xref: /freebsd/sys/fs/ext2fs/ext2_inode.c (revision 094fc1ed0f2627525c7b0342efcbad5be7a8546a)
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  * 3. 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 #include <fs/ext2fs/ext2_extattr.h>
57 
58 static int ext2_indirtrunc(struct inode *, daddr_t, daddr_t,
59 	    daddr_t, int, e4fs_daddr_t *);
60 
61 /*
62  * Update the access, modified, and inode change times as specified by the
63  * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively.  Write the inode
64  * to disk if the IN_MODIFIED flag is set (it may be set initially, or by
65  * the timestamp update).  The IN_LAZYMOD flag is set to force a write
66  * later if not now.  If we write now, then clear both IN_MODIFIED and
67  * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is
68  * set, then wait for the write to complete.
69  */
70 int
71 ext2_update(struct vnode *vp, 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 	error = ext2_i2ei(ip, (struct ext2fs_dinode *)((char *)bp->b_data +
94 	    EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)));
95 	if (error) {
96 		brelse(bp);
97 		return (error);
98 	}
99 	if (waitfor && !DOINGASYNC(vp))
100 		return (bwrite(bp));
101 	else {
102 		bdwrite(bp);
103 		return (0);
104 	}
105 }
106 
107 #define	SINGLE	0	/* index of single indirect block */
108 #define	DOUBLE	1	/* index of double indirect block */
109 #define	TRIPLE	2	/* index of triple indirect block */
110 /*
111  * Truncate the inode oip to at most length size, freeing the
112  * disk blocks.
113  */
114 int
115 ext2_truncate(struct vnode *vp, off_t length, int flags, 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[EXT2_NIADDR], indir_lbn[EXT2_NIADDR];
122 	uint32_t oldblks[EXT2_NDADDR + EXT2_NIADDR];
123 	uint32_t newblks[EXT2_NDADDR + EXT2_NIADDR];
124 	struct m_ext2fs *fs;
125 	struct buf *bp;
126 	int offset, size, level;
127 	e4fs_daddr_t count, nblocks, blocksreleased = 0;
128 	int error, i, allerror;
129 	off_t osize;
130 #ifdef INVARIANTS
131 	struct bufobj *bo;
132 #endif
133 
134 	oip = VTOI(ovp);
135 #ifdef INVARIANTS
136 	bo = &ovp->v_bufobj;
137 #endif
138 
139 	ASSERT_VOP_LOCKED(vp, "ext2_truncate");
140 
141 	if (length < 0)
142 		return (EINVAL);
143 
144 	if (ovp->v_type == VLNK &&
145 	    oip->i_size < ovp->v_mount->mnt_maxsymlinklen) {
146 #ifdef INVARIANTS
147 		if (length != 0)
148 			panic("ext2_truncate: partial truncate of symlink");
149 #endif
150 		bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
151 		oip->i_size = 0;
152 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
153 		return (ext2_update(ovp, 1));
154 	}
155 	if (oip->i_size == length) {
156 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
157 		return (ext2_update(ovp, 0));
158 	}
159 	fs = oip->i_e2fs;
160 	osize = oip->i_size;
161 	/*
162 	 * Lengthen the size of the file. We must ensure that the
163 	 * last byte of the file is allocated. Since the smallest
164 	 * value of osize is 0, length will be at least 1.
165 	 */
166 	if (osize < length) {
167 		if (length > oip->i_e2fs->e2fs_maxfilesize)
168 			return (EFBIG);
169 		vnode_pager_setsize(ovp, length);
170 		offset = blkoff(fs, length - 1);
171 		lbn = lblkno(fs, length - 1);
172 		flags |= BA_CLRBUF;
173 		error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags);
174 		if (error) {
175 			vnode_pager_setsize(vp, osize);
176 			return (error);
177 		}
178 		oip->i_size = length;
179 		if (bp->b_bufsize == fs->e2fs_bsize)
180 			bp->b_flags |= B_CLUSTEROK;
181 		if (flags & IO_SYNC)
182 			bwrite(bp);
183 		else if (DOINGASYNC(ovp))
184 			bdwrite(bp);
185 		else
186 			bawrite(bp);
187 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
188 		return (ext2_update(ovp, !DOINGASYNC(ovp)));
189 	}
190 	/*
191 	 * Shorten the size of the file. If the file is not being
192 	 * truncated to a block boundary, the contents of the
193 	 * partial block following the end of the file must be
194 	 * zero'ed in case it ever become accessible again because
195 	 * of subsequent file growth.
