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 * SPDX-License-Identifier: BSD-3-Clause
9 *
10 * Copyright (c) 1982, 1986, 1989, 1993
11 * The Regents of the University of California. All rights reserved.
12 *
13 * Redistribution and use in source and binary forms, with or without
14 * modification, are permitted provided that the following conditions
15 * are met:
16 * 1. Redistributions of source code must retain the above copyright
17 * notice, this list of conditions and the following disclaimer.
18 * 2. Redistributions in binary form must reproduce the above copyright
19 * notice, this list of conditions and the following disclaimer in the
20 * documentation and/or other materials provided with the distribution.
21 * 3. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
35 * SUCH DAMAGE.
36 *
37 * @(#)ffs_inode.c 8.5 (Berkeley) 12/30/93
38 * $FreeBSD$
39 */
40
41 #include <sys/param.h>
42 #include <sys/systm.h>
43 #include <sys/mount.h>
44 #include <sys/bio.h>
45 #include <sys/buf.h>
46 #include <sys/vnode.h>
47 #include <sys/malloc.h>
48 #include <sys/rwlock.h>
49 #include <sys/sdt.h>
50
51 #include <vm/vm.h>
52 #include <vm/vm_extern.h>
53
54 #include <fs/ext2fs/fs.h>
55 #include <fs/ext2fs/inode.h>
56 #include <fs/ext2fs/ext2_mount.h>
57 #include <fs/ext2fs/ext2fs.h>
58 #include <fs/ext2fs/fs.h>
59 #include <fs/ext2fs/ext2_extern.h>
60 #include <fs/ext2fs/ext2_extattr.h>
61
62 /*
63 * Update the access, modified, and inode change times as specified by the
64 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode
65 * to disk if the IN_MODIFIED flag is set (it may be set initially, or by
66 * the timestamp update). The IN_LAZYMOD flag is set to force a write
67 * later if not now. If we write now, then clear both IN_MODIFIED and
68 * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is
69 * set, then wait for the write to complete.
70 */
71 int
72 ext2_update(struct vnode *vp, int waitfor)
73 {
74 struct m_ext2fs *fs;
75 struct buf *bp;
76 struct inode *ip;
77 int error;
78
79 ASSERT_VOP_ELOCKED(vp, "ext2_update");
80 ext2_itimes(vp);
81 ip = VTOI(vp);
82 if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
83 return (0);
84 ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED);
85 fs = ip->i_e2fs;
86 if (fs->e2fs_ronly)
87 return (0);
88 if ((error = bread(ip->i_devvp,
89 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
90 (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
91 brelse(bp);
92 return (error);
93 }
94 error = ext2_i2ei(ip, (struct ext2fs_dinode *)((char *)bp->b_data +
95 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)));
96 if (error) {
97 brelse(bp);
98 return (error);
99 }
100 if (waitfor && !DOINGASYNC(vp))
101 return (bwrite(bp));
102 else {
103 bdwrite(bp);
104 return (0);
105 }
106 }
107
108 #define SINGLE 0 /* index of single indirect block */
109 #define DOUBLE 1 /* index of double indirect block */
110 #define TRIPLE 2 /* index of triple indirect block */
111
112 /*
113 * Release blocks associated with the inode ip and stored in the indirect
114 * block bn. Blocks are free'd in LIFO order up to (but not including)
115 * lastbn. If level is greater than SINGLE, the block is an indirect block
116 * and recursive calls to indirtrunc must be used to cleanse other indirect
117 * blocks.
118 *
119 * NB: triple indirect blocks are untested.
120 */
121 static int
122 ext2_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn,
123 daddr_t lastbn, int level, e4fs_daddr_t *countp)
124 {
125 struct buf *bp;
126 struct m_ext2fs *fs = ip->i_e2fs;
127 struct vnode *vp;
128 e2fs_daddr_t *bap, *copy;
129 int i, nblocks, error = 0, allerror = 0;
130 e2fs_lbn_t nb, nlbn, last;
131 e4fs_daddr_t blkcount, factor, blocksreleased = 0;
132
133 /*
134 * Calculate index in current block of last
135 * block to be kept. -1 indicates the entire
136 * block so we need not calculate the index.
