xref: /freebsd/sys/fs/ext2fs/ext2_inode.c (revision bc96366c864c07ef352edb92017357917c75b36c)
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 bufobj *bo;
119 	struct m_ext2fs *fs;
120 	struct buf *bp;
121 	int offset, size, level;
122 	e4fs_daddr_t count, nblocks, blocksreleased = 0;
123 	int error, i, allerror;
124 	off_t osize;
125 
126 	oip = VTOI(ovp);
127 	bo = &ovp->v_bufobj;
128 
129 	ASSERT_VOP_LOCKED(vp, "ext2_truncate");
130 
131 	if (length < 0)
132 	    return (EINVAL);
133 
134 	if (ovp->v_type == VLNK &&
135 	    oip->i_size < ovp->v_mount->mnt_maxsymlinklen) {
136 #ifdef INVARIANTS
137 		if (length != 0)
138 			panic("ext2_truncate: partial truncate of symlink");
139 #endif
140 		bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
141 		oip->i_size = 0;
142 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
143 		return (ext2_update(ovp, 1));
144 	}
145 	if (oip->i_size == length) {
146 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
147 		return (ext2_update(ovp, 0));
148 	}
149 	fs = oip->i_e2fs;
150 	osize = oip->i_size;
151 	/*
152 	 * Lengthen the size of the file. We must ensure that the
153 	 * last byte of the file is allocated. Since the smallest
154 	 * value of osize is 0, length will be at least 1.
155 	 */
156 	if (osize < length) {
157 		if (length > oip->i_e2fs->e2fs_maxfilesize)
158 			return (EFBIG);
159 		vnode_pager_setsize(ovp, length);
160 		offset = blkoff(fs, length - 1);
161 		lbn = lblkno(fs, length - 1);
162 		flags |= BA_CLRBUF;
163 		error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags);
164 		if (error) {
165 			vnode_pager_setsize(vp, osize);
166 			return (error);
167 		}
168 		oip->i_size = length;
169 		if (bp->b_bufsize == fs->e2fs_bsize)
170 			bp->b_flags |= B_CLUSTEROK;
171 		if (flags & IO_SYNC)
172 			bwrite(bp);
173 		else if (DOINGASYNC(ovp))
174 			bdwrite(bp);
175 		else
176 			bawrite(bp);
177 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
178 		return (ext2_update(ovp, !DOINGASYNC(ovp)));
179 	}
180 	/*
181 	 * Shorten the size of the file. If the file is not being
182 	 * truncated to a block boundry, the contents of the
183 	 * partial block following the end of the file must be
184 	 * zero'ed in case it ever become accessible again because
185 	 * of subsequent file growth.
186 	 */
187 	/* I don't understand the comment above */
188 	offset = blkoff(fs, length);
189 	if (offset == 0) {
190 		oip->i_size = length;
191 	} else {
192 		lbn = lblkno(fs, length);
193 		flags |= BA_CLRBUF;
194 		error = ext2_balloc(oip, lbn, offset, cred, &bp, flags);
195 		if (error)
196 			return (error);
197 		oip->i_size = length;
198 		size = blksize(fs, oip, lbn);
199 		bzero((char *)bp->b_data + offset, (u_int)(size - offset));
200 		allocbuf(bp, size);
201 		if (bp->b_bufsize == fs->e2fs_bsize)
202 			bp->b_flags |= B_CLUSTEROK;
203 		if (flags & IO_SYNC)
204 			bwrite(bp);
205 		else if (DOINGASYNC(ovp))
206 			bdwrite(bp);
207 		else
208 			bawrite(bp);
209 	}
210 	/*
211 	 * Calculate index into inode's block list of
212 	 * last direct and indirect blocks (if any)
213 	 * which we want to keep.  Lastblock is -1 when
214 	 * the file is truncated to 0.
215 	 */
216 	lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
217 	lastiblock[SINGLE] = lastblock - NDADDR;
218 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
219 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
220 	nblocks = btodb(fs->e2fs_bsize);
221 	/*
222 	 * Update file and block pointers on disk before we start freeing
223 	 * blocks.  If we crash before free'ing blocks below, the blocks
224 	 * will be returned to the free list.  lastiblock values are also
225 	 * normalized to -1 for calls to ext2_indirtrunc below.
226 	 */
227 	for (level = TRIPLE; level >= SINGLE; level--) {
228 		oldblks[NDADDR + level] = oip->i_ib[level];
229 		if (lastiblock[level] < 0) {
230 			oip->i_ib[level] = 0;
231 			lastiblock[level] = -1;
232 		}
233 	}
234 	for (i = 0; i < NDADDR; i++) {
235 		oldblks[i] = oip->i_db[i];
236 		if (i > lastblock)
237 			oip->i_db[i] = 0;
238 	}
239 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
240 	allerror = ext2_update(ovp, !DOINGASYNC(ovp));
241 
242 	/*
243 	 * Having written the new inode to disk, save its new configuration
244 	 * and put back the old block pointers long enough to process them.
245 	 * Note that we save the new block configuration so we can check it
246 	 * when we are done.
