xref: /freebsd/sys/fs/ext2fs/ext2_inode.c (revision 5686c6c38a3e1cc78804eaf5f880bda23dcf592f)
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(struct vnode *vp, int waitfor)
70 {
71 	struct m_ext2fs *fs;
72 	struct buf *bp;
73 	struct inode *ip;
74 	int error;
75 
76 	ASSERT_VOP_ELOCKED(vp, "ext2_update");
77 	ext2_itimes(vp);
78 	ip = VTOI(vp);
79 	if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
80 		return (0);
81 	ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED);
82 	fs = ip->i_e2fs;
83 	if(fs->e2fs_ronly)
84 		return (0);
85 	if ((error = bread(ip->i_devvp,
86 	    fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
87 		(int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
88 		brelse(bp);
89 		return (error);
90 	}
91 	ext2_i2ei(ip, (struct ext2fs_dinode *)((char *)bp->b_data +
92 	    EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)));
93 	if (waitfor && !DOINGASYNC(vp))
94 		return (bwrite(bp));
95 	else {
96 		bdwrite(bp);
97 		return (0);
98 	}
99 }
100 
101 #define	SINGLE	0	/* index of single indirect block */
102 #define	DOUBLE	1	/* index of double indirect block */
103 #define	TRIPLE	2	/* index of triple indirect block */
104 /*
105  * Truncate the inode oip to at most length size, freeing the
106  * disk blocks.
107  */
108 int
109 ext2_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred,
110     struct thread *td)
111 {
112 	struct vnode *ovp = vp;
113 	int32_t lastblock;
114 	struct inode *oip;
115 	int32_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
116 	uint32_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
117 	struct bufobj *bo;
118 	struct m_ext2fs *fs;
119 	struct buf *bp;
120 	int offset, size, level;
121 	long count, nblocks, blocksreleased = 0;
122 	int error, i, allerror;
123 	off_t osize;
124 
125 	oip = VTOI(ovp);
126 	bo = &ovp->v_bufobj;
127 
128 	ASSERT_VOP_LOCKED(vp, "ext2_truncate");
129 
130 	if (length < 0)
131 	    return (EINVAL);
132 
133 	if (ovp->v_type == VLNK &&
134 	    oip->i_size < ovp->v_mount->mnt_maxsymlinklen) {
135 #ifdef DIAGNOSTIC
136 		if (length != 0)
137 			panic("ext2_truncate: partial truncate of symlink");
138 #endif
139 		bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
140 		oip->i_size = 0;
141 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
142 		return (ext2_update(ovp, 1));
143 	}
144 	if (oip->i_size == length) {
145 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
146 		return (ext2_update(ovp, 0));
147 	}
148 	fs = oip->i_e2fs;
149 	osize = oip->i_size;
150 	/*
151 	 * Lengthen the size of the file. We must ensure that the
152 	 * last byte of the file is allocated. Since the smallest
153 	 * value of osize is 0, length will be at least 1.
154 	 */
155 	if (osize < length) {
156 		if (length > oip->i_e2fs->e2fs_maxfilesize)
157 			return (EFBIG);
158 		vnode_pager_setsize(ovp, length);
159 		offset = blkoff(fs, length - 1);
160 		lbn = lblkno(fs, length - 1);
161 		flags |= BA_CLRBUF;
162 		error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags);
163 		if (error) {
164 			vnode_pager_setsize(vp, osize);
165 			return (error);
166 		}
167 		oip->i_size = length;
168 		if (bp->b_bufsize == fs->e2fs_bsize)
169 			bp->b_flags |= B_CLUSTEROK;
170 		if (flags & IO_SYNC)
171 			bwrite(bp);
172 		else if (DOINGASYNC(ovp))
173 			bdwrite(bp);
174 		else
175 			bawrite(bp);
176 		oip->i_flag |= IN_CHANGE | IN_UPDATE;
177 		return (ext2_update(ovp, !DOINGASYNC(ovp)));
178 	}
179 	/*
180 	 * Shorten the size of the file. If the file is not being
181 	 * truncated to a block boundry, the contents of the
182 	 * partial block following the end of the file must be
183 	 * zero'ed in case it ever become accessible again because
184 	 * of subsequent file growth.
