT2FS_ITIMES(ip,
179 	    atime ? atime : &ts,
180 	    mtime ? mtime : &ts);
181 	if ((ip->i_flag & IN_MODIFIED) == 0)
182 		return (0);
183 	ip->i_flag &= ~IN_MODIFIED;
184 	fs = ip->i_e2fs;
185 	error = bread(ip->i_devvp,
186 			  fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
187 			  (int)fs->e2fs_bsize, NOCRED, &bp);
188 	if (error) {
189 		brelse(bp);
190 		return (error);
191 	}
192 	ip->i_flag &= ~(IN_MODIFIED);
193 	cp = (caddr_t)bp->b_data +
194 	    (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE(fs));
195 
196 	/*
197 	 * See note about 16-bit UID/GID limitation in ext2fs_vget(). Now
198 	 * that we are about to write the inode, construct the split UID and
199 	 * GID fields out of the two 32-bit fields we kept in memory.
200 	 */
201 	ip->i_e2fs_uid_low = (u_int16_t)ip->i_e2fs_uid;
202 	ip->i_e2fs_gid_low = (u_int16_t)ip->i_e2fs_gid;
203 	ip->i_e2fs_uid_high = ip->i_e2fs_uid >> 16;
204 	ip->i_e2fs_gid_high = ip->i_e2fs_gid >> 16;
205 
206 	e2fs_isave(&ip->i_e2din, (struct ext2fs_dinode *)cp);
207 	if (waitfor)
208 		return (bwrite(bp));
209 	else {
210 		bdwrite(bp);
211 		return (0);
212 	}
213 }
214 
215 #define	SINGLE	0	/* index of single indirect block */
216 #define	DOUBLE	1	/* index of double indirect block */
217 #define	TRIPLE	2	/* index of triple indirect block */
218 /*
219  * Truncate the inode oip to at most length size, freeing the
220  * disk blocks.
221  */
222 int
ext2fs_truncate(struct inode * oip,off_t length,int flags,struct ucred * cred)223 ext2fs_truncate(struct inode *oip, off_t length, int flags, struct ucred *cred)
224 {
225 	struct vnode *ovp = ITOV(oip);
226 	ufs1_daddr_t lastblock;
227 	ufs1_daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
228 	ufs1_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
229 	struct m_ext2fs *fs;
230 	struct buf *bp;
231 	int offset, size, level;
232 	long count, nblocks, vflags, blocksreleased = 0;
233 	int i;
234 	int aflags, error, allerror;
235 	off_t osize;
236 
237 	if (length < 0)
238 		return (EINVAL);
239 
240 	if (ovp->v_type != VREG &&
241 	    ovp->v_type != VDIR &&
242 	    ovp->v_type != VLNK)
243 		return (0);
244 
245 	if (ovp->v_type == VLNK &&
246 		(ext2fs_size(oip) < 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 		bzero((char *)&oip->i_e2din.e2di_shortlink,
254 			(u_int)ext2fs_size(oip));
255 		(void)ext2fs_setsize(oip, 0);
256 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
257 		return (ext2fs_update(oip, NULL, NULL, 1));
258 	}
259 
260 	if (ext2fs_size(oip) == length) {
261 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
262 		return (ext2fs_update(oip, NULL, NULL, 0));
263 	}
264 	fs = oip->i_e2fs;
265 	osize = ext2fs_size(oip);
266 	/*
267 	 * Lengthen the size of the file. We must ensure that the
268 	 * last byte of the file is allocated. Since the smallest
269 	 * value of osize is 0, length will be at least 1.
270 	 */
271 	if (osize < length) {
272 #if 0 /* XXX */
273 		if (length > fs->fs_maxfilesize)
274 			return (EFBIG);
275 #endif
276 		offset = blkoff(fs, length - 1);
277 		lbn = lblkno(fs, length - 1);
278 		aflags = B_CLRBUF;
279 		if (flags & IO_SYNC)
280 			aflags |= B_SYNC;
281 		error = ext2fs_buf_alloc(oip, lbn, offset + 1, cred, &bp,
282 		    aflags);
283 		if (error)
284 			return (error);
285 		(void)ext2fs_setsize(oip, length);
286 		uvm_vnp_setsize(ovp, length);
287 		uvm_vnp_uncache(ovp);
288 		if (aflags & B_SYNC)
289 			bwrite(bp);
290 		else
291 			bawrite(bp);
292 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
293 		return (ext2fs_update(oip, NULL, NULL, 1));
294 	}
295 	/*
296 	 * Shorten the size of the file. If the file is not being
297 	 * truncated to a block boundry, the contents of the
298 	 * partial block following the end of the file must be
299 	 * zero'ed in case it ever become accessible again because
300 	 * of subsequent file growth.
