Initial import from FreeBSD RELENG_4:
[dragonfly.git] / sys / vfs / ufs / ffs_inode.c
CommitLineData
984263bc
MD
1/*
2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
20 *
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31 * SUCH DAMAGE.
32 *
33 * @(#)ffs_inode.c 8.13 (Berkeley) 4/21/95
34 * $FreeBSD: src/sys/ufs/ffs/ffs_inode.c,v 1.56.2.5 2002/02/05 18:35:03 dillon Exp $
35 */
36
37#include "opt_quota.h"
38
39#include <sys/param.h>
40#include <sys/systm.h>
41#include <sys/mount.h>
42#include <sys/proc.h>
43#include <sys/buf.h>
44#include <sys/vnode.h>
45#include <sys/kernel.h>
46#include <sys/malloc.h>
47#include <sys/resourcevar.h>
48#include <sys/vmmeter.h>
49
50#include <vm/vm.h>
51#include <vm/vm_extern.h>
52
53#include <ufs/ufs/quota.h>
54#include <ufs/ufs/ufsmount.h>
55#include <ufs/ufs/inode.h>
56#include <ufs/ufs/ufs_extern.h>
57
58#include <ufs/ffs/fs.h>
59#include <ufs/ffs/ffs_extern.h>
60
61static int ffs_indirtrunc __P((struct inode *, ufs_daddr_t, ufs_daddr_t,
62 ufs_daddr_t, int, long *));
63
64/*
65 * Update the access, modified, and inode change times as specified by the
66 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode
67 * to disk if the IN_MODIFIED flag is set (it may be set initially, or by
68 * the timestamp update). The IN_LAZYMOD flag is set to force a write
69 * later if not now. If we write now, then clear both IN_MODIFIED and
70 * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is
71 * set, then wait for the write to complete.
72 */
73int
74ffs_update(vp, waitfor)
75 struct vnode *vp;
76 int waitfor;
77{
78 register struct fs *fs;
79 struct buf *bp;
80 struct inode *ip;
81 int error;
82
83 ufs_itimes(vp);
84 ip = VTOI(vp);
85 if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
86 return (0);
87 ip->i_flag &= ~(IN_LAZYMOD | IN_MODIFIED);
88 fs = ip->i_fs;
89 if (fs->fs_ronly)
90 return (0);
91 /*
92 * Ensure that uid and gid are correct. This is a temporary
93 * fix until fsck has been changed to do the update.
94 */
95 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
96 ip->i_din.di_ouid = ip->i_uid; /* XXX */
97 ip->i_din.di_ogid = ip->i_gid; /* XXX */
98 } /* XXX */
99 error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
100 (int)fs->fs_bsize, NOCRED, &bp);
101 if (error) {
102 brelse(bp);
103 return (error);
104 }
105 if (DOINGSOFTDEP(vp))
106 softdep_update_inodeblock(ip, bp, waitfor);
107 else if (ip->i_effnlink != ip->i_nlink)
108 panic("ffs_update: bad link cnt");
109 *((struct dinode *)bp->b_data +
110 ino_to_fsbo(fs, ip->i_number)) = ip->i_din;
111 if (waitfor && !DOINGASYNC(vp)) {
112 return (bwrite(bp));
113 } else if (vm_page_count_severe() || buf_dirty_count_severe()) {
114 return (bwrite(bp));
115 } else {
116 if (bp->b_bufsize == fs->fs_bsize)
117 bp->b_flags |= B_CLUSTEROK;
118 bdwrite(bp);
119 return (0);
120 }
121}
122
123#define SINGLE 0 /* index of single indirect block */
124#define DOUBLE 1 /* index of double indirect block */
125#define TRIPLE 2 /* index of triple indirect block */
126/*
127 * Truncate the inode oip to at most length size, freeing the
128 * disk blocks.