185 	 */
186 	/* I don't understand the comment above */
187 	offset = blkoff(fs, length);
188 	if (offset == 0) {
189 		oip->i_size = length;
190 	} else {
191 		lbn = lblkno(fs, length);
192 		flags |= BA_CLRBUF;
193 		error = ext2_balloc(oip, lbn, offset, cred, &bp, flags);
194 		if (error)
195 			return (error);
196 		oip->i_size = length;
197 		size = blksize(fs, oip, lbn);
198 		bzero((char *)bp->b_data + offset, (u_int)(size - offset));
199 		allocbuf(bp, size);
200 		if (bp->b_bufsize == fs->e2fs_bsize)
201 			bp->b_flags |= B_CLUSTEROK;
202 		if (flags & IO_SYNC)
203 			bwrite(bp);
204 		else if (DOINGASYNC(ovp))
205 			bdwrite(bp);
206 		else
207 			bawrite(bp);
208 	}
209 	/*
210 	 * Calculate index into inode's block list of
211 	 * last direct and indirect blocks (if any)
212 	 * which we want to keep.  Lastblock is -1 when
213 	 * the file is truncated to 0.
214 	 */
215 	lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
216 	lastiblock[SINGLE] = lastblock - NDADDR;
217 	lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
218 	lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
219 	nblocks = btodb(fs->e2fs_bsize);
220 	/*
221 	 * Update file and block pointers on disk before we start freeing
222 	 * blocks.  If we crash before free'ing blocks below, the blocks
223 	 * will be returned to the free list.  lastiblock values are also
224 	 * normalized to -1 for calls to ext2_indirtrunc below.
225 	 */
226 	bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof(oldblks));
227 	for (level = TRIPLE; level >= SINGLE; level--)
228 		if (lastiblock[level] < 0) {
229 			oip->i_ib[level] = 0;
230 			lastiblock[level] = -1;
231 		}
232 	for (i = NDADDR - 1; i > lastblock; i--)
233 		oip->i_db[i] = 0;
234 	oip->i_flag |= IN_CHANGE | IN_UPDATE;
235 	allerror = ext2_update(ovp, !DOINGASYNC(ovp));
236 
237 	/*
238 	 * Having written the new inode to disk, save its new configuration
239 	 * and put back the old block pointers long enough to process them.
240 	 * Note that we save the new block configuration so we can check it
241 	 * when we are done.
242 	 */
243 	bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof(newblks));
244 	bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof(oldblks));
245 	oip->i_size = osize;
246 	error = vtruncbuf(ovp, cred, length, (int)fs->e2fs_bsize);
247 	if (error && (allerror == 0))
248 		allerror = error;
249 	vnode_pager_setsize(ovp, length);
250 
251 	/*
252 	 * Indirect blocks first.
253 	 */
254 	indir_lbn[SINGLE] = -NDADDR;
255 	indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
256 	indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
257 	for (level = TRIPLE; level >= SINGLE; level--) {
258 		bn = oip->i_ib[level];
259 		if (bn != 0) {
260 			error = ext2_indirtrunc(oip, indir_lbn[level],
261 			    fsbtodb(fs, bn), lastiblock[level], level, &count);
262 			if (error)
263 				allerror = error;
264 			blocksreleased += count;
265 			if (lastiblock[level] < 0) {
266 				oip->i_ib[level] = 0;
267 				ext2_blkfree(oip, bn, fs->e2fs_fsize);
268 				blocksreleased += nblocks;
269 			}
270 		}
271 		if (lastiblock[level] >= 0)
272 			goto done;
273 	}
274 
275 	/*
276 	 * All whole direct blocks or frags.
277 	 */
278 	for (i = NDADDR - 1; i > lastblock; i--) {
279 		long bsize;
280 
281 		bn = oip->i_db[i];
282 		if (bn == 0)
283 			continue;
284 		oip->i_db[i] = 0;
285 		bsize = blksize(fs, oip, i);
286 		ext2_blkfree(oip, bn, bsize);
287 		blocksreleased += btodb(bsize);
288 	}
289 	if (lastblock < 0)
290 		goto done;
291 
292 	/*
293 	 * Finally, look for a change in size of the
294 	 * last direct block; release any frags.
