xref: /freebsd/sys/fs/ext2fs/ext2_inode.c (revision 10b9d77bf1ccf2f3affafa6261692cb92cf7e992)
1 /*-
2  *  modified for Lites 1.1
3  *
4  *  Aug 1995, Godmar Back (gback@cs.utah.edu)
5  *  University of Utah, Department of Computer Science
6  */
7 /*-
8  * Copyright (c) 1982, 1986, 1989, 1993
9  *	The Regents of the University of California.  All rights reserved.
10  *
11  * Redistribution and use in source and binary forms, with or without
12  * modification, are permitted provided that the following conditions
13  * are met:
14  * 1. Redistributions of source code must retain the above copyright
15  *    notice, this list of conditions and the following disclaimer.
16  * 2. Redistributions in binary form must reproduce the above copyright
17  *    notice, this list of conditions and the following disclaimer in the
18  *    documentation and/or other materials provided with the distribution.
19  * 4. Neither the name of the University nor the names of its contributors
20  *    may be used to endorse or promote products derived from this software
21  *    without specific prior written permission.
22  *
23  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
24  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
25  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
26  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
27  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
28  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
29  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
30  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
31  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
32  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33  * SUCH DAMAGE.
34  *
35  *	@(#)ffs_inode.c	8.5 (Berkeley) 12/30/93
36  * $FreeBSD$
37  */
38 
39 #include <sys/param.h>
40 #include <sys/systm.h>
41 #include <sys/mount.h>
42 #include <sys/bio.h>
43 #include <sys/buf.h>
44 #include <sys/vnode.h>
45 #include <sys/malloc.h>
46 
47 #include <vm/vm.h>
48 #include <vm/vm_extern.h>
49 
50 #include <fs/ext2fs/inode.h>
51 #include <fs/ext2fs/ext2_mount.h>
52 #include <fs/ext2fs/ext2fs.h>
53 #include <fs/ext2fs/fs.h>
54 #include <fs/ext2fs/ext2_extern.h>
55 
56 static int ext2_indirtrunc(struct inode *, int32_t, int32_t, int32_t, int,
57 	    long *);
58 
59 /*
60  * Update the access, modified, and inode change times as specified by the
61  * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively.  Write the inode
62  * to disk if the IN_MODIFIED flag is set (it may be set initially, or by
63  * the timestamp update).  The IN_LAZYMOD flag is set to force a write
64  * later if not now.  If we write now, then clear both IN_MODIFIED and
65  * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is
66  * set, then wait for the write to complete.
67  */
68 int
69 ext2_update(vp, waitfor)
70 	struct vnode *vp;
71 	int waitfor;
72 {
73 	struct m_ext2fs *fs;
74 	struct buf *bp;
75 	struct inode *ip;
76 	int error;
77 
78 	ASSERT_VOP_ELOCKED(vp, "ext2_update");
79 	ext2_itimes(vp);
80 	ip = VTOI(vp);
81 	if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
82 		return (0);
83 	ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED);
84 	fs = ip->i_e2fs;
85 	if(fs->e2fs_ronly)
86 		return (0);
87 	if ((error = bread(ip->i_devvp,
88 	    fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
89 		(int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
90 		brelse(bp);
91 		return (error);
92 	}
93 	ext2_i2ei(ip, (struct ext2fs_dinode *)((char *)bp->b_data +
94 	    EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)));
95 	if (waitfor && (vp->v_mount->mnt_kern_flag & MNTK_ASYNC) == 0)
96 		return (bwrite(bp));
97 	else {
98 		bdwrite(bp);
99 		return (0);
100 	}
101 }
102 
103 #define	SINGLE	0	/* index of single indirect block */
104 #define	DOUBLE	1	/* index of double indirect block */
105 #define	TRIPLE	2	/* index of triple indirect block */
106 /*
107  * Truncate the inode oip to at most length size, freeing the
108  * disk blocks.
109  */
110 int
111 ext2_truncate(vp, length, flags, cred, td)
112 	struct vnode *vp;
113 	off_t length;
114 	int flags;
115 	struct ucred *cred;
116 	struct thread *td;
117 {
118 	struct vnode *ovp = vp;
119 	int32_t lastblock;
120 	struct inode *oip;
121 	int32_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
122 	int32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
123 	struct bufobj *bo;
124 	struct m_ext2fs *fs;
125 	struct buf *bp;
126 	int offset, size, level;
127 	long count, nblocks, blocksreleased = 0;
128 	int aflags, error, i, allerror;
129 	off_t osize;
130 
131 	oip = VTOI(ovp);
132 	bo = &ovp->v_bufobj;
133 
134 	ASSERT_VOP_LOCKED(vp, "ext2_truncate");
135 
136 	if (length < 0)
137 	    return (EINVAL);
138 
139 	if (ovp->v_type == VLNK &&
140 	    oip->i_size < ovp->v_mount->mnt_maxsymlinklen) {
141 #ifdef DIAGNOSTIC
142 		if (length != 0)
143 			panic("ext2_truncate: partial truncate of symlink");
144 #endif
145 		bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
146 		oip->i_size = 0;
147 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
148 		return (ext2_update(ovp, 1));
149 	}
150 	if (oip->i_size == length) {
151 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
152 		return (ext2_update(ovp, 0));
153 	}
154 	fs = oip->i_e2fs;
155 	osize = oip->i_size;
156 	/*
157 	 * Lengthen the size of the file. We must ensure that the
158 	 * last byte of the file is allocated. Since the smallest
159 	 * value of osize is 0, length will be at least 1.
