xref: /freebsd/sys/fs/ext2fs/ext2_inode.c (revision 884d26c84cba3ffc3d4e626306098fcdfe6a0c2b)
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 #include <sys/rwlock.h>
47 
48 #include <vm/vm.h>
49 #include <vm/vm_extern.h>
50 
51 #include <fs/ext2fs/inode.h>
52 #include <fs/ext2fs/ext2_mount.h>
53 #include <fs/ext2fs/ext2fs.h>
54 #include <fs/ext2fs/fs.h>
55 #include <fs/ext2fs/ext2_extern.h>
56 
57 static int ext2_indirtrunc(struct inode *, daddr_t, daddr_t,
58 	    daddr_t, int, e4fs_daddr_t *);
59 
60 /*
61  * Update the access, modified, and inode change times as specified by the
62  * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively.  Write the inode
63  * to disk if the IN_MODIFIED flag is set (it may be set initially, or by
64  * the timestamp update).  The IN_LAZYMOD flag is set to force a write
65  * later if not now.  If we write now, then clear both IN_MODIFIED and
66  * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is
67  * set, then wait for the write to complete.
68  */
69 int
70 ext2_update(struct vnode *vp, int waitfor)
71 {
72 	struct m_ext2fs *fs;
73 	struct buf *bp;
74 	struct inode *ip;
75 	int error;
76 
77 	ASSERT_VOP_ELOCKED(vp, "ext2_update");
78 	ext2_itimes(vp);
79 	ip = VTOI(vp);
80 	if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
81 		return (0);
82 	ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED);
83 	fs = ip->i_e2fs;
84 	if(fs->e2fs_ronly)
85 		return (0);
86 	if ((error = bread(ip->i_devvp,
87 	    fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
88 		(int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
89 		brelse(bp);
90 		return (error);
91 	}
92 	ext2_i2ei(ip, (struct ext2fs_dinode *)((char *)bp->b_data +
93 	    EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)));
94 	if (waitfor && !DOINGASYNC(vp))
95 		return (bwrite(bp));
96 	else {
97 		bdwrite(bp);
98 		return (0);
99 	}
100 }
101 
102 #define	SINGLE	0	/* index of single indirect block */
103 #define	DOUBLE	1	/* index of double indirect block */
104 #define	TRIPLE	2	/* index of triple indirect block */
105 /*
106  * Truncate the inode oip to at most length size, freeing the
107  * disk blocks.
108  */
109 int
110 ext2_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred,
111     struct thread *td)
112 {
113 	struct vnode *ovp = vp;
114 	int32_t lastblock;
115 	struct inode *oip;
116 	int32_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
117 	uint32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
118 	struct m_ext2fs *fs;
119 	struct buf *bp;
120 	int offset, size, level;
121 	e4fs_daddr_t count, nblocks, blocksreleased = 0;
122 	int error, i, allerror;
123 	off_t osize;
124 #ifdef INVARIANTS
125 	struct bufobj *bo;
126 #endif
127 
128 	oip = VTOI(ovp);
129 #ifdef INVARIANTS
130 	bo = &ovp->v_bufobj;
131 #endif
132 
133 	ASSERT_VOP_LOCKED(vp, "ext2_truncate");
134 
135 	if (length < 0)
136 	    return (EINVAL);
137 
138 	if (ovp->v_type == VLNK &&
139 	    oip->i_size < ovp->v_mount->mnt_maxsymlinklen) {
140 #ifdef INVARIANTS
141 		if (length != 0)
142 			panic("ext2_truncate: partial truncate of symlink");
143 #endif
144 		bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
145 		oip->i_size = 0;
146 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
147 		return (ext2_update(ovp, 1));
148 	}
149 	if (oip->i_size == length) {
150 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
151 		return (ext2_update(ovp, 0));
152 	}
153 	fs = oip->i_e2fs;
154 	osize = oip->i_size;
155 	/*
156 	 * Lengthen the size of the file. We must ensure that the
157 	 * last byte of the file is allocated. Since the smallest
158 	 * value of osize is 0, length will be at least 1.