129 */
130int
131ffs_truncate(vp, length, flags, cred, p)
132 struct vnode *vp;
133 off_t length;
134 int flags;
135 struct ucred *cred;
136 struct proc *p;
137{
138 register struct vnode *ovp = vp;
139 ufs_daddr_t lastblock;
140 register struct inode *oip;
141 ufs_daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
142 ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
143 register struct fs *fs;
144 struct buf *bp;
145 int offset, size, level;
146 long count, nblocks, blocksreleased = 0;
147 register int i;
148 int aflags, error, allerror;
149 off_t osize;
150
151 oip = VTOI(ovp);
152 fs = oip->i_fs;
153 if (length < 0)
154 return (EINVAL);
155 if (length > fs->fs_maxfilesize)
156 return (EFBIG);
157 if (ovp->v_type == VLNK &&
158 (oip->i_size < ovp->v_mount->mnt_maxsymlinklen || oip->i_din.di_blocks == 0)) {
159#ifdef DIAGNOSTIC
160 if (length != 0)
161 panic("ffs_truncate: partial truncate of symlink");
162#endif
163 bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
164 oip->i_size = 0;
165 oip->i_flag |= IN_CHANGE | IN_UPDATE;
166 return (UFS_UPDATE(ovp, 1));
167 }
168 if (oip->i_size == length) {
169 oip->i_flag |= IN_CHANGE | IN_UPDATE;
170 return (UFS_UPDATE(ovp, 0));
171 }
172 if (fs->fs_ronly)
173 panic("ffs_truncate: read-only filesystem");
174#ifdef QUOTA
175 error = getinoquota(oip);
176 if (error)
177 return (error);
178#endif
179 ovp->v_lasta = ovp->v_clen = ovp->v_cstart = ovp->v_lastw = 0;
180 if (DOINGSOFTDEP(ovp)) {
181 if (length > 0 || softdep_slowdown(ovp)) {
182 /*
183 * If a file is only partially truncated, then
184 * we have to clean up the data structures
185 * describing the allocation past the truncation
186 * point. Finding and deallocating those structures
187 * is a lot of work. Since partial truncation occurs
188 * rarely, we solve the problem by syncing the file
189 * so that it will have no data structures left.
190 */
191 if ((error = VOP_FSYNC(ovp, cred, MNT_WAIT,
192 p)) != 0)
193 return (error);
194 } else {
195#ifdef QUOTA
196 (void) chkdq(oip, -oip->i_blocks, NOCRED, 0);
197#endif
198 softdep_setup_freeblocks(oip, length);
199 vinvalbuf(ovp, 0, cred, p, 0, 0);
200 oip->i_flag |= IN_CHANGE | IN_UPDATE;
201 return (ffs_update(ovp, 0));
202 }
203 }
204 osize = oip->i_size;
205 /*
206 * Lengthen the size of the file. We must ensure that the
207 * last byte of the file is allocated. Since the smallest
208 * value of osize is 0, length will be at least 1.
209 */
210 if (osize < length) {
211 vnode_pager_setsize(ovp, length);
212 aflags = B_CLRBUF;
213 if (flags & IO_SYNC)
214 aflags |= B_SYNC;
215 error = VOP_BALLOC(ovp, length - 1, 1,
216 cred, aflags, &bp);
217 if (error)
218 return (error);
219 oip->i_size = length;
220 if (bp->b_bufsize == fs->fs_bsize)
221 bp->b_flags |= B_CLUSTEROK;
222 if (aflags & B_SYNC)
223 bwrite(bp);
224 else
225 bawrite(bp);
226 oip->i_flag |= IN_CHANGE | IN_UPDATE;
227 return (UFS_UPDATE(ovp, 1));
228 }
229 /*
230 * Shorten the size of the file. If the file is not being
231 * truncated to a block boundary, the contents of the
232 * partial block following the end of the file must be
233 * zero'ed in case it ever becomes accessible again because
234 * of subsequent file growth. Directories however are not
235 * zero'ed as they should grow back initialized to empty.
236 */
237 offset = blkoff(fs, length);
238 if (offset == 0) {
239 oip->i_size = length;
240 } else {
241 lbn = lblkno(fs, length);
242 aflags = B_CLRBUF;
243 if (flags & IO_SYNC)
244 aflags |= B_SYNC;
245 error = VOP_BALLOC(ovp, length - 1, 1, cred, aflags, &bp);
246 if (error) {
247 return (error);
248 }
249 /*
250 * When we are doing soft updates and the UFS_BALLOC
251 * above fills in a direct block hole with a full sized
252 * block that will be truncated down to a fragment below,
253 * we must flush out the block dependency with an FSYNC
254 * so that we do not get a soft updates inconsistency
255 * when we create the fragment below.
256 */
257 if (DOINGSOFTDEP(ovp) && lbn < NDADDR &&
258 fragroundup(fs, blkoff(fs, length)) < fs->fs_bsize &&
259 (error = VOP_FSYNC(ovp, cred, MNT_WAIT, p)) != 0) {
260 return (error);
261 }
262 oip->i_size = length;
263 size = blksize(fs, oip, lbn);
264 if (ovp->v_type != VDIR)
265 bzero((char *)bp->b_data + offset,
266 (u_int)(size - offset));
267 /* Kirk's code has reallocbuf(bp, size, 1) here */
268 allocbuf(bp, size);
269 if (bp->b_bufsize == fs->fs_bsize)
270 bp->b_flags |= B_CLUSTEROK;
271 if (aflags & B_SYNC)
272 bwrite(bp);
273 else
274 bawrite(bp);
275 }
276 /*
277 * Calculate index into inode's block list of
278 * last direct and indirect blocks (if any)
279 * which we want to keep. Lastblock is -1 when
280 * the file is truncated to 0.
