xref: /freebsd/sys/fs/ext2fs/ext2_inode.c (revision cc16dea626cf2fc80cde667ac4798065108e596c)
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 	bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof(oldblks));
228 	for (level = TRIPLE; level >= SINGLE; level--)
229 		if (lastiblock[level] < 0) {
230 			oip->i_ib[level] = 0;
231 			lastiblock[level] = -1;
232 		}
233 	for (i = NDADDR - 1; i > lastblock; i--)
234 		oip->i_db[i] = 0;
235 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
236 	allerror = ext2_update(ovp, !DOINGASYNC(ovp));
237 
238 	/*
239 	 * Having written the new inode to disk, save its new configuration
240 	 * and put back the old block pointers long enough to process them.
241 	 * Note that we save the new block configuration so we can check it
242 	 * when we are done.
243 	 */
244 	bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof(newblks));
245 	bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof(oldblks));
246 	oip->i_size = osize;
247 	error = vtruncbuf(ovp, cred, length, (int)fs->e2fs_bsize);
248 	if (error && (allerror == 0))
249 		allerror = error;
250 	vnode_pager_setsize(ovp, length);
251 
252 	/*
253 	 * Indirect blocks first.
254 	 */
255 	indir_lbn[SINGLE] = -NDADDR;
256 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
257 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
258 	for (level = TRIPLE; level >= SINGLE; level--) {
259 		bn = oip->i_ib[level];
260 		if (bn != 0) {
261 			error = ext2_indirtrunc(oip, indir_lbn[level],
262 			    fsbtodb(fs, bn), lastiblock[level], level, &count);
263 			if (error)
264 				allerror = error;
265 			blocksreleased += count;
266 			if (lastiblock[level] < 0) {
267 				oip->i_ib[level] = 0;
268 				ext2_blkfree(oip, bn, fs->e2fs_fsize);
269 				blocksreleased += nblocks;
270 			}
271 		}
272 		if (lastiblock[level] >= 0)
273 			goto done;
274 	}
275 
276 	/*
277 	 * All whole direct blocks or frags.
278 	 */
279 	for (i = NDADDR - 1; i > lastblock; i--) {
280 		long bsize;
281 
282 		bn = oip->i_db[i];
283 		if (bn == 0)
284 			continue;
285 		oip->i_db[i] = 0;
286 		bsize = blksize(fs, oip, i);
287 		ext2_blkfree(oip, bn, bsize);
288 		blocksreleased += btodb(bsize);
289 	}
290 	if (lastblock < 0)
291 		goto done;
292 
293 	/*
294 	 * Finally, look for a change in size of the
295 	 * last direct block; release any frags.
296 	 */
297 	bn = oip->i_db[lastblock];
298 	if (bn != 0) {
299 		long oldspace, newspace;
300 
301 		/*
302 		 * Calculate amount of space we're giving
303 		 * back as old block size minus new block size.
304 		 */
305 		oldspace = blksize(fs, oip, lastblock);
306 		oip->i_size = length;
307 		newspace = blksize(fs, oip, lastblock);
308 		if (newspace == 0)
309 			panic("ext2_truncate: newspace");
310 		if (oldspace - newspace > 0) {
311 			/*
312 			 * Block number of space to be free'd is
313 			 * the old block # plus the number of frags
314 			 * required for the storage we're keeping.
315 			 */
316 			bn += numfrags(fs, newspace);
317 			ext2_blkfree(oip, bn, oldspace - newspace);
318 			blocksreleased += btodb(oldspace - newspace);
319 		}
320 	}
321 done:
322 #ifdef INVARIANTS
323 	for (level = SINGLE; level <= TRIPLE; level++)
324 		if (newblks[NDADDR + level] != oip->i_ib[level])
325 			panic("itrunc1");
326 	for (i = 0; i < NDADDR; i++)
327 		if (newblks[i] != oip->i_db[i])
328 			panic("itrunc2");
329 	BO_LOCK(bo);
330 	if (length == 0 && (bo->bo_dirty.bv_cnt != 0 ||
331 	    bo->bo_clean.bv_cnt != 0))
332 		panic("itrunc3");
333 	BO_UNLOCK(bo);
334 #endif /* INVARIANTS */
335 	/*
336 	 * Put back the real size.
337 	 */
338 	oip->i_size = length;
339 	if (oip->i_blocks >= blocksreleased)
340 		oip->i_blocks -= blocksreleased;
341 	else				/* sanity */
342 		oip->i_blocks = 0;
343 	oip->i_flag |= IN_CHANGE;
344 	vnode_pager_setsize(ovp, length);
345 	return (allerror);
346 }
347 
348 /*
349  * Release blocks associated with the inode ip and stored in the indirect
350  * block bn.  Blocks are free'd in LIFO order up to (but not including)
351  * lastbn.  If level is greater than SINGLE, the block is an indirect block
352  * and recursive calls to indirtrunc must be used to cleanse other indirect
353  * blocks.
354  *
355  * NB: triple indirect blocks are untested.