247 	 */
248 	for (i = 0; i < NDADDR; i++) {
249 		newblks[i] = oip->i_db[i];
250 		oip->i_db[i] = oldblks[i];
251 	}
252 	for (i = 0; i < NIADDR; i++) {
253 		newblks[NDADDR + i] = oip->i_ib[i];
254 		oip->i_ib[i] = oldblks[NDADDR + i];
255 	}
256 	oip->i_size = osize;
257 	error = vtruncbuf(ovp, cred, length, (int)fs->e2fs_bsize);
258 	if (error && (allerror == 0))
259 		allerror = error;
260 	vnode_pager_setsize(ovp, length);
261 
262 	/*
263 	 * Indirect blocks first.
264 	 */
265 	indir_lbn[SINGLE] = -NDADDR;
266 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
267 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
268 	for (level = TRIPLE; level >= SINGLE; level--) {
269 		bn = oip->i_ib[level];
270 		if (bn != 0) {
271 			error = ext2_indirtrunc(oip, indir_lbn[level],
272 			    fsbtodb(fs, bn), lastiblock[level], level, &count);
273 			if (error)
274 				allerror = error;
275 			blocksreleased += count;
276 			if (lastiblock[level] < 0) {
277 				oip->i_ib[level] = 0;
278 				ext2_blkfree(oip, bn, fs->e2fs_fsize);
279 				blocksreleased += nblocks;
280 			}
281 		}
282 		if (lastiblock[level] >= 0)
283 			goto done;
284 	}
285 
286 	/*
287 	 * All whole direct blocks or frags.
288 	 */
289 	for (i = NDADDR - 1; i > lastblock; i--) {
290 		long bsize;
291 
292 		bn = oip->i_db[i];
293 		if (bn == 0)
294 			continue;
295 		oip->i_db[i] = 0;
296 		bsize = blksize(fs, oip, i);
297 		ext2_blkfree(oip, bn, bsize);
298 		blocksreleased += btodb(bsize);
299 	}
300 	if (lastblock < 0)
301 		goto done;
302 
303 	/*
304 	 * Finally, look for a change in size of the
305 	 * last direct block; release any frags.
306 	 */
307 	bn = oip->i_db[lastblock];
308 	if (bn != 0) {
309 		long oldspace, newspace;
310 
311 		/*
312 		 * Calculate amount of space we're giving
313 		 * back as old block size minus new block size.
314 		 */
315 		oldspace = blksize(fs, oip, lastblock);
316 		oip->i_size = length;
317 		newspace = blksize(fs, oip, lastblock);
318 		if (newspace == 0)
319 			panic("ext2_truncate: newspace");
320 		if (oldspace - newspace > 0) {
321 			/*
322 			 * Block number of space to be free'd is
323 			 * the old block # plus the number of frags
324 			 * required for the storage we're keeping.
325 			 */
326 			bn += numfrags(fs, newspace);
327 			ext2_blkfree(oip, bn, oldspace - newspace);
328 			blocksreleased += btodb(oldspace - newspace);
329 		}
330 	}
331 done:
332 #ifdef INVARIANTS
333 	for (level = SINGLE; level <= TRIPLE; level++)
334 		if (newblks[NDADDR + level] != oip->i_ib[level])
335 			panic("itrunc1");
336 	for (i = 0; i < NDADDR; i++)
337 		if (newblks[i] != oip->i_db[i])
338 			panic("itrunc2");
339 	BO_LOCK(bo);
340 	if (length == 0 && (bo->bo_dirty.bv_cnt != 0 ||
341 	    bo->bo_clean.bv_cnt != 0))
342 		panic("itrunc3");
343 	BO_UNLOCK(bo);
344 #endif /* INVARIANTS */
345 	/*
346 	 * Put back the real size.
347 	 */
348 	oip->i_size = length;
349 	if (oip->i_blocks >= blocksreleased)
350 		oip->i_blocks -= blocksreleased;
351 	else				/* sanity */
352 		oip->i_blocks = 0;
353 	oip->i_flag |= IN_CHANGE;
354 	vnode_pager_setsize(ovp, length);
355 	return (allerror);
356 }
357 
358 /*
359  * Release blocks associated with the inode ip and stored in the indirect
360  * block bn.  Blocks are free'd in LIFO order up to (but not including)
361  * lastbn.  If level is greater than SINGLE, the block is an indirect block
362  * and recursive calls to indirtrunc must be used to cleanse other indirect
363  * blocks.
364  *
365  * NB: triple indirect blocks are untested.
366  */
367 
368 static int
369 ext2_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn,
370     daddr_t lastbn, int level, e4fs_daddr_t *countp)
371 {
372 	struct buf *bp;
373 	struct m_ext2fs *fs = ip->i_e2fs;
374 	struct vnode *vp;
375 	e2fs_daddr_t *bap, *copy;
376 	int i, nblocks, error = 0, allerror = 0;
377 	e2fs_lbn_t nb, nlbn, last;
378 	e4fs_daddr_t blkcount, factor, blocksreleased = 0;
379 
380 	/*
381 	 * Calculate index in current block of last
382 	 * block to be kept.  -1 indicates the entire
383 	 * block so we need not calculate the index.