295 	 */
296 	bn = oip->i_db[lastblock];
297 	if (bn != 0) {
298 		long oldspace, newspace;
299 
300 		/*
301 		 * Calculate amount of space we're giving
302 		 * back as old block size minus new block size.
303 		 */
304 		oldspace = blksize(fs, oip, lastblock);
305 		oip->i_size = length;
306 		newspace = blksize(fs, oip, lastblock);
307 		if (newspace == 0)
308 			panic("ext2_truncate: newspace");
309 		if (oldspace - newspace > 0) {
310 			/*
311 			 * Block number of space to be free'd is
312 			 * the old block # plus the number of frags
313 			 * required for the storage we're keeping.
314 			 */
315 			bn += numfrags(fs, newspace);
316 			ext2_blkfree(oip, bn, oldspace - newspace);
317 			blocksreleased += btodb(oldspace - newspace);
318 		}
319 	}
320 done:
321 #ifdef DIAGNOSTIC
322 	for (level = SINGLE; level <= TRIPLE; level++)
323 		if (newblks[NDADDR + level] != oip->i_ib[level])
324 			panic("itrunc1");
325 	for (i = 0; i < NDADDR; i++)
326 		if (newblks[i] != oip->i_db[i])
327 			panic("itrunc2");
328 	BO_LOCK(bo);
329 	if (length == 0 && (bo->bo_dirty.bv_cnt != 0 ||
330 	    bo->bo_clean.bv_cnt != 0))
331 		panic("itrunc3");
332 	BO_UNLOCK(bo);
333 #endif /* DIAGNOSTIC */
334 	/*
335 	 * Put back the real size.
336 	 */
337 	oip->i_size = length;
338 	if (oip->i_blocks >= blocksreleased)
339 		oip->i_blocks -= blocksreleased;
340 	else				/* sanity */
341 		oip->i_blocks = 0;
342 	oip->i_flag |= IN_CHANGE;
343 	vnode_pager_setsize(ovp, length);
344 	return (allerror);
345 }
346 
347 /*
348  * Release blocks associated with the inode ip and stored in the indirect
349  * block bn.  Blocks are free'd in LIFO order up to (but not including)
350  * lastbn.  If level is greater than SINGLE, the block is an indirect block
351  * and recursive calls to indirtrunc must be used to cleanse other indirect
352  * blocks.
353  *
354  * NB: triple indirect blocks are untested.
355  */
356 
357 static int
358 ext2_indirtrunc(struct inode *ip, int32_t lbn, int32_t dbn, int32_t lastbn,
359     int level, long *countp)
360 {
361 	struct buf *bp;
362 	struct m_ext2fs *fs = ip->i_e2fs;
363 	struct vnode *vp;
364 	int32_t *bap, *copy, nb, nlbn, last;
365 	long blkcount, factor;
366 	int i, nblocks, blocksreleased = 0;
367 	int error = 0, allerror = 0;
368 
369 	/*
370 	 * Calculate index in current block of last
371 	 * block to be kept.  -1 indicates the entire
372 	 * block so we need not calculate the index.
373 	 */
374 	factor = 1;
375 	for (i = SINGLE; i < level; i++)
376 		factor *= NINDIR(fs);
377 	last = lastbn;
378 	if (lastbn > 0)
379 		last /= factor;
380 	nblocks = btodb(fs->e2fs_bsize);
381 	/*
382 	 * Get buffer of block pointers, zero those entries corresponding
383 	 * to blocks to be free'd, and update on disk copy first.  Since
384 	 * double(triple) indirect before single(double) indirect, calls
385 	 * to bmap on these blocks will fail.  However, we already have
386 	 * the on disk address, so we have to set the b_blkno field
387 	 * explicitly instead of letting bread do everything for us.
