ffs_inode.c revision 1.24.2.1 1 /* $NetBSD: ffs_inode.c,v 1.24.2.1 1998/11/09 06:06:36 chs Exp $ */
2
3 /*
4 * Copyright (c) 1982, 1986, 1989, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. All advertising materials mentioning features or use of this software
16 * must display the following acknowledgement:
17 * This product includes software developed by the University of
18 * California, Berkeley and its contributors.
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 8.13 (Berkeley) 4/21/95
36 */
37
38 #if defined(_KERNEL) && !defined(_LKM)
39 #include "opt_quota.h"
40 #include "opt_uvm.h"
41 #endif
42
43 #include <sys/param.h>
44 #include <sys/systm.h>
45 #include <sys/mount.h>
46 #include <sys/proc.h>
47 #include <sys/file.h>
48 #include <sys/buf.h>
49 #include <sys/vnode.h>
50 #include <sys/kernel.h>
51 #include <sys/malloc.h>
52 #include <sys/trace.h>
53 #include <sys/resourcevar.h>
54
55 #include <vm/vm.h>
56
57 #if defined(UVM)
58 #include <uvm/uvm_extern.h>
59 #endif
60
61 #include <ufs/ufs/quota.h>
62 #include <ufs/ufs/inode.h>
63 #include <ufs/ufs/ufsmount.h>
64 #include <ufs/ufs/ufs_extern.h>
65 #include <ufs/ufs/ufs_bswap.h>
66
67 #include <ufs/ffs/fs.h>
68 #include <ufs/ffs/ffs_extern.h>
69
70 static int ffs_indirtrunc __P((struct inode *, ufs_daddr_t, ufs_daddr_t,
71 ufs_daddr_t, int, long *));
72
73 /*
74 * Update the access, modified, and inode change times as specified
75 * by the IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively.
76 * The IN_MODIFIED flag is used to specify that the inode needs to be
77 * updated but that the times have already been set. The access
78 * and modified times are taken from the second and third parameters;
79 * the inode change time is always taken from the current time. If
80 * waitfor is set, then wait for the disk write of the inode to
81 * complete.
82 */
83
84 int
85 ffs_update(v)
86 void *v;
87 {
88 struct vop_update_args /* {
89 struct vnode *a_vp;
90 struct timespec *a_access;
91 struct timespec *a_modify;
92 int a_waitfor;
93 } */ *ap = v;
94 register struct fs *fs;
95 struct buf *bp;
96 struct inode *ip;
97 int error;
98 struct timespec ts;
99 caddr_t cp;
100
101 if (ap->a_vp->v_mount->mnt_flag & MNT_RDONLY)
102 return (0);
103 ip = VTOI(ap->a_vp);
104 TIMEVAL_TO_TIMESPEC(&time, &ts);
105 FFS_ITIMES(ip, ap->a_access, ap->a_modify, &ts);
106 if ((ip->i_flag & IN_MODIFIED) == 0)
107 return (0);
108 ip->i_flag &= ~IN_MODIFIED;
109 fs = ip->i_fs;
110 /*
111 * Ensure that uid and gid are correct. This is a temporary
112 * fix until fsck has been changed to do the update.
113 */
114 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
115 ip->i_din.ffs_din.di_ouid = ip->i_ffs_uid; /* XXX */
116 ip->i_din.ffs_din.di_ogid = ip->i_ffs_gid; /* XXX */
117 } /* XXX */
118 error = bread(ip->i_devvp,
119 fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
120 (int)fs->fs_bsize, NOCRED, &bp);
121 if (error) {
122 brelse(bp);
123 return (error);
124 }
125 cp = (caddr_t)bp->b_data +
126 (ino_to_fsbo(fs, ip->i_number) * DINODE_SIZE);
127 #ifdef FFS_EI
128 if (UFS_MPNEEDSWAP(ap->a_vp->v_mount))
129 ffs_dinode_swap(&ip->i_din.ffs_din, (struct dinode *)cp);
130 else
131 #endif
132 memcpy(cp, &ip->i_din.ffs_din, DINODE_SIZE);
133 if (ap->a_waitfor && (ap->a_vp->v_mount->mnt_flag & MNT_ASYNC) == 0)
134 return (bwrite(bp));
135 else {
136 bdwrite(bp);
137 return (0);
138 }
139 }
140
141 #define SINGLE 0 /* index of single indirect block */
142 #define DOUBLE 1 /* index of double indirect block */
143 #define TRIPLE 2 /* index of triple indirect block */
144 /*
145 * Truncate the inode oip to at most length size, freeing the
146 * disk blocks.