196 	 */
197 	/* I don't understand the comment above */
198 	offset = blkoff(fs, length);
199 	if (offset == 0) {
200 		oip->i_size = length;
201 	} else {
202 		lbn = lblkno(fs, length);
203 		flags |= BA_CLRBUF;
204 		error = ext2_balloc(oip, lbn, offset, cred, &bp, flags);
205 		if (error)
206 			return (error);
207 		oip->i_size = length;
208 		size = blksize(fs, oip, lbn);
209 		bzero((char *)bp->b_data + offset, (u_int)(size - offset));
210 		allocbuf(bp, size);
211 		if (bp->b_bufsize == fs->e2fs_bsize)
212 			bp->b_flags |= B_CLUSTEROK;
213 		if (flags & IO_SYNC)
214 			bwrite(bp);
215 		else if (DOINGASYNC(ovp))
216 			bdwrite(bp);
217 		else
218 			bawrite(bp);
219 	}
220 	/*
221 	 * Calculate index into inode's block list of
222 	 * last direct and indirect blocks (if any)
223 	 * which we want to keep.  Lastblock is -1 when
224 	 * the file is truncated to 0.
225 	 */
226 	lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
227 	lastiblock[SINGLE] = lastblock - EXT2_NDADDR;
228 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
229 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
230 	nblocks = btodb(fs->e2fs_bsize);
231 	/*
232 	 * Update file and block pointers on disk before we start freeing
233 	 * blocks.  If we crash before free'ing blocks below, the blocks
234 	 * will be returned to the free list.  lastiblock values are also
235 	 * normalized to -1 for calls to ext2_indirtrunc below.
236 	 */
237 	for (level = TRIPLE; level >= SINGLE; level--) {
238 		oldblks[EXT2_NDADDR + level] = oip->i_ib[level];
239 		if (lastiblock[level] < 0) {
240 			oip->i_ib[level] = 0;
241 			lastiblock[level] = -1;
242 		}
243 	}
244 	for (i = 0; i < EXT2_NDADDR; i++) {
245 		oldblks[i] = oip->i_db[i];
246 		if (i > lastblock)
247 			oip->i_db[i] = 0;
248 	}
249 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
250 	allerror = ext2_update(ovp, !DOINGASYNC(ovp));
251 
252 	/*
253 	 * Having written the new inode to disk, save its new configuration
254 	 * and put back the old block pointers long enough to process them.
255 	 * Note that we save the new block configuration so we can check it
256 	 * when we are done.
257 	 */
258 	for (i = 0; i < EXT2_NDADDR; i++) {
259 		newblks[i] = oip->i_db[i];
260 		oip->i_db[i] = oldblks[i];
261 	}
262 	for (i = 0; i < EXT2_NIADDR; i++) {
263 		newblks[EXT2_NDADDR + i] = oip->i_ib[i];
264 		oip->i_ib[i] = oldblks[EXT2_NDADDR + i];
265 	}
266 	oip->i_size = osize;
267 	error = vtruncbuf(ovp, cred, length, (int)fs->e2fs_bsize);
268 	if (error && (allerror == 0))
269 		allerror = error;
270 	vnode_pager_setsize(ovp, length);
271 
272 	/*
273 	 * Indirect blocks first.
274 	 */
275 	indir_lbn[SINGLE] = -EXT2_NDADDR;
276 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
277 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
278 	for (level = TRIPLE; level >= SINGLE; level--) {
279 		bn = oip->i_ib[level];
280 		if (bn != 0) {
281 			error = ext2_indirtrunc(oip, indir_lbn[level],
282 			    fsbtodb(fs, bn), lastiblock[level], level, &count);
283 			if (error)
284 				allerror = error;
285 			blocksreleased += count;
286 			if (lastiblock[level] < 0) {
287 				oip->i_ib[level] = 0;
288 				ext2_blkfree(oip, bn, fs->e2fs_fsize);
289 				blocksreleased += nblocks;
290 			}
291 		}
292 		if (lastiblock[level] >= 0)
293 			goto done;
294 	}
295 
296 	/*
297 	 * All whole direct blocks or frags.