160 	 */
161 	if (osize < length) {
162 		if (length > oip->i_e2fs->e2fs_maxfilesize)
163 			return (EFBIG);
164 		vnode_pager_setsize(ovp, length);
165 		offset = blkoff(fs, length - 1);
166 		lbn = lblkno(fs, length - 1);
167 		aflags = B_CLRBUF;
168 		if (flags & IO_SYNC)
169 			aflags |= B_SYNC;
170 		error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, aflags);
171 		if (error) {
172 			vnode_pager_setsize(vp, osize);
173 			return (error);
174 		}
175 		oip->i_size = length;
176 		if (bp->b_bufsize == fs->e2fs_bsize)
177 			bp->b_flags |= B_CLUSTEROK;
178 		if (aflags & B_SYNC)
179 			bwrite(bp);
180 		else if (ovp->v_mount->mnt_flag & MNT_ASYNC)
181 			bdwrite(bp);
182 		else
183 			bawrite(bp);
184 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
185 		return (ext2_update(ovp, 1));
186 	}
187 	/*
188 	 * Shorten the size of the file. If the file is not being
189 	 * truncated to a block boundry, the contents of the
190 	 * partial block following the end of the file must be
191 	 * zero'ed in case it ever become accessible again because
192 	 * of subsequent file growth.
193 	 */
194 	/* I don't understand the comment above */
195 	offset = blkoff(fs, length);
196 	if (offset == 0) {
197 		oip->i_size = length;
198 	} else {
199 		lbn = lblkno(fs, length);
200 		aflags = B_CLRBUF;
201 		if (flags & IO_SYNC)
202 			aflags |= B_SYNC;
203 		error = ext2_balloc(oip, lbn, offset, cred, &bp, aflags);
204 		if (error)
205 			return (error);
206 		oip->i_size = length;
207 		size = blksize(fs, oip, lbn);
208 		bzero((char *)bp->b_data + offset, (u_int)(size - offset));
209 		allocbuf(bp, size);
210 		if (bp->b_bufsize == fs->e2fs_bsize)
211 			bp->b_flags |= B_CLUSTEROK;
212 		if (aflags & B_SYNC)
213 			bwrite(bp);
214 		else if (ovp->v_mount->mnt_flag & MNT_ASYNC)
215 			bdwrite(bp);
216 		else
217 			bawrite(bp);
218 	}
219 	/*
220 	 * Calculate index into inode's block list of
221 	 * last direct and indirect blocks (if any)
222 	 * which we want to keep.  Lastblock is -1 when
223 	 * the file is truncated to 0.
224 	 */
225 	lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
226 	lastiblock[SINGLE] = lastblock - NDADDR;
227 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
228 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
229 	nblocks = btodb(fs->e2fs_bsize);
230 	/*
231 	 * Update file and block pointers on disk before we start freeing
232 	 * blocks.  If we crash before free'ing blocks below, the blocks
233 	 * will be returned to the free list.  lastiblock values are also
234 	 * normalized to -1 for calls to ext2_indirtrunc below.
235 	 */
236 	bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks);
237 	for (level = TRIPLE; level >= SINGLE; level--)
238 		if (lastiblock[level] < 0) {
239 			oip->i_ib[level] = 0;
240 			lastiblock[level] = -1;
241 		}
242 	for (i = NDADDR - 1; i > lastblock; i--)
243 		oip->i_db[i] = 0;
244 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
245 	allerror = ext2_update(ovp, 1);
246 
247 	/*
248 	 * Having written the new inode to disk, save its new configuration
249 	 * and put back the old block pointers long enough to process them.
250 	 * Note that we save the new block configuration so we can check it
251 	 * when we are done.
252 	 */
253 	bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks);
254 	bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks);
255 	oip->i_size = osize;
256 	error = vtruncbuf(ovp, cred, td, length, (int)fs->e2fs_bsize);
257 	if (error && (allerror == 0))
258 		allerror = error;
259 	vnode_pager_setsize(ovp, length);
260 
261 	/*
262 	 * Indirect blocks first.