147 */
148 int
149 ffs_truncate(v)
150 void *v;
151 {
152 struct vop_truncate_args /* {
153 struct vnode *a_vp;
154 off_t a_length;
155 int a_flags;
156 struct ucred *a_cred;
157 struct proc *a_p;
158 } */ *ap = v;
159 register struct vnode *ovp = ap->a_vp;
160 register ufs_daddr_t lastblock;
161 register struct inode *oip;
162 ufs_daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
163 ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
164 off_t length = ap->a_length;
165 register struct fs *fs;
166 struct buf *bp;
167 int offset, size, level;
168 long count, nblocks, vflags, blocksreleased = 0;
169 struct timespec ts;
170 register int i;
171 int aflags, error, allerror;
172 off_t osize;
173
174 if (length < 0)
175 return (EINVAL);
176 oip = VTOI(ovp);
177 TIMEVAL_TO_TIMESPEC(&time, &ts);
178 if (ovp->v_type == VLNK &&
179 (oip->i_ffs_size < ovp->v_mount->mnt_maxsymlinklen ||
180 (ovp->v_mount->mnt_maxsymlinklen == 0 &&
181 oip->i_din.ffs_din.di_blocks == 0))) {
182 #ifdef DIAGNOSTIC
183 if (length != 0)
184 panic("ffs_truncate: partial truncate of symlink");
185 #endif
186 memset((char *)&oip->i_ffs_shortlink, 0, (u_int)oip->i_ffs_size);
187 oip->i_ffs_size = 0;
188 oip->i_flag |= IN_CHANGE | IN_UPDATE;
189 return (VOP_UPDATE(ovp, &ts, &ts, 1));
190 }
191 if (oip->i_ffs_size == length) {
192 oip->i_flag |= IN_CHANGE | IN_UPDATE;
193 return (VOP_UPDATE(ovp, &ts, &ts, 0));
194 }
195 #ifdef QUOTA
196 if ((error = getinoquota(oip)) != 0)
197 return (error);
198 #endif
199 fs = oip->i_fs;
200 osize = oip->i_ffs_size;
201 ovp->v_lasta = ovp->v_clen = ovp->v_cstart = ovp->v_lastw = 0;
202 /*
203 * Lengthen the size of the file. We must ensure that the
204 * last byte of the file is allocated. Since the smallest
205 * value of osize is 0, length will be at least 1.
206 */
207 if (osize < length) {
208 if (length > fs->fs_maxfilesize)
209 return (EFBIG);
210 offset = blkoff(fs, length - 1);
211 lbn = lblkno(fs, length - 1);
212 aflags = B_CLRBUF;
213 if (ap->a_flags & IO_SYNC)
214 aflags |= B_SYNC;
215 error = ffs_balloc(oip, lbn, offset + 1, ap->a_cred, &bp,
216 aflags);
217 if (error)
218 return (error);
219 oip->i_ffs_size = length;
220 #if defined(UVM)
221 uvm_vnp_setsize(ovp, length);
222 (void) uvm_vnp_uncache(ovp);
223 #else
224 vnode_pager_setsize(ovp, length);
225 (void) vnode_pager_uncache(ovp);
226 #endif
227 #ifdef UBC
228 #else
229 if (aflags & B_SYNC)
230 bwrite(bp);
231 else
232 bawrite(bp);
233 #endif
234 oip->i_flag |= IN_CHANGE | IN_UPDATE;
235 return (VOP_UPDATE(ovp, &ts, &ts, 1));
236 }
237 /*
238 * Shorten the size of the file. If the file is not being
239 * truncated to a block boundry, the contents of the
240 * partial block following the end of the file must be
241 * zero'ed in case it ever become accessable again because
242 * of subsequent file growth.
243 */
244 offset = blkoff(fs, length);
245 if (offset == 0) {
246 oip->i_ffs_size = length;
247 } else {
248 lbn = lblkno(fs, length);
249 aflags = B_CLRBUF;
250 if (ap->a_flags & IO_SYNC)
251 aflags |= B_SYNC;
252 error = ffs_balloc(oip, lbn, offset, ap->a_cred, &bp, aflags);
253 if (error)
254 return (error);
255 oip->i_ffs_size = length;
256 size = blksize(fs, oip, lbn);
257 #if defined(UVM)
258 (void) uvm_vnp_uncache(ovp);
259 #else
260 (void) vnode_pager_uncache(ovp);
261 #endif
262 memset((char *)bp->b_data + offset, 0, (u_int)(size - offset));
263 allocbuf(bp, size);
264 if (aflags & B_SYNC)
265 bwrite(bp);
266 else
267 bawrite(bp);
268 }
269 #if defined(UVM)
270 uvm_vnp_setsize(ovp, length);
271 #else
272 vnode_pager_setsize(ovp, length);
273 #endif
274 /*
275 * Calculate index into inode's block list of
276 * last direct and indirect blocks (if any)
277 * which we want to keep. Lastblock is -1 when
278 * the file is truncated to 0.