137 */
138 factor = 1;
139 for (i = SINGLE; i < level; i++)
140 factor *= NINDIR(fs);
141 last = lastbn;
142 if (lastbn > 0)
143 last /= factor;
144 nblocks = btodb(fs->e2fs_bsize);
145 /*
146 * Get buffer of block pointers, zero those entries corresponding
147 * to blocks to be free'd, and update on disk copy first. Since
148 * double(triple) indirect before single(double) indirect, calls
149 * to bmap on these blocks will fail. However, we already have
150 * the on disk address, so we have to set the b_blkno field
151 * explicitly instead of letting bread do everything for us.
152 */
153 vp = ITOV(ip);
154 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0, 0);
155 if ((bp->b_flags & (B_DONE | B_DELWRI)) == 0) {
156 bp->b_iocmd = BIO_READ;
157 if (bp->b_bcount > bp->b_bufsize)
158 panic("ext2_indirtrunc: bad buffer size");
159 bp->b_blkno = dbn;
160 vfs_busy_pages(bp, 0);
161 bp->b_iooffset = dbtob(bp->b_blkno);
162 bstrategy(bp);
163 error = bufwait(bp);
164 }
165 if (error) {
166 brelse(bp);
167 *countp = 0;
168 return (error);
169 }
170 bap = (e2fs_daddr_t *)bp->b_data;
171 copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK);
172 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize);
173 bzero((caddr_t)&bap[last + 1],
174 (NINDIR(fs) - (last + 1)) * sizeof(e2fs_daddr_t));
175 if (last == -1)
176 bp->b_flags |= B_INVAL;
177 if (DOINGASYNC(vp)) {
178 bdwrite(bp);
179 } else {
180 error = bwrite(bp);
181 if (error)
182 allerror = error;
183 }
184 bap = copy;
185
186 /*
187 * Recursively free totally unused blocks.
188 */
189 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
190 i--, nlbn += factor) {
191 nb = bap[i];
192 if (nb == 0)
193 continue;
194 if (level > SINGLE) {
195 if ((error = ext2_indirtrunc(ip, nlbn,
196 fsbtodb(fs, nb), (int32_t)-1, level - 1, &blkcount)) != 0)
197 allerror = error;
198 blocksreleased += blkcount;
199 }
200 ext2_blkfree(ip, nb, fs->e2fs_bsize);
201 blocksreleased += nblocks;
202 }
203
204 /*
205 * Recursively free last partial block.
206 */
207 if (level > SINGLE && lastbn >= 0) {
208 last = lastbn % factor;
209 nb = bap[i];
210 if (nb != 0) {
211 if ((error = ext2_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
212 last, level - 1, &blkcount)) != 0)
213 allerror = error;
214 blocksreleased += blkcount;
215 }
216 }
217 free(copy, M_TEMP);
218 *countp = blocksreleased;
219 return (allerror);
220 }
221
222 /*
223 * Truncate the inode oip to at most length size, freeing the
224 * disk blocks.
225 */
226 static int
227 ext2_ind_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred,
228 struct thread *td)
229 {
230 struct vnode *ovp = vp;
231 e4fs_daddr_t lastblock;
232 struct inode *oip;
233 e4fs_daddr_t bn, lbn, lastiblock[EXT2_NIADDR], indir_lbn[EXT2_NIADDR];
234 uint32_t oldblks[EXT2_NDADDR + EXT2_NIADDR];
235 uint32_t newblks[EXT2_NDADDR + EXT2_NIADDR];
236 struct m_ext2fs *fs;
237 struct buf *bp;
238 int offset, size, level;
239 e4fs_daddr_t count, nblocks, blocksreleased = 0;
240 int error, i, allerror;
241 off_t osize;
242 #ifdef INVARIANTS
243 struct bufobj *bo;
244 #endif
245
246 oip = VTOI(ovp);
247 #ifdef INVARIANTS
248 bo = &ovp->v_bufobj;
249 #endif
250
251 fs = oip->i_e2fs;
252 osize = oip->i_size;
253 /*
254 * Lengthen the size of the file. We must ensure that the
255 * last byte of the file is allocated. Since the smallest
256 * value of osize is 0, length will be at least 1.