263 	 */
264 	indir_lbn[SINGLE] = -NDADDR;
265 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
266 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
267 	for (level = TRIPLE; level >= SINGLE; level--) {
268 		bn = oip->i_ib[level];
269 		if (bn != 0) {
270 			error = ext2_indirtrunc(oip, indir_lbn[level],
271 			    fsbtodb(fs, bn), lastiblock[level], level, &count);
272 			if (error)
273 				allerror = error;
274 			blocksreleased += count;
275 			if (lastiblock[level] < 0) {
276 				oip->i_ib[level] = 0;
277 				ext2_blkfree(oip, bn, fs->e2fs_fsize);
278 				blocksreleased += nblocks;
279 			}
280 		}
281 		if (lastiblock[level] >= 0)
282 			goto done;
283 	}
284 
285 	/*
286 	 * All whole direct blocks or frags.
287 	 */
288 	for (i = NDADDR - 1; i > lastblock; i--) {
289 		long bsize;
290 
291 		bn = oip->i_db[i];
292 		if (bn == 0)
293 			continue;
294 		oip->i_db[i] = 0;
295 		bsize = blksize(fs, oip, i);
296 		ext2_blkfree(oip, bn, bsize);
297 		blocksreleased += btodb(bsize);
298 	}
299 	if (lastblock < 0)
300 		goto done;
301 
302 	/*
303 	 * Finally, look for a change in size of the
304 	 * last direct block; release any frags.
305 	 */
306 	bn = oip->i_db[lastblock];
307 	if (bn != 0) {
308 		long oldspace, newspace;
309 
310 		/*
311 		 * Calculate amount of space we're giving
312 		 * back as old block size minus new block size.
313 		 */
314 		oldspace = blksize(fs, oip, lastblock);
315 		oip->i_size = length;
316 		newspace = blksize(fs, oip, lastblock);
317 		if (newspace == 0)
318 			panic("itrunc: newspace");
319 		if (oldspace - newspace > 0) {
320 			/*
321 			 * Block number of space to be free'd is
322 			 * the old block # plus the number of frags
323 			 * required for the storage we're keeping.
324 			 */
325 			bn += numfrags(fs, newspace);
326 			ext2_blkfree(oip, bn, oldspace - newspace);
327 			blocksreleased += btodb(oldspace - newspace);
328 		}
329 	}
330 done:
331 #ifdef DIAGNOSTIC
332 	for (level = SINGLE; level <= TRIPLE; level++)
333 		if (newblks[NDADDR + level] != oip->i_ib[level])
334 			panic("itrunc1");
335 	for (i = 0; i < NDADDR; i++)
336 		if (newblks[i] != oip->i_db[i])
337 			panic("itrunc2");
338 	BO_LOCK(bo);
339 	if (length == 0 && (bo->bo_dirty.bv_cnt != 0 ||
340 	    bo->bo_clean.bv_cnt != 0))
341 		panic("itrunc3");
342 	BO_UNLOCK(bo);
343 #endif /* DIAGNOSTIC */
344 	/*
345 	 * Put back the real size.
346 	 */
347 	oip->i_size = length;
348 	oip->i_blocks -= blocksreleased;
349 	if (oip->i_blocks < 0)			/* sanity */
350 		oip->i_blocks = 0;
351 	oip->i_flag |= IN_CHANGE;
352 	vnode_pager_setsize(ovp, length);
353 	return (allerror);
354 }
355 
356 /*
357  * Release blocks associated with the inode ip and stored in the indirect
358  * block bn.  Blocks are free'd in LIFO order up to (but not including)
359  * lastbn.  If level is greater than SINGLE, the block is an indirect block
360  * and recursive calls to indirtrunc must be used to cleanse other indirect
361  * blocks.
362  *
363  * NB: triple indirect blocks are untested.
364  */
365 
366 static int
367 ext2_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
368 	struct inode *ip;
369 	int32_t lbn, lastbn;
370 	int32_t dbn;
371 	int level;
372 	long *countp;
373 {
374 	struct buf *bp;
375 	struct m_ext2fs *fs = ip->i_e2fs;
376 	struct vnode *vp;
377 	int32_t *bap, *copy, nb, nlbn, last;
378 	long blkcount, factor;
379 	int i, nblocks, blocksreleased = 0;
380 	int error = 0, allerror = 0;
381 
382 	/*
383 	 * Calculate index in current block of last
384 	 * block to be kept.  -1 indicates the entire
385 	 * block so we need not calculate the index.
