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/endian.h>
47 #include <sys/vnode.h>
48 #include <sys/malloc.h>
49 #include <sys/rwlock.h>
50 #include <sys/sdt.h>
51
52 #include <vm/vm.h>
53 #include <vm/vm_extern.h>
54
55 #include <fs/ext2fs/fs.h>
56 #include <fs/ext2fs/inode.h>
57 #include <fs/ext2fs/ext2_mount.h>
58 #include <fs/ext2fs/ext2fs.h>
59 #include <fs/ext2fs/fs.h>
60 #include <fs/ext2fs/ext2_extern.h>
61 #include <fs/ext2fs/ext2_extattr.h>
62
63 /*
64 * Update the access, modified, and inode change times as specified by the
65 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode
66 * to disk if the IN_MODIFIED flag is set (it may be set initially, or by
67 * the timestamp update). The IN_LAZYMOD flag is set to force a write
68 * later if not now. If we write now, then clear both IN_MODIFIED and
69 * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is
70 * set, then wait for the write to complete.
71 */
72 int
73 ext2_update(struct vnode *vp, int waitfor)
74 {
75 struct m_ext2fs *fs;
76 struct buf *bp;
77 struct inode *ip;
78 int error;
79
80 ASSERT_VOP_ELOCKED(vp, "ext2_update");
81 ext2_itimes(vp);
82 ip = VTOI(vp);
83 if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
84 return (0);
85 ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED);
86 fs = ip->i_e2fs;
87 if (fs->e2fs_ronly)
88 return (0);
89 if ((error = bread(ip->i_devvp,
90 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
91 (int)fs->e2fs_bsize, NOCRED, &bp)) != 0) {
92 brelse(bp);
93 return (error);
94 }
95 error = ext2_i2ei(ip, (struct ext2fs_dinode *)((char *)bp->b_data +
96 EXT2_INODE_SIZE(fs) * ino_to_fsbo(fs, ip->i_number)));
97 if (error) {
98 brelse(bp);
99 return (error);
100 }
101 if (waitfor && !DOINGASYNC(vp))
102 return (bwrite(bp));
103 else {
104 bdwrite(bp);
105 return (0);
106 }
107 }
108
109 #define SINGLE 0 /* index of single indirect block */
110 #define DOUBLE 1 /* index of double indirect block */
111 #define TRIPLE 2 /* index of triple indirect block */
112
113 /*
114 * Release blocks associated with the inode ip and stored in the indirect
115 * block bn. Blocks are free'd in LIFO order up to (but not including)
116 * lastbn. If level is greater than SINGLE, the block is an indirect block
117 * and recursive calls to indirtrunc must be used to cleanse other indirect
118 * blocks.
119 *
120 * NB: triple indirect blocks are untested.
121 */
122 static int
123 ext2_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn,
124 daddr_t lastbn, int level, e4fs_daddr_t *countp)
125 {
126 struct buf *bp;
127 struct m_ext2fs *fs = ip->i_e2fs;
128 struct vnode *vp;
129 e2fs_daddr_t *bap, *copy;
130 int i, nblocks, error = 0, allerror = 0;
131 e2fs_lbn_t nb, nlbn, last;
132 e4fs_daddr_t blkcount, factor, blocksreleased = 0;
133
134 /*
135 * Calculate index in current block of last
136 * block to be kept. -1 indicates the entire
137 * block so we need not calculate the index.
138 */
139 factor = 1;
140 for (i = SINGLE; i < level; i++)
141 factor *= NINDIR(fs);
142 last = lastbn;
143 if (lastbn > 0)
144 last /= factor;
145 nblocks = btodb(fs->e2fs_bsize);
146 /*
147 * Get buffer of block pointers, zero those entries corresponding
148 * to blocks to be free'd, and update on disk copy first. Since
149 * double(triple) indirect before single(double) indirect, calls
150 * to bmap on these blocks will fail. However, we already have
151 * the on disk address, so we have to set the b_blkno field
152 * explicitly instead of letting bread do everything for us.