388 	 */
389 	vp = ITOV(ip);
390 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0, 0);
391 	if ((bp->b_flags & (B_DONE | B_DELWRI)) == 0) {
392 		bp->b_iocmd = BIO_READ;
393 		if (bp->b_bcount > bp->b_bufsize)
394 			panic("ext2_indirtrunc: bad buffer size");
395 		bp->b_blkno = dbn;
396 		vfs_busy_pages(bp, 0);
397 		bp->b_iooffset = dbtob(bp->b_blkno);
398 		bstrategy(bp);
399 		error = bufwait(bp);
400 	}
401 	if (error) {
402 		brelse(bp);
403 		*countp = 0;
404 		return (error);
405 	}
406 
407 	bap = (int32_t *)bp->b_data;
408 	copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK);
409 	bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize);
410 	bzero((caddr_t)&bap[last + 1],
411 	  (u_int)(NINDIR(fs) - (last + 1)) * sizeof(int32_t));
412 	if (last == -1)
413 		bp->b_flags |= B_INVAL;
414 	if (DOINGASYNC(vp)) {
415 		bdwrite(bp);
416 	} else {
417 		error = bwrite(bp);
418 		if (error)
419 			allerror = error;
420 	}
421 	bap = copy;
422 
423 	/*
424 	 * Recursively free totally unused blocks.
425 	 */
426 	for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
427 	    i--, nlbn += factor) {
428 		nb = bap[i];
429 		if (nb == 0)
430 			continue;
431 		if (level > SINGLE) {
432 			if ((error = ext2_indirtrunc(ip, nlbn,
433 			    fsbtodb(fs, nb), (int32_t)-1, level - 1, &blkcount)) != 0)
434 				allerror = error;
435 			blocksreleased += blkcount;
436 		}
437 		ext2_blkfree(ip, nb, fs->e2fs_bsize);
438 		blocksreleased += nblocks;
439 	}
440 
441 	/*
442 	 * Recursively free last partial block.
443 	 */
444 	if (level > SINGLE && lastbn >= 0) {
445 		last = lastbn % factor;
446 		nb = bap[i];
447 		if (nb != 0) {
448 			if ((error = ext2_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
449 			    last, level - 1, &blkcount)) != 0)
450 				allerror = error;
451 			blocksreleased += blkcount;
452 		}
453 	}
454 	free(copy, M_TEMP);
455 	*countp = blocksreleased;
456 	return (allerror);
457 }
458 
459 /*
460  *	discard preallocated blocks
461  */
462 int
463 ext2_inactive(struct vop_inactive_args *ap)
464 {
465 	struct vnode *vp = ap->a_vp;
466 	struct inode *ip = VTOI(vp);
467 	struct thread *td = ap->a_td;
468 	int mode, error = 0;
469 
470 	/*
471 	 * Ignore inodes related to stale file handles.
472 	 */
473 	if (ip->i_mode == 0)
474 		goto out;
475 	if (ip->i_nlink <= 0) {
476 		error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td);
477 		ip->i_rdev = 0;
478 		mode = ip->i_mode;
479 		ip->i_mode = 0;
480 		ip->i_flag |= IN_CHANGE | IN_UPDATE;
481 		ext2_vfree(vp, ip->i_number, mode);
482 	}
483 	if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
484 		ext2_update(vp, 0);
485 out:
486 	/*
487 	 * If we are done with the inode, reclaim it
488 	 * so that it can be reused immediately.
489 	 */
490 	if (ip->i_mode == 0)
491 		vrecycle(vp);
492 	return (error);
493 }
494 
495 /*
496  * Reclaim an inode so that it can be used for other purposes.
497  */
498 int
499 ext2_reclaim(struct vop_reclaim_args *ap)
500 {
501 	struct inode *ip;
502 	struct vnode *vp = ap->a_vp;
503 
504 	ip = VTOI(vp);
505 	if (ip->i_flag & IN_LAZYMOD) {
506 		ip->i_flag |= IN_MODIFIED;
507 		ext2_update(vp, 0);
508 	}
509 	vfs_hash_remove(vp);
510 	free(vp->v_data, M_EXT2NODE);
511 	vp->v_data = 0;
512 	vnode_destroy_vobject(vp);
513 	return (0);
514 }
515