Register keyword removal
[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 $
12e4aaff 35 * $DragonFly: src/sys/vfs/ufs/ffs_inode.c,v 1.6 2003/07/03 17:24:03 dillon Exp $
984263bc
MD
36 */
37
38#include "opt_quota.h"
39
40#include <sys/param.h>
41#include <sys/systm.h>
42#include <sys/mount.h>
43#include <sys/proc.h>
44#include <sys/buf.h>
45#include <sys/vnode.h>
46#include <sys/kernel.h>
47#include <sys/malloc.h>
48#include <sys/resourcevar.h>
49#include <sys/vmmeter.h>
50
51#include <vm/vm.h>
52#include <vm/vm_extern.h>
53
54#include <ufs/ufs/quota.h>
55#include <ufs/ufs/ufsmount.h>
56#include <ufs/ufs/inode.h>
57#include <ufs/ufs/ufs_extern.h>
58
59#include <ufs/ffs/fs.h>
60#include <ufs/ffs/ffs_extern.h>
61
12e4aaff
MD
62#include <vm/vm_page2.h>
63
984263bc
MD
64static int ffs_indirtrunc __P((struct inode *, ufs_daddr_t, ufs_daddr_t,
65 ufs_daddr_t, int, long *));
66
67/*
68 * Update the access, modified, and inode change times as specified by the
69 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode
70 * to disk if the IN_MODIFIED flag is set (it may be set initially, or by
71 * the timestamp update). The IN_LAZYMOD flag is set to force a write
72 * later if not now. If we write now, then clear both IN_MODIFIED and
73 * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is
74 * set, then wait for the write to complete.
75 */
76int
77ffs_update(vp, waitfor)
78 struct vnode *vp;
79 int waitfor;
80{
81 register struct fs *fs;
82 struct buf *bp;
83 struct inode *ip;
84 int error;
85
86 ufs_itimes(vp);
87 ip = VTOI(vp);
88 if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
89 return (0);
90 ip->i_flag &= ~(IN_LAZYMOD | IN_MODIFIED);
91 fs = ip->i_fs;
92 if (fs->fs_ronly)
93 return (0);
94 /*
95 * Ensure that uid and gid are correct. This is a temporary
96 * fix until fsck has been changed to do the update.
97 */
98 if (fs->fs_inodefmt < FS_44INODEFMT) { /* XXX */
99 ip->i_din.di_ouid = ip->i_uid; /* XXX */
100 ip->i_din.di_ogid = ip->i_gid; /* XXX */
101 } /* XXX */
102 error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
3b568787 103 (int)fs->fs_bsize, &bp);
984263bc
MD
104 if (error) {
105 brelse(bp);
106 return (error);
107 }
108 if (DOINGSOFTDEP(vp))
109 softdep_update_inodeblock(ip, bp, waitfor);
110 else if (ip->i_effnlink != ip->i_nlink)
111 panic("ffs_update: bad link cnt");
112 *((struct dinode *)bp->b_data +
113 ino_to_fsbo(fs, ip->i_number)) = ip->i_din;
114 if (waitfor && !DOINGASYNC(vp)) {
115 return (bwrite(bp));
116 } else if (vm_page_count_severe() || buf_dirty_count_severe()) {
117 return (bwrite(bp));
118 } else {
119 if (bp->b_bufsize == fs->fs_bsize)
120 bp->b_flags |= B_CLUSTEROK;
121 bdwrite(bp);
122 return (0);
123 }
124}
125
126#define SINGLE 0 /* index of single indirect block */
127#define DOUBLE 1 /* index of double indirect block */
128#define TRIPLE 2 /* index of triple indirect block */
129/*
130 * Truncate the inode oip to at most length size, freeing the
131 * disk blocks.
