1 /* $OpenBSD: ffs_inode.c,v 1.83 2024/02/03 18:51:58 beck Exp $ */
2 /* $NetBSD: ffs_inode.c,v 1.10 1996/05/11 18:27:19 mycroft Exp $ */
3
4 /*
5 * Copyright (c) 1982, 1986, 1989, 1993
6 * The Regents of the University of California. All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions
10 * are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the name of the University nor the names of its contributors
17 * may be used to endorse or promote products derived from this software
18 * without specific prior written permission.
19 *
20 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
30 * SUCH DAMAGE.
31 *
32 * @(#)ffs_inode.c 8.8 (Berkeley) 10/19/94
33 */
34
35 #include <sys/param.h>
36 #include <sys/systm.h>
37 #include <sys/mount.h>
38 #include <sys/proc.h>
39 #include <sys/buf.h>
40 #include <sys/vnode.h>
41 #include <sys/kernel.h>
42 #include <sys/malloc.h>
43 #include <sys/resourcevar.h>
44
45 #include <ufs/ufs/quota.h>
46 #include <ufs/ufs/inode.h>
47 #include <ufs/ufs/ufsmount.h>
48 #include <ufs/ufs/ufs_extern.h>
49
50 #include <ufs/ffs/fs.h>
51 #include <ufs/ffs/ffs_extern.h>
52
53 int ffs_indirtrunc(struct inode *, daddr_t, daddr_t, daddr_t, int, long *);
54
55 /*
56 * Update the access, modified, and inode change times as specified by the
57 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. The IN_MODIFIED
58 * flag is used to specify that the inode needs to be updated but that the
59 * times have already been set. The IN_LAZYMOD flag is used to specify
60 * that the inode needs to be updated at some point, by reclaim if not
61 * in the course of other changes; this is used to defer writes just to
62 * update device timestamps. If waitfor is set, then wait for the disk
63 * write of the inode to complete.
64 */
65 int
ffs_update(struct inode * ip,int waitfor)66 ffs_update(struct inode *ip, int waitfor)
67 {
68 struct vnode *vp;
69 struct fs *fs;
70 struct buf *bp;
71 int error;
72
73 vp = ITOV(ip);
74 ufs_itimes(vp);
75
76 if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
77 return (0);
78
79 ip->i_flag &= ~(IN_MODIFIED | IN_LAZYMOD);
80 fs = ip->i_fs;
81
82 /*
83 * Ensure that uid and gid are correct. This is a temporary
84 * fix until fsck has been changed to do the update.
85 */
86 if (fs->fs_magic == FS_UFS1_MAGIC && fs->fs_inodefmt < FS_44INODEFMT) {
87 ip->i_din1->di_ouid = ip->i_ffs1_uid;
88 ip->i_din1->di_ogid = ip->i_ffs1_gid;
89 }
90
91 error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
92 (int)fs->fs_bsize, &bp);
93 if (error) {
94 brelse(bp);
95 return (error);
96 }
97
98 if (ip->i_effnlink != DIP(ip, nlink))
99 panic("ffs_update: bad link cnt");
100
101 #ifdef FFS2
102 if (ip->i_ump->um_fstype == UM_UFS2)
103 *((struct ufs2_dinode *)bp->b_data +
104 ino_to_fsbo(fs, ip->i_number)) = *ip->i_din2;
105 else
106 #endif
107 *((struct ufs1_dinode *)bp->b_data +
108 ino_to_fsbo(fs, ip->i_number)) = *ip->i_din1;
109
110 if (waitfor && !DOINGASYNC(vp)) {
111 return (bwrite(bp));
112 } else {
113 bdwrite(bp);
114 return (0);
115 }
116 }
117
118 #define SINGLE 0 /* index of single indirect block */
119 #define DOUBLE 1 /* index of double indirect block */
120 #define TRIPLE 2 /* index of triple indirect block */
121
122 /*
123 * Truncate the inode oip to at most length size, freeing the
124 * disk blocks.