257 */
258 if (osize < length) {
259 if (length > oip->i_e2fs->e2fs_maxfilesize)
260 return (EFBIG);
261 vnode_pager_setsize(ovp, length);
262 offset = blkoff(fs, length - 1);
263 lbn = lblkno(fs, length - 1);
264 flags |= BA_CLRBUF;
265 error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags);
266 if (error) {
267 vnode_pager_setsize(vp, osize);
268 return (error);
269 }
270 oip->i_size = length;
271 if (bp->b_bufsize == fs->e2fs_bsize)
272 bp->b_flags |= B_CLUSTEROK;
273 if (flags & IO_SYNC)
274 bwrite(bp);
275 else if (DOINGASYNC(ovp))
276 bdwrite(bp);
277 else
278 bawrite(bp);
279 oip->i_flag |= IN_CHANGE | IN_UPDATE;
280 return (ext2_update(ovp, !DOINGASYNC(ovp)));
281 }
282 /*
283 * Shorten the size of the file. If the file is not being
284 * truncated to a block boundary, the contents of the
285 * partial block following the end of the file must be
286 * zero'ed in case it ever become accessible again because
287 * of subsequent file growth.
288 */
289 /* I don't understand the comment above */
290 offset = blkoff(fs, length);
291 if (offset == 0) {
292 oip->i_size = length;
293 } else {
294 lbn = lblkno(fs, length);
295 flags |= BA_CLRBUF;
296 error = ext2_balloc(oip, lbn, offset, cred, &bp, flags);
297 if (error)
298 return (error);
299 oip->i_size = length;
300 size = blksize(fs, oip, lbn);
301 bzero((char *)bp->b_data + offset, (u_int)(size - offset));
302 allocbuf(bp, size);
303 if (bp->b_bufsize == fs->e2fs_bsize)
304 bp->b_flags |= B_CLUSTEROK;
305 if (flags & IO_SYNC)
306 bwrite(bp);
307 else if (DOINGASYNC(ovp))
308 bdwrite(bp);
309 else
310 bawrite(bp);
311 }
312 /*
313 * Calculate index into inode's block list of
314 * last direct and indirect blocks (if any)
315 * which we want to keep. Lastblock is -1 when
316 * the file is truncated to 0.
317 */
318 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
319 lastiblock[SINGLE] = lastblock - EXT2_NDADDR;
320 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
321 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
322 nblocks = btodb(fs->e2fs_bsize);
323 /*
324 * Update file and block pointers on disk before we start freeing
325 * blocks. If we crash before free'ing blocks below, the blocks
326 * will be returned to the free list. lastiblock values are also
327 * normalized to -1 for calls to ext2_indirtrunc below.
328 */
329 for (level = TRIPLE; level >= SINGLE; level--) {
330 oldblks[EXT2_NDADDR + level] = oip->i_ib[level];
331 if (lastiblock[level] < 0) {
332 oip->i_ib[level] = 0;
333 lastiblock[level] = -1;
334 }
335 }
336 for (i = 0; i < EXT2_NDADDR; i++) {
337 oldblks[i] = oip->i_db[i];
338 if (i > lastblock)
339 oip->i_db[i] = 0;
340 }
341 oip->i_flag |= IN_CHANGE | IN_UPDATE;
342 allerror = ext2_update(ovp, !DOINGASYNC(ovp));
343
344 /*
345 * Having written the new inode to disk, save its new configuration
346 * and put back the old block pointers long enough to process them.
347 * Note that we save the new block configuration so we can check it
348 * when we are done.
349 */
350 for (i = 0; i < EXT2_NDADDR; i++) {
351 newblks[i] = oip->i_db[i];
352 oip->i_db[i] = oldblks[i];
353 }
354 for (i = 0; i < EXT2_NIADDR; i++) {
355 newblks[EXT2_NDADDR + i] = oip->i_ib[i];
356 oip->i_ib[i] = oldblks[EXT2_NDADDR + i];
357 }
358 oip->i_size = osize;
359 error = vtruncbuf(ovp, length, (int)fs->e2fs_bsize);
360 if (error && (allerror == 0))
361 allerror = error;
362 vnode_pager_setsize(ovp, length);
363
364 /*
365 * Indirect blocks first.
366 */
367 indir_lbn[SINGLE] = -EXT2_NDADDR;
368 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
369 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
370 for (level = TRIPLE; level >= SINGLE; level--) {
371 bn = oip->i_ib[level];
372 if (bn != 0) {
373 error = ext2_indirtrunc(oip, indir_lbn[level],
374 fsbtodb(fs, bn), lastiblock[level], level, &count);
375 if (error)
376 allerror = error;
377 blocksreleased += count;
378 if (lastiblock[level] < 0) {
379 oip->i_ib[level] = 0;
380 ext2_blkfree(oip, bn, fs->e2fs_fsize);
381 blocksreleased += nblocks;
382 }
383 }
384 if (lastiblock[level] >= 0)
385 goto done;
386 }
387
388 /*
389 * All whole direct blocks or frags.
