The thread/proc pointer argument in the VFS subsystem originally existed
[dragonfly.git] / sys / vfs / gnu / ext2fs / ext2_inode.c
CommitLineData
984263bc
MD
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 * Copyright (c) 1982, 1986, 1989, 1993
9 * The Regents of the University of California. All rights reserved.
10 *
11 * Redistribution and use in source and binary forms, with or without
12 * modification, are permitted provided that the following conditions
13 * are met:
14 * 1. Redistributions of source code must retain the above copyright
15 * notice, this list of conditions and the following disclaimer.
16 * 2. Redistributions in binary form must reproduce the above copyright
17 * notice, this list of conditions and the following disclaimer in the
18 * documentation and/or other materials provided with the distribution.
19 * 3. All advertising materials mentioning features or use of this software
20 * must display the following acknowledgement:
21 * This product includes software developed by the University of
22 * California, Berkeley and its contributors.
23 * 4. Neither the name of the University nor the names of its contributors
24 * may be used to endorse or promote products derived from this software
25 * without specific prior written permission.
26 *
27 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
28 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
29 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
30 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
31 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
32 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
33 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
34 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
35 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
36 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * SUCH DAMAGE.
38 *
39 * @(#)ext2_inode.c 8.5 (Berkeley) 12/30/93
40 * $FreeBSD: src/sys/gnu/ext2fs/ext2_inode.c,v 1.24.2.1 2000/08/03 00:52:57 peter Exp $
87de5057 41 * $DragonFly: src/sys/vfs/gnu/ext2fs/ext2_inode.c,v 1.18 2006/05/06 02:43:13 dillon Exp $
984263bc
MD
42 */
43
44#include "opt_quota.h"
45
46#include <sys/param.h>
47#include <sys/systm.h>
48#include <sys/mount.h>
49#include <sys/buf.h>
50#include <sys/vnode.h>
51#include <sys/malloc.h>
52
53#include <vm/vm.h>
54#include <vm/vm_extern.h>
55
1f1db49f
MD
56#include "quota.h"
57#include "inode.h"
58#include "ext2mount.h"
984263bc 59
1f2de5d4
MD
60#include "ext2_fs.h"
61#include "ext2_fs_sb.h"
62#include "fs.h"
63#include "ext2_extern.h"
984263bc 64
54078292
MD
65static int ext2_indirtrunc (struct inode *, daddr_t, off_t, daddr_t,
66 int, long *);
984263bc 67
984263bc
MD
68/*
69 * Update the access, modified, and inode change times as specified by the
70 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode
71 * to disk if the IN_MODIFIED flag is set (it may be set initially, or by
72 * the timestamp update). The IN_LAZYMOD flag is set to force a write
73 * later if not now. If we write now, then clear both IN_MODIFIED and
74 * IN_LAZYMOD to reflect the presumably successful write, and if waitfor is
75 * set, then wait for the write to complete.
76 */
77int
0f7f7a49 78ext2_update(struct vnode *vp, int waitfor)
984263bc 79{
f7aae92f 80 struct ext2_sb_info *fs;
984263bc
MD
81 struct buf *bp;
82 struct inode *ip;
83 int error;
84
1f1db49f 85 ext2_itimes(vp);
984263bc
MD
86 ip = VTOI(vp);
87 if ((ip->i_flag & IN_MODIFIED) == 0)
88 return (0);
89 ip->i_flag &= ~(IN_LAZYMOD | IN_MODIFIED);
90 if (vp->v_mount->mnt_flag & MNT_RDONLY)
91 return (0);
92 fs = ip->i_e2fs;
54078292
MD
93 error = bread(ip->i_devvp,
94 fsbtodoff(fs, ino_to_fsba(fs, ip->i_number)),
95 (int)fs->s_blocksize, &bp);
96 if (error) {
984263bc
MD
97 brelse(bp);
98 return (error);
99 }
100 ext2_di2ei( &ip->i_din, (struct ext2_inode *) ((char *)bp->b_data + EXT2_INODE_SIZE *
101 ino_to_fsbo(fs, ip->i_number)));
102/*
103 if (waitfor && (vp->v_mount->mnt_flag & MNT_ASYNC) == 0)
104 return (bwrite(bp));
105 else {
106*/
107 bdwrite(bp);
108 return (0);
109/*
110 }
111*/
112}
113
114#define SINGLE 0 /* index of single indirect block */
115#define DOUBLE 1 /* index of double indirect block */
116#define TRIPLE 2 /* index of triple indirect block */
117/*
118 * Truncate the inode oip to at most length size, freeing the
119 * disk blocks.