159 	 */
160 	if (osize < length) {
161 		if (length > oip->i_e2fs->e2fs_maxfilesize)
162 			return (EFBIG);
163 		vnode_pager_setsize(ovp, length);
164 		offset = blkoff(fs, length - 1);
165 		lbn = lblkno(fs, length - 1);
166 		flags |= BA_CLRBUF;
167 		error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags);
168 		if (error) {
169 			vnode_pager_setsize(vp, osize);
170 			return (error);
171 		}
172 		oip->i_size = length;
173 		if (bp->b_bufsize == fs->e2fs_bsize)
174 			bp->b_flags |= B_CLUSTEROK;
175 		if (flags & IO_SYNC)
176 			bwrite(bp);
177 		else if (DOINGASYNC(ovp))
178 			bdwrite(bp);
179 		else
180 			bawrite(bp);
181 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
182 		return (ext2_update(ovp, !DOINGASYNC(ovp)));
183 	}
184 	/*
185 	 * Shorten the size of the file. If the file is not being
186 	 * truncated to a block boundary, the contents of the
187 	 * partial block following the end of the file must be
188 	 * zero'ed in case it ever become accessible again because
189 	 * of subsequent file growth.
190 	 */
191 	/* I don't understand the comment above */
192 	offset = blkoff(fs, length);
193 	if (offset == 0) {
194 		oip->i_size = length;
195 	} else {
196 		lbn = lblkno(fs, length);
197 		flags |= BA_CLRBUF;
198 		error = ext2_balloc(oip, lbn, offset, cred, &bp, flags);
199 		if (error)
200 			return (error);
201 		oip->i_size = length;
202 		size = blksize(fs, oip, lbn);
203 		bzero((char *)bp->b_data + offset, (u_int)(size - offset));
204 		allocbuf(bp, size);
205 		if (bp->b_bufsize == fs->e2fs_bsize)
206 			bp->b_flags |= B_CLUSTEROK;
207 		if (flags & IO_SYNC)
208 			bwrite(bp);
209 		else if (DOINGASYNC(ovp))
210 			bdwrite(bp);
211 		else
212 			bawrite(bp);
213 	}
214 	/*
215 	 * Calculate index into inode's block list of
216 	 * last direct and indirect blocks (if any)
217 	 * which we want to keep.  Lastblock is -1 when
218 	 * the file is truncated to 0.
219 	 */
220 	lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
221 	lastiblock[SINGLE] = lastblock - NDADDR;
222 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
223 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
224 	nblocks = btodb(fs->e2fs_bsize);
225 	/*
226 	 * Update file and block pointers on disk before we start freeing
227 	 * blocks.  If we crash before free'ing blocks below, the blocks
228 	 * will be returned to the free list.  lastiblock values are also
229 	 * normalized to -1 for calls to ext2_indirtrunc below.
230 	 */
231 	for (level = TRIPLE; level >= SINGLE; level--) {
232 		oldblks[NDADDR + level] = oip->i_ib[level];
233 		if (lastiblock[level] < 0) {
234 			oip->i_ib[level] = 0;
235 			lastiblock[level] = -1;
236 		}
237 	}
238 	for (i = 0; i < NDADDR; i++) {
239 		oldblks[i] = oip->i_db[i];
240 		if (i > lastblock)
241 			oip->i_db[i] = 0;
242 	}
243 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
244 	allerror = ext2_update(ovp, !DOINGASYNC(ovp));
245 
246 	/*
247 	 * Having written the new inode to disk, save its new configuration
248 	 * and put back the old block pointers long enough to process them.
249 	 * Note that we save the new block configuration so we can check it
250 	 * when we are done.
251 	 */
252 	for (i = 0; i < NDADDR; i++) {
253 		newblks[i] = oip->i_db[i];
254 		oip->i_db[i] = oldblks[i];
255 	}
256 	for (i = 0; i < NIADDR; i++) {
257 		newblks[NDADDR + i] = oip->i_ib[i];
258 		oip->i_ib[i] = oldblks[NDADDR + i];
259 	}
260 	oip->i_size = osize;
261 	error = vtruncbuf(ovp, cred, length, (int)fs->e2fs_bsize);
262 	if (error && (allerror == 0))
263 		allerror = error;
264 	vnode_pager_setsize(ovp, length);
265 
266 	/*
267 	 * Indirect blocks first.
268 	 */
269 	indir_lbn[SINGLE] = -NDADDR;
270 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
271 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
272 	for (level = TRIPLE; level >= SINGLE; level--) {
273 		bn = oip->i_ib[level];
274 		if (bn != 0) {
275 			error = ext2_indirtrunc(oip, indir_lbn[level],
276 			    fsbtodb(fs, bn), lastiblock[level], level, &count);
277 			if (error)
278 				allerror = error;
279 			blocksreleased += count;
280 			if (lastiblock[level] < 0) {
281 				oip->i_ib[level] = 0;
282 				ext2_blkfree(oip, bn, fs->e2fs_fsize);
283 				blocksreleased += nblocks;
284 			}
285 		}
286 		if (lastiblock[level] >= 0)
287 			goto done;
288 	}
289 
290 	/*
291 	 * All whole direct blocks or frags.