281 */
282 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
283 lastiblock[SINGLE] = lastblock - NDADDR;
284 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
285 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
286 nblocks = btodb(fs->fs_bsize);
287 /*
288 * Update file and block pointers on disk before we start freeing
289 * blocks. If we crash before free'ing blocks below, the blocks
290 * will be returned to the free list. lastiblock values are also
291 * normalized to -1 for calls to ffs_indirtrunc below.
292 */
293 bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks);
294 for (level = TRIPLE; level >= SINGLE; level--)
295 if (lastiblock[level] < 0) {
296 oip->i_ib[level] = 0;
297 lastiblock[level] = -1;
298 }
299 for (i = NDADDR - 1; i > lastblock; i--)
300 oip->i_db[i] = 0;
301 oip->i_flag |= IN_CHANGE | IN_UPDATE;
302 allerror = UFS_UPDATE(ovp, 1);
303
304 /*
305 * Having written the new inode to disk, save its new configuration
306 * and put back the old block pointers long enough to process them.
307 * Note that we save the new block configuration so we can check it
308 * when we are done.
309 */
310 bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks);
311 bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks);
312 oip->i_size = osize;
313
314 error = vtruncbuf(ovp, cred, p, length, fs->fs_bsize);
315 if (error && (allerror == 0))
316 allerror = error;
317
318 /*
319 * Indirect blocks first.
320 */
321 indir_lbn[SINGLE] = -NDADDR;
322 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
323 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
324 for (level = TRIPLE; level >= SINGLE; level--) {
325 bn = oip->i_ib[level];
326 if (bn != 0) {
327 error = ffs_indirtrunc(oip, indir_lbn[level],
328 fsbtodb(fs, bn), lastiblock[level], level, &count);
329 if (error)
330 allerror = error;
331 blocksreleased += count;
332 if (lastiblock[level] < 0) {
333 oip->i_ib[level] = 0;
334 ffs_blkfree(oip, bn, fs->fs_bsize);
335 blocksreleased += nblocks;
336 }
337 }
338 if (lastiblock[level] >= 0)
339 goto done;
340 }
341
342 /*
343 * All whole direct blocks or frags.
344 */
345 for (i = NDADDR - 1; i > lastblock; i--) {
346 register long bsize;
347
348 bn = oip->i_db[i];
349 if (bn == 0)
350 continue;
351 oip->i_db[i] = 0;
352 bsize = blksize(fs, oip, i);
353 ffs_blkfree(oip, bn, bsize);
354 blocksreleased += btodb(bsize);
355 }
356 if (lastblock < 0)
357 goto done;
358
359 /*
360 * Finally, look for a change in size of the
361 * last direct block; release any frags.
362 */
363 bn = oip->i_db[lastblock];
364 if (bn != 0) {
365 long oldspace, newspace;
366
367 /*
368 * Calculate amount of space we're giving
369 * back as old block size minus new block size.
370 */
371 oldspace = blksize(fs, oip, lastblock);
372 oip->i_size = length;
373 newspace = blksize(fs, oip, lastblock);
374 if (newspace == 0)
375 panic("ffs_truncate: newspace");
376 if (oldspace - newspace > 0) {
377 /*
378 * Block number of space to be free'd is
379 * the old block # plus the number of frags
380 * required for the storage we're keeping.
381 */
382 bn += numfrags(fs, newspace);
383 ffs_blkfree(oip, bn, oldspace - newspace);
384 blocksreleased += btodb(oldspace - newspace);
385 }
386 }
387done:
388#ifdef DIAGNOSTIC
389 for (level = SINGLE; level <= TRIPLE; level++)
390 if (newblks[NDADDR + level] != oip->i_ib[level])
391 panic("ffs_truncate1");
392 for (i = 0; i < NDADDR; i++)
393 if (newblks[i] != oip->i_db[i])
394 panic("ffs_truncate2");
395 if (length == 0 &&
396 (!TAILQ_EMPTY(&ovp->v_dirtyblkhd) ||
397 !TAILQ_EMPTY(&ovp->v_cleanblkhd)))
398 panic("ffs_truncate3");
399#endif /* DIAGNOSTIC */
400 /*
401 * Put back the real size.