153 */
154 vp = ITOV(ip);
155 bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0, 0);
156 if ((bp->b_flags & (B_DONE | B_DELWRI)) == 0) {
157 bp->b_iocmd = BIO_READ;
158 if (bp->b_bcount > bp->b_bufsize)
159 panic("ext2_indirtrunc: bad buffer size");
160 bp->b_blkno = dbn;
161 vfs_busy_pages(bp, 0);
162 bp->b_iooffset = dbtob(bp->b_blkno);
163 bstrategy(bp);
164 error = bufwait(bp);
165 }
166 if (error) {
167 brelse(bp);
168 *countp = 0;
169 return (error);
170 }
171 bap = (e2fs_daddr_t *)bp->b_data;
172 copy = malloc(fs->e2fs_bsize, M_TEMP, M_WAITOK);
173 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->e2fs_bsize);
174 bzero((caddr_t)&bap[last + 1],
175 (NINDIR(fs) - (last + 1)) * sizeof(e2fs_daddr_t));
176 if (last == -1)
177 bp->b_flags |= B_INVAL;
178 if (DOINGASYNC(vp)) {
179 bdwrite(bp);
180 } else {
181 error = bwrite(bp);
182 if (error)
183 allerror = error;
184 }
185 bap = copy;
186
187 /*
188 * Recursively free totally unused blocks.
189 */
190 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
191 i--, nlbn += factor) {
192 nb = le32toh(bap[i]);
193 if (nb == 0)
194 continue;
195 if (level > SINGLE) {
196 if ((error = ext2_indirtrunc(ip, nlbn,
197 fsbtodb(fs, nb), (int32_t)-1, level - 1, &blkcount)) != 0)
198 allerror = error;
199 blocksreleased += blkcount;
200 }
201 ext2_blkfree(ip, nb, fs->e2fs_bsize);
202 blocksreleased += nblocks;
203 }
204
205 /*
206 * Recursively free last partial block.
207 */
208 if (level > SINGLE && lastbn >= 0) {
209 last = lastbn % factor;
210 nb = le32toh(bap[i]);
211 if (nb != 0) {
212 if ((error = ext2_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
213 last, level - 1, &blkcount)) != 0)
214 allerror = error;
215 blocksreleased += blkcount;
216 }
217 }
218 free(copy, M_TEMP);
219 *countp = blocksreleased;
220 return (allerror);
221 }
222
223 /*
224 * Truncate the inode oip to at most length size, freeing the
225 * disk blocks.
226 */
227 static int
228 ext2_ind_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred,
229 struct thread *td)
230 {
231 struct vnode *ovp = vp;
232 e4fs_daddr_t lastblock;
233 struct inode *oip;
234 e4fs_daddr_t bn, lbn, lastiblock[EXT2_NIADDR], indir_lbn[EXT2_NIADDR];
235 uint32_t oldblks[EXT2_NDADDR + EXT2_NIADDR];
236 uint32_t newblks[EXT2_NDADDR + EXT2_NIADDR];
237 struct m_ext2fs *fs;
238 struct buf *bp;
239 int offset, size, level;
240 e4fs_daddr_t count, nblocks, blocksreleased = 0;
241 int error, i, allerror;
242 off_t osize;
243 #ifdef INVARIANTS
244 struct bufobj *bo;
245 #endif
246
247 oip = VTOI(ovp);
248 #ifdef INVARIANTS
249 bo = &ovp->v_bufobj;
250 #endif
251
252 fs = oip->i_e2fs;
253 osize = oip->i_size;
254 /*
255 * Lengthen the size of the file. We must ensure that the
256 * last byte of the file is allocated. Since the smallest
257 * value of osize is 0, length will be at least 1.
258 */
259 if (osize < length) {
260 if (length > oip->i_e2fs->e2fs_maxfilesize)
261 return (EFBIG);
262 vnode_pager_setsize(ovp, length);
263 offset = blkoff(fs, length - 1);
264 lbn = lblkno(fs, length - 1);
265 flags |= BA_CLRBUF;
266 error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags);
267 if (error) {
268 vnode_pager_setsize(vp, osize);
269 return (error);
270 }
271 oip->i_size = length;
272 if (bp->b_bufsize == fs->e2fs_bsize)
273 bp->b_flags |= B_CLUSTEROK;
274 if (flags & IO_SYNC)
275 bwrite(bp);
276 else if (DOINGASYNC(ovp))
277 bdwrite(bp);
278 else
279 bawrite(bp);
280 oip->i_flag |= IN_CHANGE | IN_UPDATE;
281 return (ext2_update(ovp, !DOINGASYNC(ovp)));
282 }
283 /*
284 * Shorten the size of the file. If the file is not being
285 * truncated to a block boundary, the contents of the
286 * partial block following the end of the file must be
287 * zero'ed in case it ever become accessible again because
288 * of subsequent file growth.
