xref: /minix/sys/ufs/ext2fs/ext2fs_inode.c (revision 0a6a1f1d)
1 /*	$NetBSD: ext2fs_inode.c,v 1.82 2015/03/28 19:24:04 maxv 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  *
47  * THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
48  * IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
49  * OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
50  * IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
51  * INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
52  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
53  * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
54  * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
55  * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF
56  * THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
57  *
58  *	@(#)ffs_inode.c	8.8 (Berkeley) 10/19/94
59  * Modified for ext2fs by Manuel Bouyer.
60  */
61 
62 #include <sys/cdefs.h>
63 __KERNEL_RCSID(0, "$NetBSD: ext2fs_inode.c,v 1.82 2015/03/28 19:24:04 maxv Exp $");
64 
65 #include <sys/param.h>
66 #include <sys/systm.h>
67 #include <sys/mount.h>
68 #include <sys/proc.h>
69 #include <sys/file.h>
70 #include <sys/buf.h>
71 #include <sys/vnode.h>
72 #include <sys/kernel.h>
73 #include <sys/kmem.h>
74 #include <sys/trace.h>
75 #include <sys/resourcevar.h>
76 #include <sys/kauth.h>
77 
78 #include <ufs/ufs/inode.h>
79 #include <ufs/ufs/ufsmount.h>
80 #include <ufs/ufs/ufs_extern.h>
81 
82 #include <ufs/ext2fs/ext2fs.h>
83 #include <ufs/ext2fs/ext2fs_extern.h>
84 
85 extern int prtactive;
86 
87 static int ext2fs_indirtrunc(struct inode *, daddr_t, daddr_t,
88 				  daddr_t, int, long *);
89 
90 /*
91  * These are fortunately the same values; it is likely that there is
92  * code that assumes they're equal. In any event, neither ought to
93  * ever change because it's a property of the on-disk formats.
94  */
95 CTASSERT(EXT2FS_NDADDR == UFS_NDADDR);
96 CTASSERT(EXT2FS_NIADDR == UFS_NIADDR);
97 
98 /*
99  * Get the size of an inode.
100  */
101 uint64_t
ext2fs_size(struct inode * ip)102 ext2fs_size(struct inode *ip)
103 {
104 	uint64_t size = ip->i_e2fs_size;
105 
106 	if ((ip->i_e2fs_mode & IFMT) == IFREG)
107 		size |= (uint64_t)ip->i_e2fs_dacl << 32;
108 	return size;
109 }
110 
111 int
ext2fs_setsize(struct inode * ip,uint64_t size)112 ext2fs_setsize(struct inode *ip, uint64_t size)
113 {
114 	if ((ip->i_e2fs_mode & IFMT) == IFREG ||
115 	    ip->i_e2fs_mode == 0) {
116 		ip->i_e2fs_dacl = size >> 32;
117 		if (size >= 0x80000000U) {
118 			struct m_ext2fs *fs = ip->i_e2fs;
119 
120 			if (fs->e2fs.e2fs_rev <= E2FS_REV0) {
121 				/* Linux automagically upgrades to REV1 here! */
122 				return EFBIG;
123 			}
124 			if (!(fs->e2fs.e2fs_features_rocompat
125 			    & EXT2F_ROCOMPAT_LARGEFILE)) {
126 				fs->e2fs.e2fs_features_rocompat |=
127 				    EXT2F_ROCOMPAT_LARGEFILE;
128 				fs->e2fs_fmod = 1;
129 			}
130 		}
131 	} else if (size >= 0x80000000U)
132 		return EFBIG;
133 
134 	ip->i_e2fs_size = size;
135 
136 	return 0;
137 }
138 
139 uint64_t
ext2fs_nblock(struct inode * ip)140 ext2fs_nblock(struct inode *ip)
141 {