390 */
391 for (i = EXT2_NDADDR - 1; i > lastblock; i--) {
392 long bsize;
393
394 bn = oip->i_db[i];
395 if (bn == 0)
396 continue;
397 oip->i_db[i] = 0;
398 bsize = blksize(fs, oip, i);
399 ext2_blkfree(oip, bn, bsize);
400 blocksreleased += btodb(bsize);
401 }
402 if (lastblock < 0)
403 goto done;
404
405 /*
406 * Finally, look for a change in size of the
407 * last direct block; release any frags.
408 */
409 bn = oip->i_db[lastblock];
410 if (bn != 0) {
411 long oldspace, newspace;
412
413 /*
414 * Calculate amount of space we're giving
415 * back as old block size minus new block size.
416 */
417 oldspace = blksize(fs, oip, lastblock);
418 oip->i_size = length;
419 newspace = blksize(fs, oip, lastblock);
420 if (newspace == 0)
421 panic("ext2_truncate: newspace");
422 if (oldspace - newspace > 0) {
423 /*
424 * Block number of space to be free'd is
425 * the old block # plus the number of frags
426 * required for the storage we're keeping.
427 */
428 bn += numfrags(fs, newspace);
429 ext2_blkfree(oip, bn, oldspace - newspace);
430 blocksreleased += btodb(oldspace - newspace);
431 }
432 }
433 done:
434 #ifdef INVARIANTS
435 for (level = SINGLE; level <= TRIPLE; level++)
436 if (newblks[EXT2_NDADDR + level] != oip->i_ib[level])
437 panic("itrunc1");
438 for (i = 0; i < EXT2_NDADDR; i++)
439 if (newblks[i] != oip->i_db[i])
440 panic("itrunc2");
441 BO_LOCK(bo);
442 if (length == 0 && (bo->bo_dirty.bv_cnt != 0 ||
443 bo->bo_clean.bv_cnt != 0))
444 panic("itrunc3");
445 BO_UNLOCK(bo);
446 #endif /* INVARIANTS */
447 /*
448 * Put back the real size.
449 */
450 oip->i_size = length;
451 if (oip->i_blocks >= blocksreleased)
452 oip->i_blocks -= blocksreleased;
453 else /* sanity */
454 oip->i_blocks = 0;
455 oip->i_flag |= IN_CHANGE;
456 vnode_pager_setsize(ovp, length);
457 return (allerror);
458 }
459
460 static int
461 ext2_ext_truncate(struct vnode *vp, off_t length, int flags,
462 struct ucred *cred, struct thread *td)
463 {
464 struct vnode *ovp = vp;
465 int32_t lastblock;
466 struct m_ext2fs *fs;
467 struct inode *oip;
468 struct buf *bp;
469 uint32_t lbn, offset;
470 int error, size;
471 off_t osize;
472
473 oip = VTOI(ovp);
474 fs = oip->i_e2fs;
475 osize = oip->i_size;
476
477 if (osize < length) {
478 if (length > oip->i_e2fs->e2fs_maxfilesize) {
479 return (EFBIG);
480 }
481 vnode_pager_setsize(ovp, length);
482 offset = blkoff(fs, length - 1);
483 lbn = lblkno(fs, length - 1);
484 flags |= BA_CLRBUF;
485 error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags);
486 if (error) {
487 vnode_pager_setsize(vp, osize);
488 return (error);
489 }
490 oip->i_size = length;
491 if (bp->b_bufsize == fs->e2fs_bsize)
492 bp->b_flags |= B_CLUSTEROK;
493 if (flags & IO_SYNC)
494 bwrite(bp);
495 else if (DOINGASYNC(ovp))
496 bdwrite(bp);
497 else
498 bawrite(bp);
499 oip->i_flag |= IN_CHANGE | IN_UPDATE;
500 return (ext2_update(ovp, !DOINGASYNC(ovp)));
501 }
502
503 lastblock = (length + fs->e2fs_bsize - 1) / fs->e2fs_bsize;
504 error = ext4_ext_remove_space(oip, lastblock, flags, cred, td);
505 if (error)
506 return (error);
507
508 offset = blkoff(fs, length);
509 if (offset == 0) {
510 oip->i_size = length;
511 } else {
512 lbn = lblkno(fs, length);
513 flags |= BA_CLRBUF;
514 error = ext2_balloc(oip, lbn, offset, cred, &bp, flags);
515 if (error) {
516 return (error);
517 }
518 oip->i_size = length;
519 size = blksize(fs, oip, lbn);
520 bzero((char *)bp->b_data + offset, (u_int)(size - offset));
521 allocbuf(bp, size);
522 if (bp->b_bufsize == fs->e2fs_bsize)
523 bp->b_flags |= B_CLUSTEROK;
524 if (flags & IO_SYNC)
525 bwrite(bp);
526 else if (DOINGASYNC(ovp))
527 bdwrite(bp);
528 else
529 bawrite(bp);
530 }
531
532 oip->i_size = osize;
533 error = vtruncbuf(ovp, length, (int)fs->e2fs_bsize);
534 if (error)
535 return (error);
536
537 vnode_pager_setsize(ovp, length);
538
539 oip->i_size = length;
540 oip->i_flag |= IN_CHANGE | IN_UPDATE;
541 error = ext2_update(ovp, !DOINGASYNC(ovp));
542
543 return (error);
544 }
545
546 /*
547 * Truncate the inode ip to at most length size, freeing the
548 * disk blocks.