289 */
290 /* I don't understand the comment above */
291 offset = blkoff(fs, length);
292 if (offset == 0) {
293 oip->i_size = length;
294 } else {
295 lbn = lblkno(fs, length);
296 flags |= BA_CLRBUF;
297 error = ext2_balloc(oip, lbn, offset, cred, &bp, flags);
298 if (error)
299 return (error);
300 oip->i_size = length;
301 size = blksize(fs, oip, lbn);
302 bzero((char *)bp->b_data + offset, (u_int)(size - offset));
303 allocbuf(bp, size);
304 if (bp->b_bufsize == fs->e2fs_bsize)
305 bp->b_flags |= B_CLUSTEROK;
306 if (flags & IO_SYNC)
307 bwrite(bp);
308 else if (DOINGASYNC(ovp))
309 bdwrite(bp);
310 else
311 bawrite(bp);
312 }
313 /*
314 * Calculate index into inode's block list of
315 * last direct and indirect blocks (if any)
316 * which we want to keep. Lastblock is -1 when
317 * the file is truncated to 0.
318 */
319 lastblock = lblkno(fs, length + fs->e2fs_bsize - 1) - 1;
320 lastiblock[SINGLE] = lastblock - EXT2_NDADDR;
321 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
322 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
323 nblocks = btodb(fs->e2fs_bsize);
324 /*
325 * Update file and block pointers on disk before we start freeing
326 * blocks. If we crash before free'ing blocks below, the blocks
327 * will be returned to the free list. lastiblock values are also
328 * normalized to -1 for calls to ext2_indirtrunc below.
329 */
330 for (level = TRIPLE; level >= SINGLE; level--) {
331 oldblks[EXT2_NDADDR + level] = oip->i_ib[level];
332 if (lastiblock[level] < 0) {
333 oip->i_ib[level] = 0;
334 lastiblock[level] = -1;
335 }
336 }
337 for (i = 0; i < EXT2_NDADDR; i++) {
338 oldblks[i] = oip->i_db[i];
339 if (i > lastblock)
340 oip->i_db[i] = 0;
341 }
342 oip->i_flag |= IN_CHANGE | IN_UPDATE;
343 allerror = ext2_update(ovp, !DOINGASYNC(ovp));
344
345 /*
346 * Having written the new inode to disk, save its new configuration
347 * and put back the old block pointers long enough to process them.
348 * Note that we save the new block configuration so we can check it
349 * when we are done.
350 */
351 for (i = 0; i < EXT2_NDADDR; i++) {
352 newblks[i] = oip->i_db[i];
353 oip->i_db[i] = oldblks[i];
354 }
355 for (i = 0; i < EXT2_NIADDR; i++) {
356 newblks[EXT2_NDADDR + i] = oip->i_ib[i];
357 oip->i_ib[i] = oldblks[EXT2_NDADDR + i];
358 }
359 oip->i_size = osize;
360 error = vtruncbuf(ovp, length, (int)fs->e2fs_bsize);
361 if (error && (allerror == 0))
362 allerror = error;
363 vnode_pager_setsize(ovp, length);
364
365 /*
366 * Indirect blocks first.
367 */
368 indir_lbn[SINGLE] = -EXT2_NDADDR;
369 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
370 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
371 for (level = TRIPLE; level >= SINGLE; level--) {
372 bn = oip->i_ib[level];
373 if (bn != 0) {
374 error = ext2_indirtrunc(oip, indir_lbn[level],
375 fsbtodb(fs, bn), lastiblock[level], level, &count);
376 if (error)
377 allerror = error;
378 blocksreleased += count;
379 if (lastiblock[level] < 0) {
380 oip->i_ib[level] = 0;
381 ext2_blkfree(oip, bn, fs->e2fs_fsize);
382 blocksreleased += nblocks;
383 }
384 }
385 if (lastiblock[level] >= 0)
386 goto done;
387 }
388
389 /*
390 * All whole direct blocks or frags.