301 	 */
302 	offset = blkoff(fs, length);
303 	if (offset == 0) {
304 		(void)ext2fs_setsize(oip, length);
305 	} else {
306 		lbn = lblkno(fs, length);
307 		aflags = B_CLRBUF;
308 		if (flags & IO_SYNC)
309 			aflags |= B_SYNC;
310 		error = ext2fs_buf_alloc(oip, lbn, offset, cred, &bp,
311 		    aflags);
312 		if (error)
313 			return (error);
314 		(void)ext2fs_setsize(oip, length);
315 		size = fs->e2fs_bsize;
316 		uvm_vnp_setsize(ovp, length);
317 		uvm_vnp_uncache(ovp);
318 		bzero((char *)bp->b_data + offset, (u_int)(size - offset));
319 		allocbuf(bp, size);
320 		if (aflags & B_SYNC)
321 			bwrite(bp);
322 		else
323 			bawrite(bp);
324 	}
325 	/*
326 	 * Calculate index into inode's block list of
327 	 * last direct and indirect blocks (if any)
328 	 * which we want to keep.  Lastblock is -1 when
329 	 * the file is truncated to 0.
330 	 */
331 	lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
332 	lastiblock[SINGLE] = lastblock - NDADDR;
333 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
334 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
335 	nblocks = btodb(fs->e2fs_bsize);
336 	/*
337 	 * Update file and block pointers on disk before we start freeing
338 	 * blocks.  If we crash before free'ing blocks below, the blocks
339 	 * will be returned to the free list.  lastiblock values are also
340 	 * normalized to -1 for calls to ext2fs_indirtrunc below.
341 	 */
342 	memcpy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
343 	for (level = TRIPLE; level >= SINGLE; level--)
344 		if (lastiblock[level] < 0) {
345 			oip->i_e2fs_blocks[NDADDR + level] = 0;
346 			lastiblock[level] = -1;
347 		}
348 	for (i = NDADDR - 1; i > lastblock; i--)
349 		oip->i_e2fs_blocks[i] = 0;
350 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
351 	if ((error = ext2fs_update(oip, NULL, NULL, 1)) != 0)
352 		allerror = error;
353 	/*
354 	 * Having written the new inode to disk, save its new configuration
355 	 * and put back the old block pointers long enough to process them.
356 	 * Note that we save the new block configuration so we can check it
357 	 * when we are done.
358 	 */
359 	bcopy((caddr_t)&oip->i_e2fs_blocks[0], (caddr_t)newblks, sizeof newblks);
360 	bcopy((caddr_t)oldblks, (caddr_t)&oip->i_e2fs_blocks[0], sizeof oldblks);
361 	(void)ext2fs_setsize(oip, osize);
362 	vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
363 	allerror = vinvalbuf(ovp, vflags, cred, curproc, 0, 0);
364 
365 	/*
366 	 * Indirect blocks first.
367 	 */
368 	indir_lbn[SINGLE] = -NDADDR;
369 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) -1;
370 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
371 	for (level = TRIPLE; level >= SINGLE; level--) {
372 		bn = fs2h32(oip->i_e2fs_blocks[NDADDR + level]);
373 		if (bn != 0) {
374 			error = ext2fs_indirtrunc(oip, indir_lbn[level],
375 			    fsbtodb(fs, bn), lastiblock[level], level, &count);
376 			if (error)
377 				allerror = error;
378 			blocksreleased += count;
379 			if (lastiblock[level] < 0) {
380 				oip->i_e2fs_blocks[NDADDR + level] = 0;
381 				ext2fs_blkfree(oip, bn);
382 				blocksreleased += nblocks;
383 			}
384 		}
385 		if (lastiblock[level] >= 0)
386 			goto done;
387 	}
388 
389 	/*
390 	 * All whole direct blocks or frags.
391 	 */
392 	for (i = NDADDR - 1; i > lastblock; i--) {
393 		bn = fs2h32(oip->i_e2fs_blocks[i]);
394 		if (bn == 0)
395 			continue;
396 		oip->i_e2fs_blocks[i] = 0;
397 		ext2fs_blkfree(oip, bn);
398 		blocksreleased += btodb(fs->e2fs_bsize);
399 	}
400 
401 done:
402 #ifdef DIAGNOSTIC
403 	for (level = SINGLE; level <= TRIPLE; level++)
404 		if (newblks[NDADDR + level] !=
405 		    oip->i_e2fs_blocks[NDADDR + level])
406 			panic("ext2fs_truncate1");
407 	for (i = 0; i < NDADDR; i++)
408 		if (newblks[i] != oip->i_e2fs_blocks[i])
409 			panic("ext2fs_truncate2");
410 	if (length == 0 &&
411 	    (!LIST_EMPTY(&ovp->v_cleanblkhd) ||
412 	     !LIST_EMPTY(&ovp->v_dirtyblkhd)))
413 		panic("ext2fs_truncate3");
414 #endif /* DIAGNOSTIC */
415 	/*
416 	 * Put back the real size.