132 */
133int
dadab5e9 134ffs_truncate(vp, length, flags, cred, td)
984263bc
MD
135 struct vnode *vp;
136 off_t length;
137 int flags;
138 struct ucred *cred;
dadab5e9 139 struct thread *td;
984263bc
MD
140{
141 register struct vnode *ovp = vp;
142 ufs_daddr_t lastblock;
143 register struct inode *oip;
144 ufs_daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
145 ufs_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
146 register struct fs *fs;
147 struct buf *bp;
148 int offset, size, level;
149 long count, nblocks, blocksreleased = 0;
150 register int i;
151 int aflags, error, allerror;
152 off_t osize;
153
154 oip = VTOI(ovp);
155 fs = oip->i_fs;
156 if (length < 0)
157 return (EINVAL);
158 if (length > fs->fs_maxfilesize)
159 return (EFBIG);
160 if (ovp->v_type == VLNK &&
161 (oip->i_size < ovp->v_mount->mnt_maxsymlinklen || oip->i_din.di_blocks == 0)) {
162#ifdef DIAGNOSTIC
163 if (length != 0)
164 panic("ffs_truncate: partial truncate of symlink");
165#endif
166 bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
167 oip->i_size = 0;
168 oip->i_flag |= IN_CHANGE | IN_UPDATE;
169 return (UFS_UPDATE(ovp, 1));
170 }
171 if (oip->i_size == length) {
172 oip->i_flag |= IN_CHANGE | IN_UPDATE;
173 return (UFS_UPDATE(ovp, 0));
174 }
175 if (fs->fs_ronly)
176 panic("ffs_truncate: read-only filesystem");
177#ifdef QUOTA
178 error = getinoquota(oip);
179 if (error)
180 return (error);
181#endif
182 ovp->v_lasta = ovp->v_clen = ovp->v_cstart = ovp->v_lastw = 0;
183 if (DOINGSOFTDEP(ovp)) {
184 if (length > 0 || softdep_slowdown(ovp)) {
185 /*
186 * If a file is only partially truncated, then
187 * we have to clean up the data structures
188 * describing the allocation past the truncation
189 * point. Finding and deallocating those structures
190 * is a lot of work. Since partial truncation occurs
191 * rarely, we solve the problem by syncing the file
192 * so that it will have no data structures left.
193 */
3b568787 194 if ((error = VOP_FSYNC(ovp, MNT_WAIT, td)) != 0)
984263bc
MD
195 return (error);
196 } else {
197#ifdef QUOTA
198 (void) chkdq(oip, -oip->i_blocks, NOCRED, 0);
199#endif
200 softdep_setup_freeblocks(oip, length);
3b568787 201 vinvalbuf(ovp, 0, td, 0, 0);
984263bc
MD
202 oip->i_flag |= IN_CHANGE | IN_UPDATE;
203 return (ffs_update(ovp, 0));
204 }
205 }
206 osize = oip->i_size;
207 /*
208 * Lengthen the size of the file. We must ensure that the
209 * last byte of the file is allocated. Since the smallest
210 * value of osize is 0, length will be at least 1.
211 */
212 if (osize < length) {
213 vnode_pager_setsize(ovp, length);
214 aflags = B_CLRBUF;
215 if (flags & IO_SYNC)
216 aflags |= B_SYNC;
217 error = VOP_BALLOC(ovp, length - 1, 1,
218 cred, aflags, &bp);
219 if (error)
220 return (error);
221 oip->i_size = length;
222 if (bp->b_bufsize == fs->fs_bsize)
223 bp->b_flags |= B_CLUSTEROK;
224 if (aflags & B_SYNC)
225 bwrite(bp);
226 else
227 bawrite(bp);
228 oip->i_flag |= IN_CHANGE | IN_UPDATE;
229 return (UFS_UPDATE(ovp, 1));
230 }
231 /*
232 * Shorten the size of the file. If the file is not being
233 * truncated to a block boundary, the contents of the
234 * partial block following the end of the file must be
235 * zero'ed in case it ever becomes accessible again because
236 * of subsequent file growth. Directories however are not
237 * zero'ed as they should grow back initialized to empty.
238 */
239 offset = blkoff(fs, length);
240 if (offset == 0) {
241 oip->i_size = length;
242 } else {
243 lbn = lblkno(fs, length);
244 aflags = B_CLRBUF;
245 if (flags & IO_SYNC)
246 aflags |= B_SYNC;
247 error = VOP_BALLOC(ovp, length - 1, 1, cred, aflags, &bp);
248 if (error) {
249 return (error);
250 }
251 /*
252 * When we are doing soft updates and the UFS_BALLOC
253 * above fills in a direct block hole with a full sized
254 * block that will be truncated down to a fragment below,
255 * we must flush out the block dependency with an FSYNC
256 * so that we do not get a soft updates inconsistency
257 * when we create the fragment below.
258 */
259 if (DOINGSOFTDEP(ovp) && lbn < NDADDR &&
260 fragroundup(fs, blkoff(fs, length)) < fs->fs_bsize &&
3b568787 261 (error = VOP_FSYNC(ovp, MNT_WAIT, td)) != 0) {
984263bc
MD
262 return (error);
263 }
264 oip->i_size = length;
265 size = blksize(fs, oip, lbn);
266 if (ovp->v_type != VDIR)
267 bzero((char *)bp->b_data + offset,
268 (u_int)(size - offset));
269 /* Kirk's code has reallocbuf(bp, size, 1) here */
270 allocbuf(bp, size);
271 if (bp->b_bufsize == fs->fs_bsize)
272 bp->b_flags |= B_CLUSTEROK;
273 if (aflags & B_SYNC)
274 bwrite(bp);
275 else
276 bawrite(bp);
277 }
278 /*
279 * Calculate index into inode's block list of
280 * last direct and indirect blocks (if any)
281 * which we want to keep. Lastblock is -1 when
282 * the file is truncated to 0.