402 */
403 oip->i_size = length;
404 oip->i_blocks -= blocksreleased;
405
406 if (oip->i_blocks < 0) /* sanity */
407 oip->i_blocks = 0;
408 oip->i_flag |= IN_CHANGE;
409#ifdef QUOTA
410 (void) chkdq(oip, -blocksreleased, NOCRED, 0);
411#endif
412 return (allerror);
413}
414
415/*
416 * Release blocks associated with the inode ip and stored in the indirect
417 * block bn. Blocks are free'd in LIFO order up to (but not including)
418 * lastbn. If level is greater than SINGLE, the block is an indirect block
419 * and recursive calls to indirtrunc must be used to cleanse other indirect
420 * blocks.
421 *
422 * NB: triple indirect blocks are untested.
423 */
424static int
425ffs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
426 register struct inode *ip;
427 ufs_daddr_t lbn, lastbn;
428 ufs_daddr_t dbn;
429 int level;
430 long *countp;
431{
432 register int i;
433 struct buf *bp;
434 register struct fs *fs = ip->i_fs;
435 register ufs_daddr_t *bap;
436 struct vnode *vp;
437 ufs_daddr_t *copy = NULL, nb, nlbn, last;
438 long blkcount, factor;
439 int nblocks, blocksreleased = 0;
440 int error = 0, allerror = 0;
441
442 /*
443 * Calculate index in current block of last
444 * block to be kept. -1 indicates the entire
445 * block so we need not calculate the index.
446 */
447 factor = 1;
448 for (i = SINGLE; i < level; i++)
449 factor *= NINDIR(fs);
450 last = lastbn;
451 if (lastbn > 0)
452 last /= factor;
453 nblocks = btodb(fs->fs_bsize);
454 /*
455 * Get buffer of block pointers, zero those entries corresponding
456 * to blocks to be free'd, and update on disk copy first. Since
457 * double(triple) indirect before single(double) indirect, calls
458 * to bmap on these blocks will fail. However, we already have
459 * the on disk address, so we have to set the b_blkno field
460 * explicitly instead of letting bread do everything for us.
461 */
462 vp = ITOV(ip);
463 bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0);
464 if ((bp->b_flags & B_CACHE) == 0) {
465 curproc->p_stats->p_ru.ru_inblock++; /* pay for read */
466 bp->b_flags |= B_READ;
467 bp->b_flags &= ~(B_ERROR|B_INVAL);
468 if (bp->b_bcount > bp->b_bufsize)
469 panic("ffs_indirtrunc: bad buffer size");
470 bp->b_blkno = dbn;
471 vfs_busy_pages(bp, 0);
472 VOP_STRATEGY(bp->b_vp, bp);
473 error = biowait(bp);
474 }
475 if (error) {
476 brelse(bp);
477 *countp = 0;
478 return (error);
479 }
480
481 bap = (ufs_daddr_t *)bp->b_data;
482 if (lastbn != -1) {
483 MALLOC(copy, ufs_daddr_t *, fs->fs_bsize, M_TEMP, M_WAITOK);
484 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->fs_bsize);
485 bzero((caddr_t)&bap[last + 1],
486 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (ufs_daddr_t));
487 if (DOINGASYNC(vp)) {
488 bawrite(bp);
489 } else {
490 error = bwrite(bp);
491 if (error)
492 allerror = error;
493 }
494 bap = copy;
495 }
496
497 /*
498 * Recursively free totally unused blocks.
499 */
500 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
501 i--, nlbn += factor) {
502 nb = bap[i];
503 if (nb == 0)
504 continue;
505 if (level > SINGLE) {
506 if ((error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
507 (ufs_daddr_t)-1, level - 1, &blkcount)) != 0)
508 allerror = error;
509 blocksreleased += blkcount;
510 }
511 ffs_blkfree(ip, nb, fs->fs_bsize);
512 blocksreleased += nblocks;
513 }
514
515 /*
516 * Recursively free last partial block.
517 */
518 if (level > SINGLE && lastbn >= 0) {
519 last = lastbn % factor;
520 nb = bap[i];
521 if (nb != 0) {
522 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
523 last, level - 1, &blkcount);
524 if (error)
525 allerror = error;
526 blocksreleased += blkcount;
527 }
528 }
529 if (copy != NULL) {
530 FREE(copy, M_TEMP);
531 } else {
532 bp->b_flags |= B_INVAL | B_NOCACHE;
533 brelse(bp);
534 }
535
536 *countp = blocksreleased;
537 return (allerror);
538}