120 */
121int
87de5057 122ext2_truncate(struct vnode *vp, off_t length, int flags, struct ucred *cred)
984263bc 123{
f7aae92f
RG
124 struct vnode *ovp = vp;
125 daddr_t lastblock;
126 struct inode *oip;
984263bc
MD
127 daddr_t bn, lbn, lastiblock[NIADDR], indir_lbn[NIADDR];
128 daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
f7aae92f 129 struct ext2_sb_info *fs;
984263bc
MD
130 struct buf *bp;
131 int offset, size, level;
132 long count, nblocks, blocksreleased = 0;
f7aae92f 133 int i;
984263bc
MD
134 int aflags, error, allerror;
135 off_t osize;
136/*
137printf("ext2_truncate called %d to %d\n", VTOI(ovp)->i_number, length);
138*/ /*
139 * negative file sizes will totally break the code below and
140 * are not meaningful anyways.
141 */
142 if (length < 0)
143 return EFBIG;
144
145 oip = VTOI(ovp);
146 if (ovp->v_type == VLNK &&
147 oip->i_size < ovp->v_mount->mnt_maxsymlinklen) {
148#if DIAGNOSTIC
149 if (length != 0)
150 panic("ext2_truncate: partial truncate of symlink");
151#endif
152 bzero((char *)&oip->i_shortlink, (u_int)oip->i_size);
153 oip->i_size = 0;
154 oip->i_flag |= IN_CHANGE | IN_UPDATE;
1f1db49f 155 return (EXT2_UPDATE(ovp, 1));
984263bc
MD
156 }
157 if (oip->i_size == length) {
158 oip->i_flag |= IN_CHANGE | IN_UPDATE;
1f1db49f 159 return (EXT2_UPDATE(ovp, 0));
984263bc
MD
160 }
161#if QUOTA
1f1db49f 162 if ((error = ext2_getinoquota(oip)) != 0)
984263bc
MD
163 return (error);
164#endif
165 fs = oip->i_e2fs;
166 osize = oip->i_size;
167 ext2_discard_prealloc(oip);
168 /*
169 * Lengthen the size of the file. We must ensure that the
170 * last byte of the file is allocated. Since the smallest
171 * value of oszie is 0, length will be at least 1.
172 */
173 if (osize < length) {
174 offset = blkoff(fs, length - 1);
175 lbn = lblkno(fs, length - 1);
176 aflags = B_CLRBUF;
177 if (flags & IO_SYNC)
178 aflags |= B_SYNC;
179 vnode_pager_setsize(ovp, length);
180 if ((error = ext2_balloc(oip, lbn, offset + 1, cred, &bp,
181 aflags)) != 0)
182 return (error);
183 oip->i_size = length;
184 if (aflags & IO_SYNC)
185 bwrite(bp);
186 else
187 bawrite(bp);
188 oip->i_flag |= IN_CHANGE | IN_UPDATE;
1f1db49f 189 return (EXT2_UPDATE(ovp, 1));
984263bc
MD
190 }
191 /*
192 * Shorten the size of the file. If the file is not being
193 * truncated to a block boundry, the contents of the
194 * partial block following the end of the file must be
195 * zero'ed in case it ever become accessable again because
196 * of subsequent file growth.
197 */
198 /* I don't understand the comment above */
199 offset = blkoff(fs, length);
200 if (offset == 0) {
201 oip->i_size = length;
202 } else {
203 lbn = lblkno(fs, length);
204 aflags = B_CLRBUF;
205 if (flags & IO_SYNC)
206 aflags |= B_SYNC;
207 if ((error = ext2_balloc(oip, lbn, offset, cred, &bp,
208 aflags)) != 0)
209 return (error);
210 oip->i_size = length;
211 size = blksize(fs, oip, lbn);
212 bzero((char *)bp->b_data + offset, (u_int)(size - offset));
213 allocbuf(bp, size);
214 if (aflags & IO_SYNC)
215 bwrite(bp);
216 else
217 bawrite(bp);
218 }
219 /*
220 * Calculate index into inode's block list of
221 * last direct and indirect blocks (if any)
222 * which we want to keep. Lastblock is -1 when
223 * the file is truncated to 0.