283 */
284 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
285 lastiblock[SINGLE] = lastblock - NDADDR;
286 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
287 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
288 nblocks = btodb(fs->fs_bsize);
289 /*
290 * Update file and block pointers on disk before we start freeing
291 * blocks. If we crash before free'ing blocks below, the blocks
292 * will be returned to the free list. lastiblock values are also
293 * normalized to -1 for calls to ffs_indirtrunc below.
294 */
295 bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks);
296 for (level = TRIPLE; level >= SINGLE; level--)
297 if (lastiblock[level] < 0) {
298 oip->i_ib[level] = 0;
299 lastiblock[level] = -1;
300 }
301 for (i = NDADDR - 1; i > lastblock; i--)
302 oip->i_db[i] = 0;
303 oip->i_flag |= IN_CHANGE | IN_UPDATE;
304 allerror = UFS_UPDATE(ovp, 1);
305
306 /*
307 * Having written the new inode to disk, save its new configuration
308 * and put back the old block pointers long enough to process them.
309 * Note that we save the new block configuration so we can check it
310 * when we are done.
311 */
312 bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks);
313 bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks);
314 oip->i_size = osize;
315
3b568787 316 error = vtruncbuf(ovp, td, length, fs->fs_bsize);
984263bc
MD
317 if (error && (allerror == 0))
318 allerror = error;
319
320 /*
321 * Indirect blocks first.
322 */
323 indir_lbn[SINGLE] = -NDADDR;
324 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
325 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
326 for (level = TRIPLE; level >= SINGLE; level--) {
327 bn = oip->i_ib[level];
328 if (bn != 0) {
329 error = ffs_indirtrunc(oip, indir_lbn[level],
330 fsbtodb(fs, bn), lastiblock[level], level, &count);
331 if (error)
332 allerror = error;
333 blocksreleased += count;
334 if (lastiblock[level] < 0) {
335 oip->i_ib[level] = 0;
336 ffs_blkfree(oip, bn, fs->fs_bsize);
337 blocksreleased += nblocks;
338 }
339 }
340 if (lastiblock[level] >= 0)
341 goto done;
342 }
343
344 /*
345 * All whole direct blocks or frags.
346 */
347 for (i = NDADDR - 1; i > lastblock; i--) {
348 register long bsize;
349
350 bn = oip->i_db[i];
351 if (bn == 0)
352 continue;
353 oip->i_db[i] = 0;
354 bsize = blksize(fs, oip, i);
355 ffs_blkfree(oip, bn, bsize);
356 blocksreleased += btodb(bsize);
357 }
358 if (lastblock < 0)
359 goto done;
360
361 /*
362 * Finally, look for a change in size of the
363 * last direct block; release any frags.
364 */
365 bn = oip->i_db[lastblock];
366 if (bn != 0) {
367 long oldspace, newspace;
368
369 /*
370 * Calculate amount of space we're giving
371 * back as old block size minus new block size.
372 */
373 oldspace = blksize(fs, oip, lastblock);
374 oip->i_size = length;
375 newspace = blksize(fs, oip, lastblock);
376 if (newspace == 0)
377 panic("ffs_truncate: newspace");
378 if (oldspace - newspace > 0) {
379 /*
380 * Block number of space to be free'd is
381 * the old block # plus the number of frags
382 * required for the storage we're keeping.
383 */
384 bn += numfrags(fs, newspace);
385 ffs_blkfree(oip, bn, oldspace - newspace);
386 blocksreleased += btodb(oldspace - newspace);
387 }
388 }
389done:
390#ifdef DIAGNOSTIC
391 for (level = SINGLE; level <= TRIPLE; level++)
392 if (newblks[NDADDR + level] != oip->i_ib[level])
393 panic("ffs_truncate1");
394 for (i = 0; i < NDADDR; i++)
395 if (newblks[i] != oip->i_db[i])
396 panic("ffs_truncate2");
397 if (length == 0 &&
398 (!TAILQ_EMPTY(&ovp->v_dirtyblkhd) ||
399 !TAILQ_EMPTY(&ovp->v_cleanblkhd)))
400 panic("ffs_truncate3");
401#endif /* DIAGNOSTIC */
402 /*
403 * Put back the real size.