125 */
126 int
ffs_truncate(struct inode * oip,off_t length,int flags,struct ucred * cred)127 ffs_truncate(struct inode *oip, off_t length, int flags, struct ucred *cred)
128 {
129 struct vnode *ovp;
130 daddr_t lastblock;
131 daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
132 daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
133 struct fs *fs;
134 struct buf *bp;
135 int offset, size, level;
136 long count, nblocks, vflags, blocksreleased = 0;
137 int i, aflags, error, allerror;
138 off_t osize;
139
140 if (length < 0)
141 return (EINVAL);
142 ovp = ITOV(oip);
143
144 if (ovp->v_type != VREG &&
145 ovp->v_type != VDIR &&
146 ovp->v_type != VLNK)
147 return (0);
148
149 if (DIP(oip, size) == length)
150 return (0);
151
152 if (ovp->v_type == VLNK &&
153 DIP(oip, size) < oip->i_ump->um_maxsymlinklen) {
154 #ifdef DIAGNOSTIC
155 if (length != 0)
156 panic("ffs_truncate: partial truncate of symlink");
157 #endif
158 memset(SHORTLINK(oip), 0, (size_t) DIP(oip, size));
159 DIP_ASSIGN(oip, size, 0);
160 oip->i_flag |= IN_CHANGE | IN_UPDATE;
161 return (UFS_UPDATE(oip, 1));
162 }
163
164 if ((error = getinoquota(oip)) != 0)
165 return (error);
166
167 fs = oip->i_fs;
168 if (length > fs->fs_maxfilesize)
169 return (EFBIG);
170
171 uvm_vnp_setsize(ovp, length);
172 oip->i_ci.ci_lasta = oip->i_ci.ci_clen
173 = oip->i_ci.ci_cstart = oip->i_ci.ci_lastw = 0;
174
175 osize = DIP(oip, size);
176 /*
177 * Lengthen the size of the file. We must ensure that the
178 * last byte of the file is allocated. Since the smallest
179 * value of osize is 0, length will be at least 1.
180 */
181 if (osize < length) {
182 aflags = B_CLRBUF;
183 if (flags & IO_SYNC)
184 aflags |= B_SYNC;
185 error = UFS_BUF_ALLOC(oip, length - 1, 1,
186 cred, aflags, &bp);
187 if (error)
188 return (error);
189 DIP_ASSIGN(oip, size, length);
190 uvm_vnp_setsize(ovp, length);
191 (void) uvm_vnp_uncache(ovp);
192 if (aflags & B_SYNC)
193 bwrite(bp);
194 else
195 bawrite(bp);
196 oip->i_flag |= IN_CHANGE | IN_UPDATE;
197 return (UFS_UPDATE(oip, 1));
198 }
199 uvm_vnp_setsize(ovp, length);
200
201 /*
202 * Shorten the size of the file. If the file is not being
203 * truncated to a block boundary, the contents of the
204 * partial block following the end of the file must be
205 * zero'ed in case it ever becomes accessible again because
206 * of subsequent file growth. Directories however are not
207 * zero'ed as they should grow back initialized to empty.
208 */
209 offset = blkoff(fs, length);
210 if (offset == 0) {
211 DIP_ASSIGN(oip, size, length);
212 } else {
213 lbn = lblkno(fs, length);
214 aflags = B_CLRBUF;
215 if (flags & IO_SYNC)
216 aflags |= B_SYNC;
217 error = UFS_BUF_ALLOC(oip, length - 1, 1,
218 cred, aflags, &bp);
219 if (error)
220 return (error);
221 DIP_ASSIGN(oip, size, length);
222 size = blksize(fs, oip, lbn);
223 (void) uvm_vnp_uncache(ovp);
224 if (ovp->v_type != VDIR)
225 memset(bp->b_data + offset, 0, size - offset);
226 buf_adjcnt(bp, size);
227 if (aflags & B_SYNC)
228 bwrite(bp);
229 else
230 bawrite(bp);
231 }
232 /*
233 * Calculate index into inode's block list of
234 * last direct and indirect blocks (if any)
235 * which we want to keep. Lastblock is -1 when
236 * the file is truncated to 0.
237 */
238 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
239 lastiblock[SINGLE] = lastblock - NDADDR;
240 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
241 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
242 nblocks = btodb(fs->fs_bsize);
243
244 /*
245 * Update file and block pointers on disk before we start freeing
246 * blocks. If we crash before free'ing blocks below, the blocks
247 * will be returned to the free list. lastiblock values are also
248 * normalized to -1 for calls to ffs_indirtrunc below.
249 */
250 for (level = TRIPLE; level >= SINGLE; level--) {
251 oldblks[NDADDR + level] = DIP(oip, ib[level]);
252 if (lastiblock[level] < 0) {
253 DIP_ASSIGN(oip, ib[level], 0);
254 lastiblock[level] = -1;
255 }
256 }
257
258 for (i = 0; i < NDADDR; i++) {
259 oldblks[i] = DIP(oip, db[i]);
260 if (i > lastblock)
261 DIP_ASSIGN(oip, db[i], 0);
262 }
263
264 oip->i_flag |= IN_CHANGE | IN_UPDATE;
265 if ((error = UFS_UPDATE(oip, 1)) != 0)
266 allerror = error;
267
268 /*
269 * Having written the new inode to disk, save its new configuration
270 * and put back the old block pointers long enough to process them.