142 	uint64_t nblock = ip->i_e2fs_nblock;
143 	struct m_ext2fs * const fs = ip->i_e2fs;
144 
145 	if (fs->e2fs.e2fs_features_rocompat & EXT2F_ROCOMPAT_HUGE_FILE) {
146 		nblock |= (uint64_t)ip->i_e2fs_nblock_high << 32;
147 
148 		if ((ip->i_e2fs_flags & EXT2_HUGE_FILE)) {
149 			nblock = EXT2_FSBTODB(fs, nblock);
150 		}
151 	}
152 
153 	return nblock;
154 }
155 
156 int
ext2fs_setnblock(struct inode * ip,uint64_t nblock)157 ext2fs_setnblock(struct inode *ip, uint64_t nblock)
158 {
159 	struct m_ext2fs * const fs = ip->i_e2fs;
160 
161 	if (nblock <= 0xffffffffULL) {
162 		CLR(ip->i_e2fs_flags, EXT2_HUGE_FILE);
163 		ip->i_e2fs_nblock = nblock;
164 		return 0;
165 	}
166 
167 	if (!ISSET(fs->e2fs.e2fs_features_rocompat, EXT2F_ROCOMPAT_HUGE_FILE))
168 		return EFBIG;
169 
170 	if (nblock <= 0xffffffffffffULL) {
171 		CLR(ip->i_e2fs_flags, EXT2_HUGE_FILE);
172 		ip->i_e2fs_nblock = nblock & 0xffffffff;
173 		ip->i_e2fs_nblock_high = (nblock >> 32) & 0xffff;
174 		return 0;
175 	}
176 
177 	if (EXT2_DBTOFSB(fs, nblock) <= 0xffffffffffffULL) {
178 		SET(ip->i_e2fs_flags, EXT2_HUGE_FILE);
179 		ip->i_e2fs_nblock = EXT2_DBTOFSB(fs, nblock) & 0xffffffff;
180 		ip->i_e2fs_nblock_high = (EXT2_DBTOFSB(fs, nblock) >> 32) & 0xffff;
181 		return 0;
182 	}
183 
184 	return EFBIG;
185 }
186 
187 /*
188  * Last reference to an inode.  If necessary, write or delete it.
189  */
190 int
ext2fs_inactive(void * v)191 ext2fs_inactive(void *v)
192 {
193 	struct vop_inactive_args /* {
194 		struct vnode *a_vp;
195 		bool *a_recycle;
196 	} */ *ap = v;
197 	struct vnode *vp = ap->a_vp;
198 	struct inode *ip = VTOI(vp);
199 	int error = 0;
200 
201 	if (prtactive && vp->v_usecount != 0)
202 		vprint("ext2fs_inactive: pushing active", vp);
203 	/* Get rid of inodes related to stale file handles. */
204 	if (ip->i_e2fs_mode == 0 || ip->i_e2fs_dtime != 0)
205 		goto out;
206 
207 	error = 0;
208 	if (ip->i_e2fs_nlink == 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
209 		/* Defer final inode free and update to reclaim.*/
210 		if (ext2fs_size(ip) != 0) {
211 			error = ext2fs_truncate(vp, (off_t)0, 0, NOCRED);
212 		}
213 		ip->i_e2fs_dtime = time_second;
214 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
215 		ip->i_omode = 1;
216 	}
217 	if (ip->i_flag & (IN_CHANGE | IN_UPDATE | IN_MODIFIED)) {
218 		ext2fs_update(vp, NULL, NULL, 0);
219 	}
220 out:
221 	/*
222 	 * If we are done with the inode, reclaim it
223 	 * so that it can be reused immediately.
224 	 */
225 	*ap->a_recycle = (ip->i_e2fs_dtime != 0);
226 	VOP_UNLOCK(vp);
227 	return (error);
228 }
229 
230 
231 /*
232  * Update the access, modified, and inode change times as specified by the
233  * IACCESS, IUPDATE, and ICHANGE flags respectively. The IMODIFIED flag is
234  * used to specify that the inode needs to be updated but that the times have
235  * already been set. The access and modified times are taken from the second
236  * and third parameters; the inode change time is always taken from the current
237  * time. If UPDATE_WAIT or UPDATE_DIROP is set, then wait for the disk
238  * write of the inode to complete.