224 */
225 lastblock = lblkno(fs, length + fs->s_blocksize - 1) - 1;
226 lastiblock[SINGLE] = lastblock - NDADDR;
227 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
228 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
229 nblocks = btodb(fs->s_blocksize);
230 /*
231 * Update file and block pointers on disk before we start freeing
232 * blocks. If we crash before free'ing blocks below, the blocks
233 * will be returned to the free list. lastiblock values are also
234 * normalized to -1 for calls to ext2_indirtrunc below.
235 */
236 bcopy((caddr_t)&oip->i_db[0], (caddr_t)oldblks, sizeof oldblks);
237 for (level = TRIPLE; level >= SINGLE; level--)
238 if (lastiblock[level] < 0) {
239 oip->i_ib[level] = 0;
240 lastiblock[level] = -1;
241 }
242 for (i = NDADDR - 1; i > lastblock; i--)
243 oip->i_db[i] = 0;
244 oip->i_flag |= IN_CHANGE | IN_UPDATE;
1f1db49f 245 allerror = EXT2_UPDATE(ovp, 1);
984263bc
MD
246
247 /*
248 * Having written the new inode to disk, save its new configuration
249 * and put back the old block pointers long enough to process them.
250 * Note that we save the new block configuration so we can check it
251 * when we are done.
252 */
253 bcopy((caddr_t)&oip->i_db[0], (caddr_t)newblks, sizeof newblks);
254 bcopy((caddr_t)oldblks, (caddr_t)&oip->i_db[0], sizeof oldblks);
255 oip->i_size = osize;
87de5057 256 error = vtruncbuf(ovp, length, (int)fs->s_blocksize);
984263bc
MD
257 if (error && (allerror == 0))
258 allerror = error;
259
260 /*
261 * Indirect blocks first.
262 */
263 indir_lbn[SINGLE] = -NDADDR;
264 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
265 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
266 for (level = TRIPLE; level >= SINGLE; level--) {
267 bn = oip->i_ib[level];
268 if (bn != 0) {
269 error = ext2_indirtrunc(oip, indir_lbn[level],
54078292 270 fsbtodoff(fs, bn), lastiblock[level], level, &count);
984263bc
MD
271 if (error)
272 allerror = error;
273 blocksreleased += count;
274 if (lastiblock[level] < 0) {
275 oip->i_ib[level] = 0;
276 ext2_blkfree(oip, bn, fs->s_frag_size);
277 blocksreleased += nblocks;
278 }
279 }
280 if (lastiblock[level] >= 0)
281 goto done;
282 }
283
284 /*
285 * All whole direct blocks or frags.
286 */
287 for (i = NDADDR - 1; i > lastblock; i--) {
f7aae92f 288 long bsize;
984263bc
MD
289
290 bn = oip->i_db[i];
291 if (bn == 0)
292 continue;
293 oip->i_db[i] = 0;
294 bsize = blksize(fs, oip, i);
295 ext2_blkfree(oip, bn, bsize);
296 blocksreleased += btodb(bsize);
297 }
298 if (lastblock < 0)
299 goto done;
300
301 /*
302 * Finally, look for a change in size of the
303 * last direct block; release any frags.
304 */
305 bn = oip->i_db[lastblock];
306 if (bn != 0) {
307 long oldspace, newspace;
308
309 /*
310 * Calculate amount of space we're giving
311 * back as old block size minus new block size.
312 */
313 oldspace = blksize(fs, oip, lastblock);
314 oip->i_size = length;
315 newspace = blksize(fs, oip, lastblock);
316 if (newspace == 0)
317 panic("itrunc: newspace");
318 if (oldspace - newspace > 0) {
319 /*
320 * Block number of space to be free'd is
321 * the old block # plus the number of frags
322 * required for the storage we're keeping.
323 */
324 bn += numfrags(fs, newspace);
325 ext2_blkfree(oip, bn, oldspace - newspace);
326 blocksreleased += btodb(oldspace - newspace);
327 }
328 }
329done:
330#if DIAGNOSTIC
331 for (level = SINGLE; level <= TRIPLE; level++)
332 if (newblks[NDADDR + level] != oip->i_ib[level])
333 panic("itrunc1");
334 for (i = 0; i < NDADDR; i++)
335 if (newblks[i] != oip->i_db[i])
336 panic("itrunc2");
6bae6177
MD
337 if (length == 0 && (!RB_EMPTY(&ovp->v_rbdirty_tree) ||
338 !RB_EMPTY(&ovp->v_rbclean_tree)))
984263bc
MD
339 panic("itrunc3");
340#endif /* DIAGNOSTIC */
341 /*
342 * Put back the real size.