384 	 */
385 	factor = 1;
386 	for (i = SINGLE; i < level; i++)
387 		factor *= NINDIR(fs);
388 	last = lastbn;
389 	if (lastbn > 0)
390 		last /= factor;
391 	nblocks = btodb(fs->e2fs_bsize);
392 	/*
393 	 * Get buffer of block pointers, zero those entries corresponding
394 	 * to blocks to be free'd, and update on disk copy first.  Since
395 	 * double(triple) indirect before single(double) indirect, calls
396 	 * to bmap on these blocks will fail.  However, we already have
397 	 * the on disk address, so we have to set the b_blkno field
398 	 * explicitly instead of letting bread do everything for us.
399 	 */
400 	vp = ITOV(ip);
401 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0, 0);
402 	if ((bp->b_flags & (B_DONE | B_DELWRI)) == 0) {
403 		bp->b_iocmd = BIO_READ;
404 		if (bp->b_bcount > bp->b_bufsize)
405 			panic("ext2_indirtrunc: bad buffer size");
406 		bp->b_blkno = dbn;
407 		vfs_busy_pages(bp, 0);
408 		bp->b_iooffset = dbtob(bp->b_blkno);
409 		bstrategy(bp);
410 		error = bufwait(bp);
411 	}
412 	if (error) {
413 		brelse(bp);
414 		*countp = 0;
415 		return (error);
416 	}
417 
418 	bap = (e2fs_daddr_t *)bp->b_data;
419 	copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK);
420 	bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize);
421 	bzero((caddr_t)&bap[last + 1],
422 	  (NINDIR(fs) - (last + 1)) * sizeof(e2fs_daddr_t));
423 	if (last == -1)
424 		bp->b_flags |= B_INVAL;
425 	if (DOINGASYNC(vp)) {
426 		bdwrite(bp);
427 	} else {
428 		error = bwrite(bp);
429 		if (error)
430 			allerror = error;
431 	}
432 	bap = copy;
433 
434 	/*
435 	 * Recursively free totally unused blocks.
436 	 */
437 	for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
438 	    i--, nlbn += factor) {
439 		nb = bap[i];
440 		if (nb == 0)
441 			continue;
442 		if (level > SINGLE) {
443 			if ((error = ext2_indirtrunc(ip, nlbn,
444 			    fsbtodb(fs, nb), (int32_t)-1, level - 1, &blkcount)) != 0)
445 				allerror = error;
446 			blocksreleased += blkcount;
447 		}
448 		ext2_blkfree(ip, nb, fs->e2fs_bsize);
449 		blocksreleased += nblocks;
450 	}
451 
452 	/*
453 	 * Recursively free last partial block.
454 	 */
455 	if (level > SINGLE && lastbn >= 0) {
456 		last = lastbn % factor;
457 		nb = bap[i];
458 		if (nb != 0) {
459 			if ((error = ext2_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
460 			    last, level - 1, &blkcount)) != 0)
461 				allerror = error;
462 			blocksreleased += blkcount;
463 		}
464 	}
465 	free(copy, M_TEMP);
466 	*countp = blocksreleased;
467 	return (allerror);
468 }
469 
470 /*
471  *	discard preallocated blocks
472  */
473 int
474 ext2_inactive(struct vop_inactive_args *ap)
475 {
476 	struct vnode *vp = ap->a_vp;
477 	struct inode *ip = VTOI(vp);
478 	struct thread *td = ap->a_td;
479 	int mode, error = 0;
480 
481 	/*
482 	 * Ignore inodes related to stale file handles.
483 	 */
484 	if (ip->i_mode == 0)
485 		goto out;
486 	if (ip->i_nlink <= 0) {
487 		error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td);
488 		ip->i_rdev = 0;
489 		mode = ip->i_mode;
490 		ip->i_mode = 0;
491 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
492 		ext2_vfree(vp, ip->i_number, mode);
493 	}
494 	if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
495 		ext2_update(vp, 0);
496 out:
497 	/*
498 	 * If we are done with the inode, reclaim it
499 	 * so that it can be reused immediately.
500 	 */
501 	if (ip->i_mode == 0)
502 		vrecycle(vp);
503 	return (error);
504 }
505 
506 /*
507  * Reclaim an inode so that it can be used for other purposes.
508  */
509 int
510 ext2_reclaim(struct vop_reclaim_args *ap)
511 {
512 	struct inode *ip;
513 	struct vnode *vp = ap->a_vp;
514 
515 	ip = VTOI(vp);
516 	if (ip->i_flag & IN_LAZYMOD) {
517 		ip->i_flag |= IN_MODIFIED;
518 		ext2_update(vp, 0);
519 	}
520 	vfs_hash_remove(vp);
521 	free(vp->v_data, M_EXT2NODE);
522 	vp->v_data = 0;
523 	vnode_destroy_vobject(vp);
524 	return (0);
525 }
526