239  */
240 int
ext2fs_update(struct vnode * vp,const struct timespec * acc,const struct timespec * mod,int updflags)241 ext2fs_update(struct vnode *vp, const struct timespec *acc,
242     const struct timespec *mod, int updflags)
243 {
244 	struct m_ext2fs *fs;
245 	struct buf *bp;
246 	struct inode *ip;
247 	int error;
248 	void *cp;
249 	int flags;
250 
251 	if (vp->v_mount->mnt_flag & MNT_RDONLY)
252 		return (0);
253 	ip = VTOI(vp);
254 	EXT2FS_ITIMES(ip, acc, mod, NULL);
255 	if (updflags & UPDATE_CLOSE)
256 		flags = ip->i_flag & (IN_MODIFIED | IN_ACCESSED);
257 	else
258 		flags = ip->i_flag & IN_MODIFIED;
259 	if (flags == 0)
260 		return (0);
261 	fs = ip->i_e2fs;
262 
263 	error = bread(ip->i_devvp,
264 			  EXT2_FSBTODB(fs, ino_to_fsba(fs, ip->i_number)),
265 			  (int)fs->e2fs_bsize, B_MODIFY, &bp);
266 	if (error) {
267 		return (error);
268 	}
269 	ip->i_flag &= ~(IN_MODIFIED | IN_ACCESSED);
270 	cp = (char *)bp->b_data +
271 	    (ino_to_fsbo(fs, ip->i_number) * EXT2_DINODE_SIZE(fs));
272 	e2fs_isave(ip->i_din.e2fs_din, (struct ext2fs_dinode *)cp);
273 	if ((updflags & (UPDATE_WAIT|UPDATE_DIROP)) != 0 &&
274 	    (flags & IN_MODIFIED) != 0 &&
275 	    (vp->v_mount->mnt_flag & MNT_ASYNC) == 0)
276 		return (bwrite(bp));
277 	else {
278 		bdwrite(bp);
279 		return (0);
280 	}
281 }
282 
283 #define	SINGLE	0	/* index of single indirect block */
284 #define	DOUBLE	1	/* index of double indirect block */
285 #define	TRIPLE	2	/* index of triple indirect block */
286 /*
287  * Truncate the inode oip to at most length size, freeing the
288  * disk blocks.
289  */
290 int
ext2fs_truncate(struct vnode * ovp,off_t length,int ioflag,kauth_cred_t cred)291 ext2fs_truncate(struct vnode *ovp, off_t length, int ioflag,
292     kauth_cred_t cred)
293 {
294 	daddr_t lastblock;
295 	struct inode *oip = VTOI(ovp);
296 	daddr_t bn, lastiblock[EXT2FS_NIADDR], indir_lbn[EXT2FS_NIADDR];
297 	/* XXX ondisk32 */
298 	int32_t oldblks[EXT2FS_NDADDR + EXT2FS_NIADDR], newblks[EXT2FS_NDADDR + EXT2FS_NIADDR];
299 	struct m_ext2fs *fs;
300 	int offset, size, level;
301 	long count, blocksreleased = 0;
302 	int i, nblocks;
303 	int error, allerror = 0;
304 	off_t osize;
305 	int sync;
306 	struct ufsmount *ump = oip->i_ump;
307 
308 	if (ovp->v_type == VCHR || ovp->v_type == VBLK ||
309 	    ovp->v_type == VFIFO || ovp->v_type == VSOCK) {
310 		return 0;
311 	}
312 
313 	if (length < 0)
314 		return (EINVAL);
315 
316 	if (ovp->v_type == VLNK &&
317 	    (ext2fs_size(oip) < ump->um_maxsymlinklen ||
318 	     (ump->um_maxsymlinklen == 0 && ext2fs_nblock(oip) == 0))) {
319 		KDASSERT(length == 0);
320 		memset((char *)&oip->i_din.e2fs_din->e2di_shortlink, 0,
321 			(u_int)ext2fs_size(oip));
322 		(void)ext2fs_setsize(oip, 0);
323 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
324 		return (ext2fs_update(ovp, NULL, NULL, 0));
325 	}
326 	if (ext2fs_size(oip) == length) {
327 		/* still do a uvm_vnp_setsize() as writesize may be larger */
328 		uvm_vnp_setsize(ovp, length);
329 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
330 		return (ext2fs_update(ovp, NULL, NULL, 0));
331 	}
332 	fs = oip->i_e2fs;
333 	if (length > ump->um_maxfilesize)
334 		return (EFBIG);
335 
336 	osize = ext2fs_size(oip);
337 
338 	/*
339 	 * Lengthen the size of the file. We must ensure that the
340 	 * last byte of the file is allocated. Since the smallest
341 	 * value of osize is 0, length will be at least 1.