343 */
344 oip->i_size = length;
345 oip->i_blocks -= blocksreleased;
346 if (oip->i_blocks < 0) /* sanity */
347 oip->i_blocks = 0;
348 oip->i_flag |= IN_CHANGE;
349 vnode_pager_setsize(ovp, length);
350#if QUOTA
1f1db49f 351 ext2_chkdq(oip, -blocksreleased, NOCRED, 0);
984263bc
MD
352#endif
353 return (allerror);
354}
355
356/*
357 * Release blocks associated with the inode ip and stored in the indirect
358 * block bn. Blocks are free'd in LIFO order up to (but not including)
359 * lastbn. If level is greater than SINGLE, the block is an indirect block
360 * and recursive calls to indirtrunc must be used to cleanse other indirect
361 * blocks.
362 *
363 * NB: triple indirect blocks are untested.
364 */
365
366static int
54078292 367ext2_indirtrunc(struct inode *ip, daddr_t lbn, off_t doffset, daddr_t lastbn,
0f7f7a49 368 int level, long *countp)
984263bc 369{
f7aae92f 370 int i;
984263bc 371 struct buf *bp;
f7aae92f
RG
372 struct ext2_sb_info *fs = ip->i_e2fs;
373 daddr_t *bap;
984263bc
MD
374 struct vnode *vp;
375 daddr_t *copy, nb, nlbn, last;
376 long blkcount, factor;
377 int nblocks, blocksreleased = 0;
378 int error = 0, allerror = 0;
379
380 /*
381 * Calculate index in current block of last
382 * block to be kept. -1 indicates the entire
383 * block so we need not calculate the index.
384 */
385 factor = 1;
386 for (i = SINGLE; i < level; i++)
387 factor *= NINDIR(fs);
388 last = lastbn;
389 if (lastbn > 0)
390 last /= factor;
391 nblocks = btodb(fs->s_blocksize);
392 /*
393 * Get buffer of block pointers, zero those entries corresponding
394 * to blocks to be free'd, and update on disk copy first. Since
395 * double(triple) indirect before single(double) indirect, calls
396 * to bmap on these blocks will fail. However, we already have
54078292 397 * the on disk address, so we have to set the bio_offset field
984263bc
MD
398 * explicitly instead of letting bread do everything for us.
399 */
400 vp = ITOV(ip);
54078292 401 bp = getblk(vp, lblktodoff(fs, lbn), (int)fs->s_blocksize, 0, 0);
10f3fee5 402 if ((bp->b_flags & B_CACHE) == 0) {
a8f169e2 403 bp->b_flags &= ~(B_ERROR | B_INVAL);
10f3fee5 404 bp->b_cmd = BUF_CMD_READ;
984263bc
MD
405 if (bp->b_bcount > bp->b_bufsize)
406 panic("ext2_indirtrunc: bad buffer size");
54078292 407 bp->b_bio2.bio_offset = doffset;
10f3fee5 408 vfs_busy_pages(bp->b_vp, bp);
81b5c339 409 vn_strategy(vp, &bp->b_bio1);
984263bc
MD
410 error = biowait(bp);
411 }
412 if (error) {
413 brelse(bp);
414 *countp = 0;
415 return (error);
416 }
417
418 bap = (daddr_t *)bp->b_data;
419 MALLOC(copy, daddr_t *, fs->s_blocksize, M_TEMP, M_WAITOK);
420 bcopy((caddr_t)bap, (caddr_t)copy, (u_int)fs->s_blocksize);
421 bzero((caddr_t)&bap[last + 1],
422 (u_int)(NINDIR(fs) - (last + 1)) * sizeof (daddr_t));
423 if (last == -1)
424 bp->b_flags |= B_INVAL;
425 error = bwrite(bp);
426 if (error)
427 allerror = error;
428 bap = copy;
429
430 /*
431 * Recursively free totally unused blocks.
432 */
433 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
434 i--, nlbn += factor) {
435 nb = bap[i];
436 if (nb == 0)
437 continue;
438 if (level > SINGLE) {
439 if ((error = ext2_indirtrunc(ip, nlbn,
54078292 440 fsbtodoff(fs, nb), (daddr_t)-1, level - 1, &blkcount)) != 0)
984263bc
MD
441 allerror = error;
442 blocksreleased += blkcount;
443 }
444 ext2_blkfree(ip, nb, fs->s_blocksize);
445 blocksreleased += nblocks;
446 }
447
448 /*
449 * Recursively free last partial block.