279 */
280 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
281 lastiblock[SINGLE] = lastblock - NDADDR;
282 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
283 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
284 nblocks = btodb(fs->fs_bsize);
285 /*
286 * Update file and block pointers on disk before we start freeing
287 * blocks. If we crash before free'ing blocks below, the blocks
288 * will be returned to the free list. lastiblock values are also
289 * normalized to -1 for calls to ffs_indirtrunc below.
290 */
291 memcpy((caddr_t)oldblks, (caddr_t)&oip->i_ffs_db[0], sizeof oldblks);
292 for (level = TRIPLE; level >= SINGLE; level--)
293 if (lastiblock[level] < 0) {
294 oip->i_ffs_ib[level] = 0;
295 lastiblock[level] = -1;
296 }
297 for (i = NDADDR - 1; i > lastblock; i--)
298 oip->i_ffs_db[i] = 0;
299 oip->i_flag |= IN_CHANGE | IN_UPDATE;
300 if ((error = VOP_UPDATE(ovp, &ts, &ts, 1)) != 0)
301 allerror = error;
302 /*
303 * Having written the new inode to disk, save its new configuration
304 * and put back the old block pointers long enough to process them.
305 * Note that we save the new block configuration so we can check it
306 * when we are done.
307 */
308 memcpy((caddr_t)newblks, (caddr_t)&oip->i_ffs_db[0], sizeof newblks);
309 memcpy((caddr_t)&oip->i_ffs_db[0], (caddr_t)oldblks, sizeof oldblks);
310 oip->i_ffs_size = osize;
311 vflags = ((length > 0) ? V_SAVE : 0) | V_SAVEMETA;
312 allerror = vinvalbuf(ovp, vflags, ap->a_cred, ap->a_p, 0, 0);
313
314 /*
315 * Indirect blocks first.
316 */
317 indir_lbn[SINGLE] = -NDADDR;
318 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
319 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
320 for (level = TRIPLE; level >= SINGLE; level--) {
321 bn = ufs_rw32(oip->i_ffs_ib[level], UFS_MPNEEDSWAP(ovp->v_mount));
322 if (bn != 0) {
323 error = ffs_indirtrunc(oip, indir_lbn[level],
324 fsbtodb(fs, bn), lastiblock[level], level, &count);
325 if (error)
326 allerror = error;
327 blocksreleased += count;
328 if (lastiblock[level] < 0) {
329 oip->i_ffs_ib[level] = 0;
330 ffs_blkfree(oip, bn, fs->fs_bsize);
331 blocksreleased += nblocks;
332 }
333 }
334 if (lastiblock[level] >= 0)
335 goto done;
336 }
337
338 /*
339 * All whole direct blocks or frags.
340 */
341 for (i = NDADDR - 1; i > lastblock; i--) {
342 register long bsize;
343
344 bn = ufs_rw32(oip->i_ffs_db[i], UFS_MPNEEDSWAP(ovp->v_mount));
345 if (bn == 0)
346 continue;
347 oip->i_ffs_db[i] = 0;
348 bsize = blksize(fs, oip, i);
349 ffs_blkfree(oip, bn, bsize);
350 blocksreleased += btodb(bsize);
351 }
352 if (lastblock < 0)
353 goto done;
354
355 /*
356 * Finally, look for a change in size of the
357 * last direct block; release any frags.
358 */
359 bn = ufs_rw32(oip->i_ffs_db[lastblock], UFS_MPNEEDSWAP(ovp->v_mount));
360 if (bn != 0) {
361 long oldspace, newspace;
362
363 /*
364 * Calculate amount of space we're giving
365 * back as old block size minus new block size.
366 */
367 oldspace = blksize(fs, oip, lastblock);
368 oip->i_ffs_size = length;
369 newspace = blksize(fs, oip, lastblock);
370 if (newspace == 0)
371 panic("itrunc: newspace");
372 if (oldspace - newspace > 0) {
373 /*
374 * Block number of space to be free'd is
375 * the old block # plus the number of frags
376 * required for the storage we're keeping.