342 	 */
343 	if (osize < length) {
344 		uvm_vnp_setwritesize(ovp, length);
345 		error = ufs_balloc_range(ovp, length - 1, 1, cred,
346 		    ioflag & IO_SYNC ? B_SYNC : 0);
347 		if (error) {
348 			(void) ext2fs_truncate(ovp, osize, ioflag & IO_SYNC,
349 			    cred);
350 			return (error);
351 		}
352 		uvm_vnp_setsize(ovp, length);
353 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
354 		KASSERT(error  || ovp->v_size == ext2fs_size(oip));
355 		return (ext2fs_update(ovp, NULL, NULL, 0));
356 	}
357 	/*
358 	 * Shorten the size of the file. If the file is not being
359 	 * truncated to a block boundry, the contents of the
360 	 * partial block following the end of the file must be
361 	 * zero'ed in case it ever become accessible again because
362 	 * of subsequent file growth.
363 	 */
364 	offset = ext2_blkoff(fs, length);
365 	if (offset != 0) {
366 		size = fs->e2fs_bsize;
367 
368 		/* XXXUBC we should handle more than just VREG */
369 		ubc_zerorange(&ovp->v_uobj, length, size - offset,
370 		    UBC_UNMAP_FLAG(ovp));
371 	}
372 	(void)ext2fs_setsize(oip, length);
373 	uvm_vnp_setsize(ovp, length);
374 	/*
375 	 * Calculate index into inode's block list of
376 	 * last direct and indirect blocks (if any)
377 	 * which we want to keep.  Lastblock is -1 when
378 	 * the file is truncated to 0.
379 	 */
380 	lastblock = ext2_lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
381 	lastiblock[SINGLE] = lastblock - EXT2FS_NDADDR;
382 	lastiblock[DOUBLE] = lastiblock[SINGLE] - EXT2_NINDIR(fs);
383 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - EXT2_NINDIR(fs) * EXT2_NINDIR(fs);
384 	nblocks = btodb(fs->e2fs_bsize);
385 	/*
386 	 * Update file and block pointers on disk before we start freeing
387 	 * blocks.  If we crash before free'ing blocks below, the blocks
388 	 * will be returned to the free list.  lastiblock values are also
389 	 * normalized to -1 for calls to ext2fs_indirtrunc below.
390 	 */
391 	memcpy((void *)oldblks, (void *)&oip->i_e2fs_blocks[0], sizeof oldblks);
392 	sync = 0;
393 	for (level = TRIPLE; level >= SINGLE; level--) {
394 		if (lastiblock[level] < 0 && oldblks[EXT2FS_NDADDR + level] != 0) {
395 			sync = 1;
396 			oip->i_e2fs_blocks[EXT2FS_NDADDR + level] = 0;
397 			lastiblock[level] = -1;
398 		}
399 	}
400 	for (i = 0; i < EXT2FS_NDADDR; i++) {
401 		if (i > lastblock && oldblks[i] != 0) {
402 			sync = 1;
403 			oip->i_e2fs_blocks[i] = 0;
404 		}
405 	}
406 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
407 	if (sync) {
408 		error = ext2fs_update(ovp, NULL, NULL, UPDATE_WAIT);
409 		if (error && !allerror)
410 			allerror = error;
411 	}
412 
413 	/*
414 	 * Having written the new inode to disk, save its new configuration
415 	 * and put back the old block pointers long enough to process them.
416 	 * Note that we save the new block configuration so we can check it
417 	 * when we are done.
418 	 */
419 	memcpy((void *)newblks, (void *)&oip->i_e2fs_blocks[0], sizeof newblks);
420 	memcpy((void *)&oip->i_e2fs_blocks[0], (void *)oldblks, sizeof oldblks);
421 
422 	(void)ext2fs_setsize(oip, osize);
423 	error = vtruncbuf(ovp, lastblock + 1, 0, 0);
424 	if (error && !allerror)
425 		allerror = error;
426 
427 	/*
428 	 * Indirect blocks first.