271 * Note that we save the new block configuration so we can check it
272 * when we are done.
273 */
274 for (i = 0; i < NDADDR; i++) {
275 newblks[i] = DIP(oip, db[i]);
276 DIP_ASSIGN(oip, db[i], oldblks[i]);
277 }
278
279 for (i = 0; i < NIADDR; i++) {
280 newblks[NDADDR + i] = DIP(oip, ib[i]);
281 DIP_ASSIGN(oip, ib[i], oldblks[NDADDR + i]);
282 }
283
284 DIP_ASSIGN(oip, size, osize);
285 vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
286 allerror = vinvalbuf(ovp, vflags, cred, curproc, 0, INFSLP);
287
288 /*
289 * Indirect blocks first.
290 */
291 indir_lbn[SINGLE] = -NDADDR;
292 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
293 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
294 for (level = TRIPLE; level >= SINGLE; level--) {
295 bn = DIP(oip, ib[level]);
296 if (bn != 0) {
297 error = ffs_indirtrunc(oip, indir_lbn[level],
298 fsbtodb(fs, bn), lastiblock[level], level, &count);
299 if (error)
300 allerror = error;
301 blocksreleased += count;
302 if (lastiblock[level] < 0) {
303 DIP_ASSIGN(oip, ib[level], 0);
304 ffs_blkfree(oip, bn, fs->fs_bsize);
305 blocksreleased += nblocks;
306 }
307 }
308 if (lastiblock[level] >= 0)
309 goto done;
310 }
311
312 /*
313 * All whole direct blocks or frags.
314 */
315 for (i = NDADDR - 1; i > lastblock; i--) {
316 long bsize;
317
318 bn = DIP(oip, db[i]);
319 if (bn == 0)
320 continue;
321
322 DIP_ASSIGN(oip, db[i], 0);
323 bsize = blksize(fs, oip, i);
324 ffs_blkfree(oip, bn, bsize);
325 blocksreleased += btodb(bsize);
326 }
327 if (lastblock < 0)
328 goto done;
329
330 /*
331 * Finally, look for a change in size of the
332 * last direct block; release any frags.
333 */
334 bn = DIP(oip, db[lastblock]);
335 if (bn != 0) {
336 long oldspace, newspace;
337
338 /*
339 * Calculate amount of space we're giving
340 * back as old block size minus new block size.
341 */
342 oldspace = blksize(fs, oip, lastblock);
343 DIP_ASSIGN(oip, size, length);
344 newspace = blksize(fs, oip, lastblock);
345 if (newspace == 0)
346 panic("ffs_truncate: newspace");
347 if (oldspace - newspace > 0) {
348 /*
349 * Block number of space to be free'd is
350 * the old block # plus the number of frags
351 * required for the storage we're keeping.
352 */
353 bn += numfrags(fs, newspace);
354 ffs_blkfree(oip, bn, oldspace - newspace);
355 blocksreleased += btodb(oldspace - newspace);
356 }
357 }
358 done:
359 #ifdef DIAGNOSTIC
360 for (level = SINGLE; level <= TRIPLE; level++)
361 if (newblks[NDADDR + level] != DIP(oip, ib[level]))
362 panic("ffs_truncate1");
363 for (i = 0; i < NDADDR; i++)
364 if (newblks[i] != DIP(oip, db[i]))
365 panic("ffs_truncate2");
366 #endif /* DIAGNOSTIC */
367 /*
368 * Put back the real size.
369 */
370 DIP_ASSIGN(oip, size, length);
371 if (DIP(oip, blocks) >= blocksreleased)
372 DIP_ADD(oip, blocks, -blocksreleased);
373 else /* sanity */
374 DIP_ASSIGN(oip, blocks, 0);
375 oip->i_flag |= IN_CHANGE;
376 (void)ufs_quota_free_blocks(oip, blocksreleased, NOCRED);
377 return (allerror);
378 }
379
380 #ifdef FFS2
381 #define BAP(ip, i) (((ip)->i_ump->um_fstype == UM_UFS2) ? bap2[i] : bap1[i])
382 #define BAP_ASSIGN(ip, i, value) \
383 do { \
384 if ((ip)->i_ump->um_fstype == UM_UFS2) \
385 bap2[i] = (value); \
386 else \
387 bap1[i] = (value); \
388 } while (0)
389 #else
390 #define BAP(ip, i) bap1[i]
391 #define BAP_ASSIGN(ip, i, value) do { bap1[i] = (value); } while (0)
392 #endif /* FFS2 */
393
394 /*
395 * Release blocks associated with the inode ip and stored in the indirect
396 * block bn. Blocks are free'd in LIFO order up to (but not including)
397 * lastbn. If level is greater than SINGLE, the block is an indirect block
398 * and recursive calls to indirtrunc must be used to cleanse other indirect
399 * blocks.