386 	 */
387 	factor = 1;
388 	for (i = SINGLE; i < level; i++)
389 		factor *= NINDIR(fs);
390 	last = lastbn;
391 	if (lastbn > 0)
392 		last /= factor;
393 	nblocks = btodb(fs->e2fs_bsize);
394 	/*
395 	 * Get buffer of block pointers, zero those entries corresponding
396 	 * to blocks to be free'd, and update on disk copy first.  Since
397 	 * double(triple) indirect before single(double) indirect, calls
398 	 * to bmap on these blocks will fail.  However, we already have
399 	 * the on disk address, so we have to set the b_blkno field
400 	 * explicitly instead of letting bread do everything for us.
401 	 */
402 	vp = ITOV(ip);
403 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0, 0);
404 	if (bp->b_flags & (B_DONE | B_DELWRI)) {
405 	} else {
406 		bp->b_iocmd = BIO_READ;
407 		if (bp->b_bcount > bp->b_bufsize)
408 			panic("ext2_indirtrunc: bad buffer size");
409 		bp->b_blkno = dbn;
410 		vfs_busy_pages(bp, 0);
411 		bp->b_iooffset = dbtob(bp->b_blkno);
412 		bstrategy(bp);
413 		error = bufwait(bp);
414 	}
415 	if (error) {
416 		brelse(bp);
417 		*countp = 0;
418 		return (error);
419 	}
420 
421 	bap = (int32_t *)bp->b_data;
422 	copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK);
423 	bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize);
424 	bzero((caddr_t)&bap[last + 1],
425 	  (u_int)(NINDIR(fs) - (last + 1)) * sizeof (int32_t));
426 	if (last == -1)
427 		bp->b_flags |= B_INVAL;
428 	error = bwrite(bp);
429 	if (error)
430 		allerror = error;
431 	bap = copy;
432 
433 	/*
434 	 * Recursively free totally unused blocks.
435 	 */
436 	for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
437 	    i--, nlbn += factor) {
438 		nb = bap[i];
439 		if (nb == 0)
440 			continue;
441 		if (level > SINGLE) {
442 			if ((error = ext2_indirtrunc(ip, nlbn,
443 			    fsbtodb(fs, nb), (int32_t)-1, level - 1, &blkcount)) != 0)
444 				allerror = error;
445 			blocksreleased += blkcount;
446 		}
447 		ext2_blkfree(ip, nb, fs->e2fs_bsize);
448 		blocksreleased += nblocks;
449 	}
450 
451 	/*
452 	 * Recursively free last partial block.
453 	 */
454 	if (level > SINGLE && lastbn >= 0) {
455 		last = lastbn % factor;
456 		nb = bap[i];
457 		if (nb != 0) {
458 			if ((error = ext2_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
459 			    last, level - 1, &blkcount)) != 0)
460 				allerror = error;
461 			blocksreleased += blkcount;
462 		}
463 	}
464 	free(copy, M_TEMP);
465 	*countp = blocksreleased;
466 	return (allerror);
467 }
468 
469 /*
470  *	discard preallocated blocks
471  */
472 int
473 ext2_inactive(ap)
474         struct vop_inactive_args /* {
475 		struct vnode *a_vp;
476 		struct thread *a_td;
477 	} */ *ap;
478 {
479 	struct vnode *vp = ap->a_vp;
480 	struct inode *ip = VTOI(vp);
481 	struct thread *td = ap->a_td;
482 	int mode, error = 0;
483 
484 	/*
485 	 * Ignore inodes related to stale file handles.
486 	 */
487 	if (ip->i_mode == 0)
488 		goto out;
489 	if (ip->i_nlink <= 0) {
490 		error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td);
491 		ip->i_rdev = 0;
492 		mode = ip->i_mode;
493 		ip->i_mode = 0;
494 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
495 		ext2_vfree(vp, ip->i_number, mode);
496 	}
497 	if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
498 		ext2_update(vp, 0);
499 out:
500 	/*
501 	 * If we are done with the inode, reclaim it
502 	 * so that it can be reused immediately.
503 	 */
504 	if (ip->i_mode == 0)
505 		vrecycle(vp, td);
506 	return (error);
507 }
508 
509 /*
510  * Reclaim an inode so that it can be used for other purposes.
511  */
512 int
513 ext2_reclaim(ap)
514 	struct vop_reclaim_args /* {
515 		struct vnode *a_vp;
516 		struct thread *a_td;
517 	} */ *ap;
518 {
519 	struct inode *ip;
520 	struct vnode *vp = ap->a_vp;
521 
522 	ip = VTOI(vp);
523 	if (ip->i_flag & IN_LAZYMOD) {
524 		ip->i_flag |= IN_MODIFIED;
525 		ext2_update(vp, 0);
526 	}
527 	vfs_hash_remove(vp);
528 	free(vp->v_data, M_EXT2NODE);
529 	vp->v_data = 0;
530 	vnode_destroy_vobject(vp);
531 	return (0);
532 }
533