356  */
357 
358 static int
359 ext2_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn,
360     daddr_t lastbn, int level, e4fs_daddr_t *countp)
361 {
362 	struct buf *bp;
363 	struct m_ext2fs *fs = ip->i_e2fs;
364 	struct vnode *vp;
365 	e2fs_daddr_t *bap, *copy;
366 	int i, nblocks, error = 0, allerror = 0;
367 	e2fs_lbn_t nb, nlbn, last;
368 	e4fs_daddr_t blkcount, factor, blocksreleased = 0;
369 
370 	/*
371 	 * Calculate index in current block of last
372 	 * block to be kept.  -1 indicates the entire
373 	 * block so we need not calculate the index.
374 	 */
375 	factor = 1;
376 	for (i = SINGLE; i < level; i++)
377 		factor *= NINDIR(fs);
378 	last = lastbn;
379 	if (lastbn > 0)
380 		last /= factor;
381 	nblocks = btodb(fs->e2fs_bsize);
382 	/*
383 	 * Get buffer of block pointers, zero those entries corresponding
384 	 * to blocks to be free'd, and update on disk copy first.  Since
385 	 * double(triple) indirect before single(double) indirect, calls
386 	 * to bmap on these blocks will fail.  However, we already have
387 	 * the on disk address, so we have to set the b_blkno field
388 	 * explicitly instead of letting bread do everything for us.
389 	 */
390 	vp = ITOV(ip);
391 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0, 0);
392 	if ((bp->b_flags & (B_DONE | B_DELWRI)) == 0) {
393 		bp->b_iocmd = BIO_READ;
394 		if (bp->b_bcount > bp->b_bufsize)
395 			panic("ext2_indirtrunc: bad buffer size");
396 		bp->b_blkno = dbn;
397 		vfs_busy_pages(bp, 0);
398 		bp->b_iooffset = dbtob(bp->b_blkno);
399 		bstrategy(bp);
400 		error = bufwait(bp);
401 	}
402 	if (error) {
403 		brelse(bp);
404 		*countp = 0;
405 		return (error);
406 	}
407 
408 	bap = (e2fs_daddr_t *)bp->b_data;
409 	copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK);
410 	bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize);
411 	bzero((caddr_t)&bap[last + 1],
412 	  (NINDIR(fs) - (last + 1)) * sizeof(e2fs_daddr_t));
413 	if (last == -1)
414 		bp->b_flags |= B_INVAL;
415 	if (DOINGASYNC(vp)) {
416 		bdwrite(bp);
417 	} else {
418 		error = bwrite(bp);
419 		if (error)
420 			allerror = error;
421 	}
422 	bap = copy;
423 
424 	/*
425 	 * Recursively free totally unused blocks.
426 	 */
427 	for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
428 	    i--, nlbn += factor) {
429 		nb = bap[i];
430 		if (nb == 0)
431 			continue;
432 		if (level > SINGLE) {
433 			if ((error = ext2_indirtrunc(ip, nlbn,
434 			    fsbtodb(fs, nb), (int32_t)-1, level - 1, &blkcount)) != 0)
435 				allerror = error;
436 			blocksreleased += blkcount;
437 		}
438 		ext2_blkfree(ip, nb, fs->e2fs_bsize);
439 		blocksreleased += nblocks;
440 	}
441 
442 	/*
443 	 * Recursively free last partial block.
444 	 */
445 	if (level > SINGLE && lastbn >= 0) {
446 		last = lastbn % factor;
447 		nb = bap[i];
448 		if (nb != 0) {
449 			if ((error = ext2_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
450 			    last, level - 1, &blkcount)) != 0)
451 				allerror = error;
452 			blocksreleased += blkcount;
453 		}
454 	}
455 	free(copy, M_TEMP);
456 	*countp = blocksreleased;
457 	return (allerror);
458 }
459 
460 /*
461  *	discard preallocated blocks
462  */
463 int
464 ext2_inactive(struct vop_inactive_args *ap)
465 {
466 	struct vnode *vp = ap->a_vp;
467 	struct inode *ip = VTOI(vp);
468 	struct thread *td = ap->a_td;
469 	int mode, error = 0;
470 
471 	/*
472 	 * Ignore inodes related to stale file handles.
473 	 */
474 	if (ip->i_mode == 0)
475 		goto out;
476 	if (ip->i_nlink <= 0) {
477 		error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td);
478 		ip->i_rdev = 0;
479 		mode = ip->i_mode;
480 		ip->i_mode = 0;
481 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
482 		ext2_vfree(vp, ip->i_number, mode);
483 	}
484 	if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
485 		ext2_update(vp, 0);
486 out:
487 	/*
488 	 * If we are done with the inode, reclaim it
489 	 * so that it can be reused immediately.
490 	 */
491 	if (ip->i_mode == 0)
492 		vrecycle(vp);
493 	return (error);
494 }
495 
496 /*
497  * Reclaim an inode so that it can be used for other purposes.
498  */
499 int
500 ext2_reclaim(struct vop_reclaim_args *ap)
501 {
502 	struct inode *ip;
503 	struct vnode *vp = ap->a_vp;
504 
505 	ip = VTOI(vp);
506 	if (ip->i_flag & IN_LAZYMOD) {
507 		ip->i_flag |= IN_MODIFIED;
508 		ext2_update(vp, 0);
509 	}
510 	vfs_hash_remove(vp);
511 	free(vp->v_data, M_EXT2NODE);
512 	vp->v_data = 0;
513 	vnode_destroy_vobject(vp);
514 	return (0);
515 }
516