391 */
392 for (i = EXT2_NDADDR - 1; i > lastblock; i--) {
393 long bsize;
394
395 bn = oip->i_db[i];
396 if (bn == 0)
397 continue;
398 oip->i_db[i] = 0;
399 bsize = blksize(fs, oip, i);
400 ext2_blkfree(oip, bn, bsize);
401 blocksreleased += btodb(bsize);
402 }
403 if (lastblock < 0)
404 goto done;
405
406 /*
407 * Finally, look for a change in size of the
408 * last direct block; release any frags.
409 */
410 bn = oip->i_db[lastblock];
411 if (bn != 0) {
412 long oldspace, newspace;
413
414 /*
415 * Calculate amount of space we're giving
416 * back as old block size minus new block size.
417 */
418 oldspace = blksize(fs, oip, lastblock);
419 oip->i_size = length;
420 newspace = blksize(fs, oip, lastblock);
421 if (newspace == 0)
422 panic("ext2_truncate: newspace");
423 if (oldspace - newspace > 0) {
424 /*
425 * Block number of space to be free'd is
426 * the old block # plus the number of frags
427 * required for the storage we're keeping.
428 */
429 bn += numfrags(fs, newspace);
430 ext2_blkfree(oip, bn, oldspace - newspace);
431 blocksreleased += btodb(oldspace - newspace);
432 }
433 }
434 done:
435 #ifdef INVARIANTS
436 for (level = SINGLE; level <= TRIPLE; level++)
437 if (newblks[EXT2_NDADDR + level] != oip->i_ib[level])
438 panic("itrunc1");
439 for (i = 0; i < EXT2_NDADDR; i++)
440 if (newblks[i] != oip->i_db[i])
441 panic("itrunc2");
442 BO_LOCK(bo);
443 if (length == 0 && (bo->bo_dirty.bv_cnt != 0 ||
444 bo->bo_clean.bv_cnt != 0))
445 panic("itrunc3");
446 BO_UNLOCK(bo);
447 #endif /* INVARIANTS */
448 /*
449 * Put back the real size.
450 */
451 oip->i_size = length;
452 if (oip->i_blocks >= blocksreleased)
453 oip->i_blocks -= blocksreleased;
454 else /* sanity */
455 oip->i_blocks = 0;
456 oip->i_flag |= IN_CHANGE;
457 vnode_pager_setsize(ovp, length);
458 return (allerror);
459 }
460
461 static int
462 ext2_ext_truncate(struct vnode *vp, off_t length, int flags,
463 struct ucred *cred, struct thread *td)
464 {
465 struct vnode *ovp = vp;
466 int32_t lastblock;
467 struct m_ext2fs *fs;
468 struct inode *oip;
469 struct buf *bp;
470 uint32_t lbn, offset;
471 int error, size;
472 off_t osize;
473
474 oip = VTOI(ovp);
475 fs = oip->i_e2fs;
476 osize = oip->i_size;
477
478 if (osize < length) {
479 if (length > oip->i_e2fs->e2fs_maxfilesize) {
480 return (EFBIG);
481 }
482 vnode_pager_setsize(ovp, length);
483 offset = blkoff(fs, length - 1);
484 lbn = lblkno(fs, length - 1);
485 flags |= BA_CLRBUF;
486 error = ext2_balloc(oip, lbn, offset + 1, cred, &bp, flags);
487 if (error) {
488 vnode_pager_setsize(vp, osize);
489 return (error);
490 }
491 oip->i_size = length;
492 if (bp->b_bufsize == fs->e2fs_bsize)
493 bp->b_flags |= B_CLUSTEROK;
494 if (flags & IO_SYNC)
495 bwrite(bp);
496 else if (DOINGASYNC(ovp))
497 bdwrite(bp);
498 else
499 bawrite(bp);
500 oip->i_flag |= IN_CHANGE | IN_UPDATE;
501 return (ext2_update(ovp, !DOINGASYNC(ovp)));
502 }
503
504 lastblock = (length + fs->e2fs_bsize - 1) / fs->e2fs_bsize;
505 error = ext4_ext_remove_space(oip, lastblock, flags, cred, td);
506 if (error)
507 return (error);
508
509 offset = blkoff(fs, length);
510 if (offset == 0) {
511 oip->i_size = length;
512 } else {
513 lbn = lblkno(fs, length);
514 flags |= BA_CLRBUF;
515 error = ext2_balloc(oip, lbn, offset, cred, &bp, flags);
516 if (error) {
517 return (error);
518 }
519 oip->i_size = length;
520 size = blksize(fs, oip, lbn);
521 bzero((char *)bp->b_data + offset, (u_int)(size - offset));
522 allocbuf(bp, size);
523 if (bp->b_bufsize == fs->e2fs_bsize)
524 bp->b_flags |= B_CLUSTEROK;
525 if (flags & IO_SYNC)
526 bwrite(bp);
527 else if (DOINGASYNC(ovp))
528 bdwrite(bp);
529 else
530 bawrite(bp);
531 }
532
533 oip->i_size = osize;
534 error = vtruncbuf(ovp, length, (int)fs->e2fs_bsize);
535 if (error)
536 return (error);
537
538 vnode_pager_setsize(ovp, length);
539
540 oip->i_size = length;
541 oip->i_flag |= IN_CHANGE | IN_UPDATE;
542 error = ext2_update(ovp, !DOINGASYNC(ovp));
543
544 return (error);
545 }
546
547 /*
548 * Truncate the inode ip to at most length size, freeing the
549 * disk blocks.