429 	 */
430 	indir_lbn[SINGLE] = -EXT2FS_NDADDR;
431 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - EXT2_NINDIR(fs) -1;
432 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - EXT2_NINDIR(fs) * EXT2_NINDIR(fs) - 1;
433 	for (level = TRIPLE; level >= SINGLE; level--) {
434 		/* XXX ondisk32 */
435 		bn = fs2h32(oip->i_e2fs_blocks[EXT2FS_NDADDR + level]);
436 		if (bn != 0) {
437 			error = ext2fs_indirtrunc(oip, indir_lbn[level],
438 			    EXT2_FSBTODB(fs, bn), lastiblock[level], level, &count);
439 			if (error)
440 				allerror = error;
441 			blocksreleased += count;
442 			if (lastiblock[level] < 0) {
443 				oip->i_e2fs_blocks[EXT2FS_NDADDR + level] = 0;
444 				ext2fs_blkfree(oip, bn);
445 				blocksreleased += nblocks;
446 			}
447 		}
448 		if (lastiblock[level] >= 0)
449 			goto done;
450 	}
451 
452 	/*
453 	 * All whole direct blocks or frags.
454 	 */
455 	for (i = EXT2FS_NDADDR - 1; i > lastblock; i--) {
456 		/* XXX ondisk32 */
457 		bn = fs2h32(oip->i_e2fs_blocks[i]);
458 		if (bn == 0)
459 			continue;
460 		oip->i_e2fs_blocks[i] = 0;
461 		ext2fs_blkfree(oip, bn);
462 		blocksreleased += btodb(fs->e2fs_bsize);
463 	}
464 
465 done:
466 #ifdef DIAGNOSTIC
467 	for (level = SINGLE; level <= TRIPLE; level++)
468 		if (newblks[EXT2FS_NDADDR + level] !=
469 		    oip->i_e2fs_blocks[EXT2FS_NDADDR + level])
470 			panic("ext2fs_truncate1");
471 	for (i = 0; i < EXT2FS_NDADDR; i++)
472 		if (newblks[i] != oip->i_e2fs_blocks[i])
473 			panic("ext2fs_truncate2");
474 	if (length == 0 &&
475 	    (!LIST_EMPTY(&ovp->v_cleanblkhd) ||
476 	     !LIST_EMPTY(&ovp->v_dirtyblkhd)))
477 		panic("ext2fs_truncate3");
478 #endif /* DIAGNOSTIC */
479 	/*
480 	 * Put back the real size.
481 	 */
482 	(void)ext2fs_setsize(oip, length);
483 	error = ext2fs_setnblock(oip, ext2fs_nblock(oip) - blocksreleased);
484 	if (error != 0)
485 		allerror = error;
486 	oip->i_flag |= IN_CHANGE;
487 	KASSERT(ovp->v_type != VREG || ovp->v_size == ext2fs_size(oip));
488 	return (allerror);
489 }
490 
491 /*
492  * Release blocks associated with the inode ip and stored in the indirect
493  * block bn.  Blocks are free'd in LIFO order up to (but not including)
494  * lastbn.  If level is greater than SINGLE, the block is an indirect block
495  * and recursive calls to indirtrunc must be used to cleanse other indirect
496  * blocks.
497  *
498  * NB: triple indirect blocks are untested.
499  */
500 static int
ext2fs_indirtrunc(struct inode * ip,daddr_t lbn,daddr_t dbn,daddr_t lastbn,int level,long * countp)501 ext2fs_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn, daddr_t lastbn,
502 		int level, long *countp)
503 {
504 	int i;
505 	struct buf *bp;
506 	struct m_ext2fs *fs = ip->i_e2fs;
507 	int32_t *bap;	/* XXX ondisk32 */
508 	struct vnode *vp;
509 	daddr_t nb, nlbn, last;
510 	int32_t *copy = NULL;	/* XXX ondisk32 */
511 	long blkcount, factor;
512 	int nblocks, blocksreleased = 0;
513 	int error = 0, allerror = 0;
514 
515 	/*
516 	 * Calculate index in current block of last
517 	 * block to be kept.  -1 indicates the entire
518 	 * block so we need not calculate the index.