298 	 */
299 	for (i = EXT2_NDADDR - 1; i > lastblock; i--) {
300 		long bsize;
301 
302 		bn = oip->i_db[i];
303 		if (bn == 0)
304 			continue;
305 		oip->i_db[i] = 0;
306 		bsize = blksize(fs, oip, i);
307 		ext2_blkfree(oip, bn, bsize);
308 		blocksreleased += btodb(bsize);
309 	}
310 	if (lastblock < 0)
311 		goto done;
312 
313 	/*
314 	 * Finally, look for a change in size of the
315 	 * last direct block; release any frags.
316 	 */
317 	bn = oip->i_db[lastblock];
318 	if (bn != 0) {
319 		long oldspace, newspace;
320 
321 		/*
322 		 * Calculate amount of space we're giving
323 		 * back as old block size minus new block size.
324 		 */
325 		oldspace = blksize(fs, oip, lastblock);
326 		oip->i_size = length;
327 		newspace = blksize(fs, oip, lastblock);
328 		if (newspace == 0)
329 			panic("ext2_truncate: newspace");
330 		if (oldspace - newspace > 0) {
331 			/*
332 			 * Block number of space to be free'd is
333 			 * the old block # plus the number of frags
334 			 * required for the storage we're keeping.
335 			 */
336 			bn += numfrags(fs, newspace);
337 			ext2_blkfree(oip, bn, oldspace - newspace);
338 			blocksreleased += btodb(oldspace - newspace);
339 		}
340 	}
341 done:
342 #ifdef INVARIANTS
343 	for (level = SINGLE; level <= TRIPLE; level++)
344 		if (newblks[EXT2_NDADDR + level] != oip->i_ib[level])
345 			panic("itrunc1");
346 	for (i = 0; i < EXT2_NDADDR; i++)
347 		if (newblks[i] != oip->i_db[i])
348 			panic("itrunc2");
349 	BO_LOCK(bo);
350 	if (length == 0 && (bo->bo_dirty.bv_cnt != 0 ||
351 	    bo->bo_clean.bv_cnt != 0))
352 		panic("itrunc3");
353 	BO_UNLOCK(bo);
354 #endif	/* INVARIANTS */
355 	/*
356 	 * Put back the real size.
357 	 */
358 	oip->i_size = length;
359 	if (oip->i_blocks >= blocksreleased)
360 		oip->i_blocks -= blocksreleased;
361 	else				/* sanity */
362 		oip->i_blocks = 0;
363 	oip->i_flag |= IN_CHANGE;
364 	vnode_pager_setsize(ovp, length);
365 	return (allerror);
366 }
367 
368 /*
369  * Release blocks associated with the inode ip and stored in the indirect
370  * block bn.  Blocks are free'd in LIFO order up to (but not including)
371  * lastbn.  If level is greater than SINGLE, the block is an indirect block
372  * and recursive calls to indirtrunc must be used to cleanse other indirect
373  * blocks.
374  *
375  * NB: triple indirect blocks are untested.
376  */
377 
378 static int
379 ext2_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn,
380     daddr_t lastbn, int level, e4fs_daddr_t *countp)
381 {
382 	struct buf *bp;
383 	struct m_ext2fs *fs = ip->i_e2fs;
384 	struct vnode *vp;
385 	e2fs_daddr_t *bap, *copy;
386 	int i, nblocks, error = 0, allerror = 0;
387 	e2fs_lbn_t nb, nlbn, last;
388 	e4fs_daddr_t blkcount, factor, blocksreleased = 0;
389 
390 	/*
391 	 * Calculate index in current block of last
392 	 * block to be kept.  -1 indicates the entire
393 	 * block so we need not calculate the index.
394 	 */
395 	factor = 1;
396 	for (i = SINGLE; i < level; i++)
397 		factor *= NINDIR(fs);
398 	last = lastbn;
399 	if (lastbn > 0)
400 		last /= factor;
401 	nblocks = btodb(fs->e2fs_bsize);
402 	/*
403 	 * Get buffer of block pointers, zero those entries corresponding
404 	 * to blocks to be free'd, and update on disk copy first.  Since
405 	 * double(triple) indirect before single(double) indirect, calls
406 	 * to bmap on these blocks will fail.  However, we already have
407 	 * the on disk address, so we have to set the b_blkno field
408 	 * explicitly instead of letting bread do everything for us.