292 	 */
293 	for (i = NDADDR - 1; i > lastblock; i--) {
294 		long bsize;
295 
296 		bn = oip->i_db[i];
297 		if (bn == 0)
298 			continue;
299 		oip->i_db[i] = 0;
300 		bsize = blksize(fs, oip, i);
301 		ext2_blkfree(oip, bn, bsize);
302 		blocksreleased += btodb(bsize);
303 	}
304 	if (lastblock < 0)
305 		goto done;
306 
307 	/*
308 	 * Finally, look for a change in size of the
309 	 * last direct block; release any frags.
310 	 */
311 	bn = oip->i_db[lastblock];
312 	if (bn != 0) {
313 		long oldspace, newspace;
314 
315 		/*
316 		 * Calculate amount of space we're giving
317 		 * back as old block size minus new block size.
318 		 */
319 		oldspace = blksize(fs, oip, lastblock);
320 		oip->i_size = length;
321 		newspace = blksize(fs, oip, lastblock);
322 		if (newspace == 0)
323 			panic("ext2_truncate: newspace");
324 		if (oldspace - newspace > 0) {
325 			/*
326 			 * Block number of space to be free'd is
327 			 * the old block # plus the number of frags
328 			 * required for the storage we're keeping.
329 			 */
330 			bn += numfrags(fs, newspace);
331 			ext2_blkfree(oip, bn, oldspace - newspace);
332 			blocksreleased += btodb(oldspace - newspace);
333 		}
334 	}
335 done:
336 #ifdef INVARIANTS
337 	for (level = SINGLE; level <= TRIPLE; level++)
338 		if (newblks[NDADDR + level] != oip->i_ib[level])
339 			panic("itrunc1");
340 	for (i = 0; i < NDADDR; i++)
341 		if (newblks[i] != oip->i_db[i])
342 			panic("itrunc2");
343 	BO_LOCK(bo);
344 	if (length == 0 && (bo->bo_dirty.bv_cnt != 0 ||
345 	    bo->bo_clean.bv_cnt != 0))
346 		panic("itrunc3");
347 	BO_UNLOCK(bo);
348 #endif /* INVARIANTS */
349 	/*
350 	 * Put back the real size.
351 	 */
352 	oip->i_size = length;
353 	if (oip->i_blocks >= blocksreleased)
354 		oip->i_blocks -= blocksreleased;
355 	else				/* sanity */
356 		oip->i_blocks = 0;
357 	oip->i_flag |= IN_CHANGE;
358 	vnode_pager_setsize(ovp, length);
359 	return (allerror);
360 }
361 
362 /*
363  * Release blocks associated with the inode ip and stored in the indirect
364  * block bn.  Blocks are free'd in LIFO order up to (but not including)
365  * lastbn.  If level is greater than SINGLE, the block is an indirect block
366  * and recursive calls to indirtrunc must be used to cleanse other indirect
367  * blocks.
368  *
369  * NB: triple indirect blocks are untested.
370  */
371 
372 static int
373 ext2_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn,
374     daddr_t lastbn, int level, e4fs_daddr_t *countp)
375 {
376 	struct buf *bp;
377 	struct m_ext2fs *fs = ip->i_e2fs;
378 	struct vnode *vp;
379 	e2fs_daddr_t *bap, *copy;
380 	int i, nblocks, error = 0, allerror = 0;
381 	e2fs_lbn_t nb, nlbn, last;
382 	e4fs_daddr_t blkcount, factor, blocksreleased = 0;
383 
384 	/*
385 	 * Calculate index in current block of last
386 	 * block to be kept.  -1 indicates the entire
387 	 * block so we need not calculate the index.
388 	 */
389 	factor = 1;
390 	for (i = SINGLE; i < level; i++)
391 		factor *= NINDIR(fs);
392 	last = lastbn;
393 	if (lastbn > 0)
394 		last /= factor;
395 	nblocks = btodb(fs->e2fs_bsize);
396 	/*
397 	 * Get buffer of block pointers, zero those entries corresponding
398 	 * to blocks to be free'd, and update on disk copy first.  Since
399 	 * double(triple) indirect before single(double) indirect, calls
400 	 * to bmap on these blocks will fail.  However, we already have
401 	 * the on disk address, so we have to set the b_blkno field
402 	 * explicitly instead of letting bread do everything for us.