404 */
405 oip->i_size = length;
406 oip->i_blocks -= blocksreleased;
407
408 if (oip->i_blocks < 0) /* sanity */
409 oip->i_blocks = 0;
410 oip->i_flag |= IN_CHANGE;
411#ifdef QUOTA
412 (void) chkdq(oip, -blocksreleased, NOCRED, 0);
413#endif
414 return (allerror);
415}
416
417/*
418 * Release blocks associated with the inode ip and stored in the indirect
419 * block bn. Blocks are free'd in LIFO order up to (but not including)
420 * lastbn. If level is greater than SINGLE, the block is an indirect block
421 * and recursive calls to indirtrunc must be used to cleanse other indirect
422 * blocks.
423 *
424 * NB: triple indirect blocks are untested.
425 */
426static int
427ffs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
428 register struct inode *ip;
429 ufs_daddr_t lbn, lastbn;
430 ufs_daddr_t dbn;
431 int level;
432 long *countp;
433{
434 register int i;
435 struct buf *bp;
436 register struct fs *fs = ip->i_fs;
437 register ufs_daddr_t *bap;
438 struct vnode *vp;
439 ufs_daddr_t *copy = NULL, nb, nlbn, last;
440 long blkcount, factor;
441 int nblocks, blocksreleased = 0;
442 int error = 0, allerror = 0;
443
444 /*
445 * Calculate index in current block of last
446 * block to be kept. -1 indicates the entire
447 * block so we need not calculate the index.
448 */
449 factor = 1;
450 for (i = SINGLE; i < level; i++)
451 factor *= NINDIR(fs);
452 last = lastbn;
453 if (lastbn > 0)
454 last /= factor;
455 nblocks = btodb(fs->fs_bsize);
456 /*
457 * Get buffer of block pointers, zero those entries corresponding
458 * to blocks to be free'd, and update on disk copy first. Since
459 * double(triple) indirect before single(double) indirect, calls
460 * to bmap on these blocks will fail. However, we already have
461 * the on disk address, so we have to set the b_blkno field
462 * explicitly instead of letting bread do everything for us.
463 */
464 vp = ITOV(ip);
465 bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0);
466 if ((bp->b_flags & B_CACHE) == 0) {
984263bc
MD
467 bp->b_flags |= B_READ;
468 bp->b_flags &= ~(B_ERROR|B_INVAL);
469 if (bp->b_bcount > bp->b_bufsize)
470 panic("ffs_indirtrunc: bad buffer size");
471 bp->b_blkno = dbn;
472 vfs_busy_pages(bp, 0);
473 VOP_STRATEGY(bp->b_vp, bp);
474 error = biowait(bp);
475 }
476 if (error) {
477 brelse(bp);
478 *countp = 0;
479 return (error);
480 }
481
482 bap = (ufs_daddr_t *)bp->b_data;
483 if (lastbn != -1) {
484 MALLOC(copy, ufs_daddr_t *, fs->fs_bsize, M_TEMP, M_WAITOK);
485 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->fs_bsize);
486 bzero((caddr_t)&bap[last + 1],
487 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (ufs_daddr_t));
488 if (DOINGASYNC(vp)) {
489 bawrite(bp);
490 } else {
491 error = bwrite(bp);
492 if (error)
493 allerror = error;
494 }
495 bap = copy;
496 }
497
498 /*
499 * Recursively free totally unused blocks.
500 */
501 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
502 i--, nlbn += factor) {
503 nb = bap[i];
504 if (nb == 0)
505 continue;
506 if (level > SINGLE) {
507 if ((error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
508 (ufs_daddr_t)-1, level - 1, &blkcount)) != 0)
509 allerror = error;
510 blocksreleased += blkcount;
511 }
512 ffs_blkfree(ip, nb, fs->fs_bsize);
513 blocksreleased += nblocks;
514 }
515
516 /*
517 * Recursively free last partial block.
518 */
519 if (level > SINGLE && lastbn >= 0) {
520 last = lastbn % factor;
521 nb = bap[i];
522 if (nb != 0) {
523 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
524 last, level - 1, &blkcount);
525 if (error)
526 allerror = error;
527 blocksreleased += blkcount;
528 }
529 }
530 if (copy != NULL) {
531 FREE(copy, M_TEMP);
532 } else {
533 bp->b_flags |= B_INVAL | B_NOCACHE;
534 brelse(bp);
535 }
536
537 *countp = blocksreleased;
538 return (allerror);
539}