409 	 */
410 	vp = ITOV(ip);
411 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0, 0);
412 	if ((bp->b_flags & (B_DONE | B_DELWRI)) == 0) {
413 		bp->b_iocmd = BIO_READ;
414 		if (bp->b_bcount > bp->b_bufsize)
415 			panic("ext2_indirtrunc: bad buffer size");
416 		bp->b_blkno = dbn;
417 		vfs_busy_pages(bp, 0);
418 		bp->b_iooffset = dbtob(bp->b_blkno);
419 		bstrategy(bp);
420 		error = bufwait(bp);
421 	}
422 	if (error) {
423 		brelse(bp);
424 		*countp = 0;
425 		return (error);
426 	}
427 	bap = (e2fs_daddr_t *)bp->b_data;
428 	copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK);
429 	bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize);
430 	bzero((caddr_t)&bap[last + 1],
431 	    (NINDIR(fs) - (last + 1)) * sizeof(e2fs_daddr_t));
432 	if (last == -1)
433 		bp->b_flags |= B_INVAL;
434 	if (DOINGASYNC(vp)) {
435 		bdwrite(bp);
436 	} else {
437 		error = bwrite(bp);
438 		if (error)
439 			allerror = error;
440 	}
441 	bap = copy;
442 
443 	/*
444 	 * Recursively free totally unused blocks.
445 	 */
446 	for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
447 	    i--, nlbn += factor) {
448 		nb = bap[i];
449 		if (nb == 0)
450 			continue;
451 		if (level > SINGLE) {
452 			if ((error = ext2_indirtrunc(ip, nlbn,
453 			    fsbtodb(fs, nb), (int32_t)-1, level - 1, &blkcount)) != 0)
454 				allerror = error;
455 			blocksreleased += blkcount;
456 		}
457 		ext2_blkfree(ip, nb, fs->e2fs_bsize);
458 		blocksreleased += nblocks;
459 	}
460 
461 	/*
462 	 * Recursively free last partial block.
463 	 */
464 	if (level > SINGLE && lastbn >= 0) {
465 		last = lastbn % factor;
466 		nb = bap[i];
467 		if (nb != 0) {
468 			if ((error = ext2_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
469 			    last, level - 1, &blkcount)) != 0)
470 				allerror = error;
471 			blocksreleased += blkcount;
472 		}
473 	}
474 	free(copy, M_TEMP);
475 	*countp = blocksreleased;
476 	return (allerror);
477 }
478 
479 /*
480  *	discard preallocated blocks
481  */
482 int
483 ext2_inactive(struct vop_inactive_args *ap)
484 {
485 	struct vnode *vp = ap->a_vp;
486 	struct inode *ip = VTOI(vp);
487 	struct thread *td = ap->a_td;
488 	int mode, error = 0;
489 
490 	/*
491 	 * Ignore inodes related to stale file handles.
492 	 */
493 	if (ip->i_mode == 0)
494 		goto out;
495 	if (ip->i_nlink <= 0) {
496 		ext2_extattr_free(ip);
497 		error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td);
498 		ip->i_rdev = 0;
499 		mode = ip->i_mode;
500 		ip->i_mode = 0;
501 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
502 		ext2_vfree(vp, ip->i_number, mode);
503 	}
504 	if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
505 		ext2_update(vp, 0);
506 out:
507 	/*
508 	 * If we are done with the inode, reclaim it
509 	 * so that it can be reused immediately.
510 	 */
511 	if (ip->i_mode == 0)
512 		vrecycle(vp);
513 	return (error);
514 }
515 
516 /*
517  * Reclaim an inode so that it can be used for other purposes.
518  */
519 int
520 ext2_reclaim(struct vop_reclaim_args *ap)
521 {
522 	struct inode *ip;
523 	struct vnode *vp = ap->a_vp;
524 
525 	ip = VTOI(vp);
526 	if (ip->i_flag & IN_LAZYMOD) {
527 		ip->i_flag |= IN_MODIFIED;
528 		ext2_update(vp, 0);
529 	}
530 	vfs_hash_remove(vp);
531 	free(vp->v_data, M_EXT2NODE);
532 	vp->v_data = 0;
533 	vnode_destroy_vobject(vp);
534 	return (0);
535 }
536