450 */
451 if (level > SINGLE && lastbn >= 0) {
452 last = lastbn % factor;
453 nb = bap[i];
454 if (nb != 0) {
54078292
MD
455 error = ext2_indirtrunc(ip, nlbn, fsbtodoff(fs, nb),
456 last, level - 1, &blkcount);
457 if (error)
984263bc
MD
458 allerror = error;
459 blocksreleased += blkcount;
460 }
461 }
462 FREE(copy, M_TEMP);
463 *countp = blocksreleased;
464 return (allerror);
465}
466
467/*
1f1db49f 468 * Last reference to an inode. If necessary, write or delete it.
0f7f7a49 469 *
87de5057 470 * ext2_inactive(struct vnode *a_vp)
984263bc
MD
471 */
472int
0f7f7a49 473ext2_inactive(struct vop_inactive_args *ap)
984263bc 474{
1f1db49f
MD
475 struct vnode *vp = ap->a_vp;
476 struct inode *ip = VTOI(vp);
1f1db49f
MD
477 int mode, error = 0;
478
479 ext2_discard_prealloc(ip);
480 if (prtactive && vp->v_usecount != 1)
481 vprint("ext2_inactive: pushing active", vp);
482
483 /*
484 * Ignore inodes related to stale file handles.
485 */
486 if (ip == NULL || ip->i_mode == 0)
487 goto out;
488 if (ip->i_nlink <= 0 && (vp->v_mount->mnt_flag & MNT_RDONLY) == 0) {
489#ifdef QUOTA
490 if (!ext2_getinoquota(ip))
491 (void)ext2_chkiq(ip, -1, NOCRED, FORCE);
492#endif
87de5057 493 error = EXT2_TRUNCATE(vp, (off_t)0, 0, NOCRED);
1f1db49f
MD
494 ip->i_rdev = 0;
495 mode = ip->i_mode;
496 ip->i_mode = 0;
497 ip->i_flag |= IN_CHANGE | IN_UPDATE;
498 EXT2_VFREE(vp, ip->i_number, mode);
499 }
500 if (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))
501 EXT2_UPDATE(vp, 0);
502out:
503 /*
504 * If we are done with the inode, reclaim it
505 * so that it can be reused immediately.
506 */
507 if (ip == NULL || ip->i_mode == 0)
87de5057 508 vrecycle(vp);
1f1db49f
MD
509 return (error);
510}
511
512/*
513 * Reclaim an inode so that it can be used for other purposes.
514 *
87de5057 515 * ext2_reclaim(struct vnode *a_vp)
1f1db49f
MD
516 */
517int
518ext2_reclaim(struct vop_reclaim_args *ap)
519{
520 struct inode *ip;
521 struct vnode *vp = ap->a_vp;
522#ifdef QUOTA
523 int i;
524#endif
525
526 if (prtactive && vp->v_usecount != 1)
527 vprint("ext2_reclaim: pushing active", vp);
528 ip = VTOI(vp);
529
530 /*
531 * Lazy updates.
532 */
533 if (ip) {
1f1db49f
MD
534 if (ip->i_flag & IN_LAZYMOD) {
535 ip->i_flag |= IN_MODIFIED;
536 EXT2_UPDATE(vp, 0);
537 }
538 }
539#ifdef INVARIANTS
540 if (ip && (ip->i_flag & (IN_ACCESS | IN_CHANGE | IN_MODIFIED | IN_UPDATE))) {
541 printf("WARNING: INODE %ld flags %08x: modified inode being released!\n", (long)ip->i_number, (int)ip->i_flag);
542 ip->i_flag |= IN_MODIFIED;
543 EXT2_UPDATE(vp, 0);
544 }
545#endif
546 /*
547 * Remove the inode from its hash chain and purge namecache
548 * data associated with the vnode.
549 */
550 vp->v_data = NULL;
551 if (ip) {
552 ext2_ihashrem(ip);
553 if (ip->i_devvp) {
554 vrele(ip->i_devvp);
555 ip->i_devvp = 0;
556 }
557#ifdef QUOTA
558 for (i = 0; i < MAXQUOTAS; i++) {
559 if (ip->i_dquot[i] != NODQUOT) {
560 ext2_dqrele(vp, ip->i_dquot[i]);
561 ip->i_dquot[i] = NODQUOT;
562 }
563 }
564#endif
565#ifdef UFS_DIRHASH
566 if (ip->i_dirhash != NULL)
567 ext2dirhash_free(ip);
568#endif
569 free(ip, VFSTOEXT2(vp->v_mount)->um_malloctype);
570 }
571 return (0);
984263bc 572}