403 	 */
404 	vp = ITOV(ip);
405 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0, 0);
406 	if ((bp->b_flags & (B_DONE | B_DELWRI)) == 0) {
407 		bp->b_iocmd = BIO_READ;
408 		if (bp->b_bcount > bp->b_bufsize)
409 			panic("ext2_indirtrunc: bad buffer size");
410 		bp->b_blkno = dbn;
411 		vfs_busy_pages(bp, 0);
412 		bp->b_iooffset = dbtob(bp->b_blkno);
413 		bstrategy(bp);
414 		error = bufwait(bp);
415 	}
416 	if (error) {
417 		brelse(bp);
418 		*countp = 0;
419 		return (error);
420 	}
421 
422 	bap = (e2fs_daddr_t *)bp->b_data;
423 	copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK);
424 	bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize);
425 	bzero((caddr_t)&bap[last + 1],
426 	  (NINDIR(fs) - (last + 1)) * sizeof(e2fs_daddr_t));
427 	if (last == -1)
428 		bp->b_flags |= B_INVAL;
429 	if (DOINGASYNC(vp)) {
430 		bdwrite(bp);
431 	} else {
432 		error = bwrite(bp);
433 		if (error)
434 			allerror = error;
435 	}
436 	bap = copy;
437 
438 	/*
439 	 * Recursively free totally unused blocks.
440 	 */
441 	for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
442 	    i--, nlbn += factor) {
443 		nb = bap[i];
444 		if (nb == 0)
445 			continue;
446 		if (level > SINGLE) {
447 			if ((error = ext2_indirtrunc(ip, nlbn,
448 			    fsbtodb(fs, nb), (int32_t)-1, level - 1, &blkcount)) != 0)
449 				allerror = error;
450 			blocksreleased += blkcount;
451 		}
452 		ext2_blkfree(ip, nb, fs->e2fs_bsize);
453 		blocksreleased += nblocks;
454 	}
455 
456 	/*
457 	 * Recursively free last partial block.
458 	 */
459 	if (level > SINGLE && lastbn >= 0) {
460 		last = lastbn % factor;
461 		nb = bap[i];
462 		if (nb != 0) {
463 			if ((error = ext2_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
464 			    last, level - 1, &blkcount)) != 0)
465 				allerror = error;
466 			blocksreleased += blkcount;
467 		}
468 	}
469 	free(copy, M_TEMP);
470 	*countp = blocksreleased;
471 	return (allerror);
472 }
473 
474 /*
475  *	discard preallocated blocks
476  */
477 int
478 ext2_inactive(struct vop_inactive_args *ap)
479 {
480 	struct vnode *vp = ap->a_vp;
481 	struct inode *ip = VTOI(vp);
482 	struct thread *td = ap->a_td;
483 	int mode, error = 0;
484 
485 	/*
486 	 * Ignore inodes related to stale file handles.
487 	 */
488 	if (ip->i_mode == 0)
489 		goto out;
490 	if (ip->i_nlink <= 0) {
491 		error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td);
492 		ip->i_rdev = 0;
493 		mode = ip->i_mode;
494 		ip->i_mode = 0;
495 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
496 		ext2_vfree(vp, ip->i_number, mode);
497 	}
498 	if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
499 		ext2_update(vp, 0);
500 out:
501 	/*
502 	 * If we are done with the inode, reclaim it
503 	 * so that it can be reused immediately.
504 	 */
505 	if (ip->i_mode == 0)
506 		vrecycle(vp);
507 	return (error);
508 }
509 
510 /*
511  * Reclaim an inode so that it can be used for other purposes.
512  */
513 int
514 ext2_reclaim(struct vop_reclaim_args *ap)
515 {
516 	struct inode *ip;
517 	struct vnode *vp = ap->a_vp;
518 
519 	ip = VTOI(vp);
520 	if (ip->i_flag & IN_LAZYMOD) {
521 		ip->i_flag |= IN_MODIFIED;
522 		ext2_update(vp, 0);
523 	}
524 	vfs_hash_remove(vp);
525 	free(vp->v_data, M_EXT2NODE);
526 	vp->v_data = 0;
527 	vnode_destroy_vobject(vp);
528 	return (0);
529 }
530