377 */
378 bn += numfrags(fs, newspace);
379 ffs_blkfree(oip, bn, oldspace - newspace);
380 blocksreleased += btodb(oldspace - newspace);
381 }
382 }
383 done:
384 #ifdef DIAGNOSTIC
385 for (level = SINGLE; level <= TRIPLE; level++)
386 if (newblks[NDADDR + level] != oip->i_ffs_ib[level])
387 panic("itrunc1");
388 for (i = 0; i < NDADDR; i++)
389 if (newblks[i] != oip->i_ffs_db[i])
390 panic("itrunc2");
391 if (length == 0 &&
392 (ovp->v_dirtyblkhd.lh_first || ovp->v_cleanblkhd.lh_first))
393 panic("itrunc3");
394 #endif /* DIAGNOSTIC */
395 /*
396 * Put back the real size.
397 */
398 oip->i_ffs_size = length;
399 oip->i_ffs_blocks -= blocksreleased;
400 if (oip->i_ffs_blocks < 0) /* sanity */
401 oip->i_ffs_blocks = 0;
402 oip->i_flag |= IN_CHANGE;
403 #ifdef QUOTA
404 (void) chkdq(oip, -blocksreleased, NOCRED, 0);
405 #endif
406 return (allerror);
407 }
408
409 /*
410 * Release blocks associated with the inode ip and stored in the indirect
411 * block bn. Blocks are free'd in LIFO order up to (but not including)
412 * lastbn. If level is greater than SINGLE, the block is an indirect block
413 * and recursive calls to indirtrunc must be used to cleanse other indirect
414 * blocks.
415 *
416 * NB: triple indirect blocks are untested.
417 */
418 static int
419 ffs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
420 register struct inode *ip;
421 ufs_daddr_t lbn, lastbn;
422 ufs_daddr_t dbn;
423 int level;
424 long *countp;
425 {
426 register int i;
427 struct buf *bp;
428 register struct fs *fs = ip->i_fs;
429 register ufs_daddr_t *bap;
430 struct vnode *vp;
431 ufs_daddr_t *copy = NULL, nb, nlbn, last;
432 long blkcount, factor;
433 int nblocks, blocksreleased = 0;
434 int error = 0, allerror = 0;
435
436 /*
437 * Calculate index in current block of last
438 * block to be kept. -1 indicates the entire
439 * block so we need not calculate the index.
440 */
441 factor = 1;
442 for (i = SINGLE; i < level; i++)
443 factor *= NINDIR(fs);
444 last = lastbn;
445 if (lastbn > 0)
446 last /= factor;
447 nblocks = btodb(fs->fs_bsize);
448 /*
449 * Get buffer of block pointers, zero those entries corresponding
450 * to blocks to be free'd, and update on disk copy first. Since
451 * double(triple) indirect before single(double) indirect, calls
452 * to bmap on these blocks will fail. However, we already have
453 * the on disk address, so we have to set the b_blkno field
454 * explicitly instead of letting bread do everything for us.
455 */
456 vp = ITOV(ip);
457 bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0);
458 if (bp->b_flags & (B_DONE | B_DELWRI)) {
459 /* Braces must be here in case trace evaluates to nothing. */
460 trace(TR_BREADHIT, pack(vp, fs->fs_bsize), lbn);
461 } else {
462 trace(TR_BREADMISS, pack(vp, fs->fs_bsize), lbn);
463 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */
464 bp->b_flags |= B_READ;
465 if (bp->b_bcount > bp->b_bufsize)
466 panic("ffs_indirtrunc: bad buffer size");
467 bp->b_blkno = dbn;
468 VOP_STRATEGY(bp);
469 error = biowait(bp);
470 }
471 if (error) {
472 brelse(bp);
473 *countp = 0;
474 return (error);
475 }
476
477 bap = (ufs_daddr_t *)bp->b_data;
478 if (lastbn != -1) {
479 MALLOC(copy, ufs_daddr_t *, fs->fs_bsize, M_TEMP, M_WAITOK);
480 memcpy((caddr_t)copy, (caddr_t)bap, (u_int)fs->fs_bsize);
481 memset((caddr_t)&bap[last + 1], 0,
482 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (ufs_daddr_t));
483 error = bwrite(bp);
484 if (error)
485 allerror = error;
486 bap = copy;
487 }
488
489 /*
490 * Recursively free totally unused blocks.
491 */
492 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
493 i--, nlbn += factor) {
494 nb = ufs_rw32(bap[i], UFS_MPNEEDSWAP(vp->v_mount));
495 if (nb == 0)
496 continue;
497 if (level > SINGLE) {
498 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
499 (ufs_daddr_t)-1, level - 1,
500 &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 = ufs_rw32(bap[i], UFS_MPNEEDSWAP(vp->v_mount));
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
524 if (copy != NULL) {
525 FREE(copy, M_TEMP);
526 } else {
527 bp->b_flags |= B_INVAL;
528 brelse(bp);
529 }
530
531 *countp = blocksreleased;
532 return (allerror);
533 }
534