417 	 */
418 	(void)ext2fs_setsize(oip, length);
419 	if (blocksreleased >= oip->i_e2fs_nblock)
420 		oip->i_e2fs_nblock = 0;
421 	else
422 		oip->i_e2fs_nblock -= blocksreleased;
423 	oip->i_flag |= IN_CHANGE;
424 	return (allerror);
425 }
426 
427 /*
428  * Release blocks associated with the inode ip and stored in the indirect
429  * block bn.  Blocks are free'd in LIFO order up to (but not including)
430  * lastbn.  If level is greater than SINGLE, the block is an indirect block
431  * and recursive calls to indirtrunc must be used to cleanse other indirect
432  * blocks.
433  *
434  * NB: triple indirect blocks are untested.
435  */
436 static int
ext2fs_indirtrunc(ip,lbn,dbn,lastbn,level,countp)437 ext2fs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
438 	struct inode *ip;
439 	ufs1_daddr_t lbn, lastbn;
440 	ufs1_daddr_t dbn;
441 	int level;
442 	long *countp;
443 {
444 	int i;
445 	struct buf *bp;
446 	struct m_ext2fs *fs = ip->i_e2fs;
447 	ufs1_daddr_t *bap;
448 	struct vnode *vp;
449 	ufs1_daddr_t *copy = NULL, nb, nlbn, last;
450 	long blkcount, factor;
451 	int nblocks, blocksreleased = 0;
452 	int error = 0, allerror = 0;
453 
454 	/*
455 	 * Calculate index in current block of last
456 	 * block to be kept.  -1 indicates the entire
457 	 * block so we need not calculate the index.
458 	 */
459 	factor = 1;
460 	for (i = SINGLE; i < level; i++)
461 		factor *= NINDIR(fs);
462 	last = lastbn;
463 	if (lastbn > 0)
464 		last /= factor;
465 	nblocks = btodb(fs->e2fs_bsize);
466 	/*
467 	 * Get buffer of block pointers, zero those entries corresponding
468 	 * to blocks to be free'd, and update on disk copy first.  Since
469 	 * double(triple) indirect before single(double) indirect, calls
470 	 * to bmap on these blocks will fail.  However, we already have
471 	 * the on disk address, so we have to set the b_blkno field
472 	 * explicitly instead of letting bread do everything for us.
473 	 */
474 	vp = ITOV(ip);
475 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0);
476 	if (!(bp->b_flags & (B_DONE | B_DELWRI))) {
477 		curproc->p_stats->p_ru.ru_inblock++;	/* pay for read */
478 		bp->b_flags |= B_READ;
479 		if (bp->b_bcount > bp->b_bufsize)
480 			panic("ext2fs_indirtrunc: bad buffer size");
481 		bp->b_blkno = dbn;
482 		VOP_STRATEGY(bp);
483 		error = biowait(bp);
484 	}
485 	if (error) {
486 		brelse(bp);
487 		*countp = 0;
488 		return (error);
489 	}
490 
491 	bap = (ufs1_daddr_t *)bp->b_data;
492 	if (lastbn >= 0) {
493 		MALLOC(copy, ufs1_daddr_t *, fs->e2fs_bsize, M_TEMP, M_WAITOK);
494 		memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->e2fs_bsize);
495 		memset((caddr_t)&bap[last + 1], 0,
496 			(u_int)(NINDIR(fs) - (last + 1)) * sizeof (u_int32_t));
497 		error = bwrite(bp);
498 		if (error)
499 			allerror = error;
500 		bap = copy;
501 	}
502 
503 	/*
504 	 * Recursively free totally unused blocks.
505 	 */
506 	for (i = NINDIR(fs) - 1,
507 		nlbn = lbn + 1 - i * factor; i > last;
508 		i--, nlbn += factor) {
509 		nb = fs2h32(bap[i]);
510 		if (nb == 0)
511 			continue;
512 		if (level > SINGLE) {
513 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
514 						   (ufs1_daddr_t)-1, level - 1,
515 						   &blkcount);
516 			if (error)
517 				allerror = error;
518 			blocksreleased += blkcount;
519 		}
520 		ext2fs_blkfree(ip, nb);
521 		blocksreleased += nblocks;
522 	}
523 
524 	/*
525 	 * Recursively free last partial block.
526 	 */
527 	if (level > SINGLE && lastbn >= 0) {
528 		last = lastbn % factor;
529 		nb = fs2h32(bap[i]);
530 		if (nb != 0) {
531 			error = ext2fs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
532 						   last, level - 1, &blkcount);
533 			if (error)
534 				allerror = error;
535 			blocksreleased += blkcount;
536 		}
537 	}
538 
539 	if (copy != NULL) {
540 		FREE(copy, M_TEMP);
541 	} else {
542 		bp->b_flags |= B_INVAL;
543 		brelse(bp);
544 	}
545 
546 	*countp = blocksreleased;
547 	return (allerror);
548 }
549