550 */
551 int
552 ext2_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred,
553 struct thread *td)
554 {
555 struct inode *ip;
556 int error;
557
558 ASSERT_VOP_LOCKED(vp, "ext2_truncate");
559
560 if (length < 0)
561 return (EINVAL);
562
563 ip = VTOI(vp);
564 if (vp->v_type == VLNK &&
565 ip->i_size < vp->v_mount->mnt_maxsymlinklen) {
566 #ifdef INVARIANTS
567 if (length != 0)
568 panic("ext2_truncate: partial truncate of symlink");
569 #endif
570 bzero((char *)&ip->i_shortlink, (u_int)ip->i_size);
571 ip->i_size = 0;
572 ip->i_flag |= IN_CHANGE | IN_UPDATE;
573 return (ext2_update(vp, 1));
574 }
575 if (ip->i_size == length) {
576 ip->i_flag |= IN_CHANGE | IN_UPDATE;
577 return (ext2_update(vp, 0));
578 }
579
580 if (ip->i_flag & IN_E4EXTENTS)
581 error = ext2_ext_truncate(vp, length, flags, cred, td);
582 else
583 error = ext2_ind_truncate(vp, length, flags, cred, td);
584
585 return (error);
586 }
587
588 /*
589 * discard preallocated blocks
590 */
591 int
592 ext2_inactive(struct vop_inactive_args *ap)
593 {
594 struct vnode *vp = ap->a_vp;
595 struct inode *ip = VTOI(vp);
596 struct thread *td = curthread;
597 int mode, error = 0;
598
599 /*
600 * Ignore inodes related to stale file handles.
601 */
602 if (ip->i_mode == 0)
603 goto out;
604 if (ip->i_nlink <= 0) {
605 ext2_extattr_free(ip);
606 error = ext2_truncate(vp, (off_t)0, 0, NOCRED, td);
607 if (!(ip->i_flag & IN_E4EXTENTS))
608 ip->i_rdev = 0;
609 mode = ip->i_mode;
610 ip->i_mode = 0;
611 ip->i_flag |= IN_CHANGE | IN_UPDATE;
612 ext2_vfree(vp, ip->i_number, mode);
613 }
614 if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
615 ext2_update(vp, 0);
616 out:
617 /*
618 * If we are done with the inode, reclaim it
619 * so that it can be reused immediately.
620 */
621 if (ip->i_mode == 0)
622 vrecycle(vp);
623 return (error);
624 }
625
626 /*
627 * Reclaim an inode so that it can be used for other purposes.
628 */
629 int
630 ext2_reclaim(struct vop_reclaim_args *ap)
631 {
632 struct inode *ip;
633 struct vnode *vp = ap->a_vp;
634
635 ip = VTOI(vp);
636 if (ip->i_flag & IN_LAZYMOD) {
637 ip->i_flag |= IN_MODIFIED;
638 ext2_update(vp, 0);
639 }
640 vfs_hash_remove(vp);
641 free(vp->v_data, M_EXT2NODE);
642 vp->v_data = 0;
643 return (0);
644 }
Cache object: ac105807fbb1843800f97219e5268bd7
|