549 */
550 int
551 ext2_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred,
552 struct thread *td)
553 {
554 struct inode *ip;
555 int error;
556
557 ASSERT_VOP_LOCKED(vp, "ext2_truncate");
558
559 if (length < 0)
560 return (EINVAL);
561
562 ip = VTOI(vp);
563 if (vp->v_type == VLNK &&
564 ip->i_size < vp->v_mount->mnt_maxsymlinklen) {
565 #ifdef INVARIANTS
566 if (length != 0)
567 panic("ext2_truncate: partial truncate of symlink");
568 #endif
569 bzero((char *)&ip->i_shortlink, (u_int)ip->i_size);
570 ip->i_size = 0;
571 ip->i_flag |= IN_CHANGE | IN_UPDATE;
572 return (ext2_update(vp, 1));
573 }
574 if (ip->i_size == length) {
575 ip->i_flag |= IN_CHANGE | IN_UPDATE;
576 return (ext2_update(vp, 0));
577 }
578
579 if (ip->i_flag & IN_E4EXTENTS)
580 error = ext2_ext_truncate(vp, length, flags, cred, td);
581 else
582 error = ext2_ind_truncate(vp, length, flags, cred, td);
583
584 return (error);
585 }
586
587 /*
588 * discard preallocated blocks
589 */
590 int
591 ext2_inactive(struct vop_inactive_args *ap)
592 {
593 struct vnode *vp = ap->a_vp;
594 struct inode *ip = VTOI(vp);
595 struct thread *td = ap->a_td;
596 int mode, error = 0;
597
598 /*
599 * Ignore inodes related to stale file handles.
600 */
601 if (ip->i_mode == 0)
602 goto out;
603 if (ip->i_nlink <= 0) {
604 ext2_extattr_free(ip);
605 error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td);
606 if (!(ip->i_flag & IN_E4EXTENTS))
607 ip->i_rdev = 0;
608 mode = ip->i_mode;
609 ip->i_mode = 0;
610 ip->i_flag |= IN_CHANGE | IN_UPDATE;
611 ext2_vfree(vp, ip->i_number, mode);
612 }
613 if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
614 ext2_update(vp, 0);
615 out:
616 /*
617 * If we are done with the inode, reclaim it
618 * so that it can be reused immediately.
619 */
620 if (ip->i_mode == 0)
621 vrecycle(vp);
622 return (error);
623 }
624
625 /*
626 * Reclaim an inode so that it can be used for other purposes.
627 */
628 int
629 ext2_reclaim(struct vop_reclaim_args *ap)
630 {
631 struct inode *ip;
632 struct vnode *vp = ap->a_vp;
633
634 ip = VTOI(vp);
635 if (ip->i_flag & IN_LAZYMOD) {
636 ip->i_flag |= IN_MODIFIED;
637 ext2_update(vp, 0);
638 }
639 vfs_hash_remove(vp);
640 free(vp->v_data, M_EXT2NODE);
641 vp->v_data = 0;
642 vnode_destroy_vobject(vp);
643 return (0);
644 }
Cache object: 2d0f43c41dfa996a5c7aff45609b7bd2
|