400 *
401 * NB: triple indirect blocks are untested.
402 */
403 int
ffs_indirtrunc(struct inode * ip,daddr_t lbn,daddr_t dbn,daddr_t lastbn,int level,long * countp)404 ffs_indirtrunc(struct inode *ip, daddr_t lbn, daddr_t dbn,
405 daddr_t lastbn, int level, long *countp)
406 {
407 int i;
408 struct buf *bp;
409 struct fs *fs = ip->i_fs;
410 struct vnode *vp;
411 void *copy = NULL;
412 daddr_t nb, nlbn, last;
413 long blkcount, factor;
414 int nblocks, blocksreleased = 0;
415 int error = 0, allerror = 0;
416 int32_t *bap1 = NULL;
417 #ifdef FFS2
418 int64_t *bap2 = NULL;
419 #endif
420
421 /*
422 * Calculate index in current block of last
423 * block to be kept. -1 indicates the entire
424 * block so we need not calculate the index.
425 */
426 factor = 1;
427 for (i = SINGLE; i < level; i++)
428 factor *= NINDIR(fs);
429 last = lastbn;
430 if (lastbn > 0)
431 last /= factor;
432 nblocks = btodb(fs->fs_bsize);
433 /*
434 * Get buffer of block pointers, zero those entries corresponding
435 * to blocks to be free'd, and update on disk copy first. Since
436 * double(triple) indirect before single(double) indirect, calls
437 * to bmap on these blocks will fail. However, we already have
438 * the on disk address, so we have to set the b_blkno field
439 * explicitly instead of letting bread do everything for us.
440 */
441 vp = ITOV(ip);
442 bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, INFSLP);
443 if (!(bp->b_flags & (B_DONE | B_DELWRI))) {
444 curproc->p_ru.ru_inblock++; /* pay for read */
445 bcstats.pendingreads++;
446 bcstats.numreads++;
447 bp->b_flags |= B_READ;
448 if (bp->b_bcount > bp->b_bufsize)
449 panic("ffs_indirtrunc: bad buffer size");
450 bp->b_blkno = dbn;
451 VOP_STRATEGY(bp->b_vp, bp);
452 error = biowait(bp);
453 }
454 if (error) {
455 brelse(bp);
456 *countp = 0;
457 return (error);
458 }
459
460 #ifdef FFS2
461 if (ip->i_ump->um_fstype == UM_UFS2)
462 bap2 = (int64_t *)bp->b_data;
463 else
464 #endif
465 bap1 = (int32_t *)bp->b_data;
466
467 if (lastbn != -1) {
468 copy = malloc(fs->fs_bsize, M_TEMP, M_WAITOK);
469 memcpy(copy, bp->b_data, fs->fs_bsize);
470
471 for (i = last + 1; i < NINDIR(fs); i++)
472 BAP_ASSIGN(ip, i, 0);
473
474 if (!DOINGASYNC(vp)) {
475 error = bwrite(bp);
476 if (error)
477 allerror = error;
478 } else {
479 bawrite(bp);
480 }
481
482 #ifdef FFS2
483 if (ip->i_ump->um_fstype == UM_UFS2)
484 bap2 = (int64_t *)copy;
485 else
486 #endif
487 bap1 = (int32_t *)copy;
488 }
489
490 /*
491 * Recursively free totally unused blocks.
492 */
493 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
494 i--, nlbn += factor) {
495 nb = BAP(ip, i);
496 if (nb == 0)
497 continue;
498 if (level > SINGLE) {
499 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
500 -1, level - 1, &blkcount);
501 if (error)
502 allerror = error;
503 blocksreleased += blkcount;
504 }
505 ffs_blkfree(ip, nb, fs->fs_bsize);
506 blocksreleased += nblocks;
507 }
508
509 /*
510 * Recursively free last partial block.
511 */
512 if (level > SINGLE && lastbn >= 0) {
513 last = lastbn % factor;
514 nb = BAP(ip, i);
515 if (nb != 0) {
516 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
517 last, level - 1, &blkcount);
518 if (error)
519 allerror = error;
520 blocksreleased += blkcount;
521 }
522 }
523 if (copy != NULL) {
524 free(copy, M_TEMP, fs->fs_bsize);
525 } else {
526 bp->b_flags |= B_INVAL;
527 brelse(bp);
528 }
529
530 *countp = blocksreleased;
531 return (allerror);
532 }
533