519 	 */
520 	factor = 1;
521 	for (i = SINGLE; i < level; i++)
522 		factor *= EXT2_NINDIR(fs);
523 	last = lastbn;
524 	if (lastbn > 0)
525 		last /= factor;
526 	nblocks = btodb(fs->e2fs_bsize);
527 	/*
528 	 * Get buffer of block pointers, zero those entries corresponding
529 	 * to blocks to be free'd, and update on disk copy first.  Since
530 	 * double(triple) indirect before single(double) indirect, calls
531 	 * to bmap on these blocks will fail.  However, we already have
532 	 * the on disk address, so we have to set the b_blkno field
533 	 * explicitly instead of letting bread do everything for us.
534 	 */
535 	vp = ITOV(ip);
536 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0);
537 	if (bp->b_oflags & (BO_DONE | BO_DELWRI)) {
538 		/* Braces must be here in case trace evaluates to nothing. */
539 		trace(TR_BREADHIT, pack(vp, fs->e2fs_bsize), lbn);
540 	} else {
541 		trace(TR_BREADMISS, pack(vp, fs->e2fs_bsize), lbn);
542 		curlwp->l_ru.ru_inblock++;	/* pay for read */
543 		bp->b_flags |= B_READ;
544 		if (bp->b_bcount > bp->b_bufsize)
545 			panic("ext2fs_indirtrunc: bad buffer size");
546 		bp->b_blkno = dbn;
547 		VOP_STRATEGY(vp, bp);
548 		error = biowait(bp);
549 	}
550 	if (error) {
551 		brelse(bp, 0);
552 		*countp = 0;
553 		return (error);
554 	}
555 
556 	bap = (int32_t *)bp->b_data;	/* XXX ondisk32 */
557 	if (lastbn >= 0) {
558 		/* XXX ondisk32 */
559 		copy = kmem_alloc(fs->e2fs_bsize, KM_SLEEP);
560 		memcpy((void *)copy, (void *)bap, (u_int)fs->e2fs_bsize);
561 		memset((void *)&bap[last + 1], 0,
562 			(u_int)(EXT2_NINDIR(fs) - (last + 1)) * sizeof (uint32_t));
563 		error = bwrite(bp);
564 		if (error)
565 			allerror = error;
566 		bap = copy;
567 	}
568 
569 	/*
570 	 * Recursively free totally unused blocks.
571 	 */
572 	for (i = EXT2_NINDIR(fs) - 1,
573 		nlbn = lbn + 1 - i * factor; i > last;
574 		i--, nlbn += factor) {
575 		/* XXX ondisk32 */
576 		nb = fs2h32(bap[i]);
577 		if (nb == 0)
578 			continue;
579 		if (level > SINGLE) {
580 			error = ext2fs_indirtrunc(ip, nlbn, EXT2_FSBTODB(fs, nb),
581 						   (daddr_t)-1, level - 1,
582 						   &blkcount);
583 			if (error)
584 				allerror = error;
585 			blocksreleased += blkcount;
586 		}
587 		ext2fs_blkfree(ip, nb);
588 		blocksreleased += nblocks;
589 	}
590 
591 	/*
592 	 * Recursively free last partial block.
593 	 */
594 	if (level > SINGLE && lastbn >= 0) {
595 		last = lastbn % factor;
596 		/* XXX ondisk32 */
597 		nb = fs2h32(bap[i]);
598 		if (nb != 0) {
599 			error = ext2fs_indirtrunc(ip, nlbn, EXT2_FSBTODB(fs, nb),
600 						   last, level - 1, &blkcount);
601 			if (error)
602 				allerror = error;
603 			blocksreleased += blkcount;
604 		}
605 	}
606 
607 	if (copy != NULL) {
608 		kmem_free(copy, fs->e2fs_bsize);
609 	} else {
610 		brelse(bp, BC_INVAL);
611 	}
612 
613 	*countp = blocksreleased;
614 	return (allerror);
615 }
616