2 * Copyright (c) 1982, 1986, 1989, 1993
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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 * 4. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)ffs_inode.c 8.13 (Berkeley) 4/21/95
32 #include <sys/cdefs.h>
33 __FBSDID("$FreeBSD$");
35 #include "opt_quota.h"
37 #include <sys/param.h>
38 #include <sys/systm.h>
39 #include <sys/mount.h>
43 #include <sys/vnode.h>
44 #include <sys/malloc.h>
45 #include <sys/resourcevar.h>
46 #include <sys/vmmeter.h>
50 #include <vm/vm_extern.h>
51 #include <vm/vm_object.h>
53 #include <ufs/ufs/extattr.h>
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>
59 #include <ufs/ffs/fs.h>
60 #include <ufs/ffs/ffs_extern.h>
62 static int ffs_indirtrunc(struct inode *, ufs2_daddr_t, ufs2_daddr_t,
63 ufs2_daddr_t, int, ufs2_daddr_t *);
66 * Update the access, modified, and inode change times as specified by the
67 * IN_ACCESS, IN_UPDATE, and IN_CHANGE flags respectively. Write the inode
68 * to disk if the IN_MODIFIED flag is set (it may be set initially, or by
69 * the timestamp update). The IN_LAZYMOD flag is set to force a write
70 * later if not now. The IN_LAZYACCESS is set instead of IN_MODIFIED if the fs
71 * is currently being suspended (or is suspended) and vnode has been accessed.
72 * If we write now, then clear IN_MODIFIED, IN_LAZYACCESS and IN_LAZYMOD to
73 * reflect the presumably successful write, and if waitfor is set, then wait
74 * for the write to complete.
77 ffs_update(vp, waitfor)
86 ASSERT_VOP_ELOCKED(vp, "ffs_update");
89 if ((ip->i_flag & IN_MODIFIED) == 0 && waitfor == 0)
91 ip->i_flag &= ~(IN_LAZYACCESS | IN_LAZYMOD | IN_MODIFIED);
93 if (fs->fs_ronly && ip->i_ump->um_fsckpid == 0)
95 error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
96 (int)fs->fs_bsize, NOCRED, &bp);
101 if (DOINGSOFTDEP(vp))
102 softdep_update_inodeblock(ip, bp, waitfor);
103 else if (ip->i_effnlink != ip->i_nlink)
104 panic("ffs_update: bad link cnt");
105 if (ip->i_ump->um_fstype == UFS1)
106 *((struct ufs1_dinode *)bp->b_data +
107 ino_to_fsbo(fs, ip->i_number)) = *ip->i_din1;
109 *((struct ufs2_dinode *)bp->b_data +
110 ino_to_fsbo(fs, ip->i_number)) = *ip->i_din2;
111 if (waitfor && !DOINGASYNC(vp)) {
113 } else if (vm_page_count_severe() || buf_dirty_count_severe()) {
116 if (bp->b_bufsize == fs->fs_bsize)
117 bp->b_flags |= B_CLUSTEROK;
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 */
127 * Truncate the inode ip to at most length size, freeing the
131 ffs_truncate(vp, length, flags, cred, td)
139 ufs2_daddr_t bn, lbn, lastblock, lastiblock[NIADDR], indir_lbn[NIADDR];
140 ufs2_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
141 ufs2_daddr_t count, blocksreleased = 0, datablocks;
145 struct ufsmount *ump;
146 int softdeptrunc, journaltrunc;
147 int needextclean, extblocks;
148 int offset, size, level, nblocks;
149 int i, error, allerror;
157 ASSERT_VOP_LOCKED(vp, "ffs_truncate");
161 if (length > fs->fs_maxfilesize)
164 error = getinoquota(ip);
169 * Historically clients did not have to specify which data
170 * they were truncating. So, if not specified, we assume
171 * traditional behavior, e.g., just the normal data.
173 if ((flags & (IO_EXT | IO_NORMAL)) == 0)
175 if (!DOINGSOFTDEP(vp) && !DOINGASYNC(vp))
178 * If we are truncating the extended-attributes, and cannot
179 * do it with soft updates, then do it slowly here. If we are
180 * truncating both the extended attributes and the file contents
181 * (e.g., the file is being unlinked), then pick it off with
182 * soft updates below.
187 journaltrunc = DOINGSUJ(vp);
188 if (journaltrunc == 0 && DOINGSOFTDEP(vp) && length == 0)
189 softdeptrunc = !softdep_slowdown(vp);
191 datablocks = DIP(ip, i_blocks);
192 if (fs->fs_magic == FS_UFS2_MAGIC && ip->i_din2->di_extsize > 0) {
193 extblocks = btodb(fragroundup(fs, ip->i_din2->di_extsize));
194 datablocks -= extblocks;
196 if ((flags & IO_EXT) && extblocks > 0) {
198 panic("ffs_truncate: partial trunc of extdata");
199 if (softdeptrunc || journaltrunc) {
200 if ((flags & IO_NORMAL) == 0)
204 if ((error = ffs_syncvnode(vp, MNT_WAIT)) != 0)
207 (void) chkdq(ip, -extblocks, NOCRED, 0);
209 vinvalbuf(vp, V_ALT, 0, 0);
211 OFF_TO_IDX(lblktosize(fs, -extblocks)), 0);
212 osize = ip->i_din2->di_extsize;
213 ip->i_din2->di_blocks -= extblocks;
214 ip->i_din2->di_extsize = 0;
215 for (i = 0; i < NXADDR; i++) {
216 oldblks[i] = ip->i_din2->di_extb[i];
217 ip->i_din2->di_extb[i] = 0;
219 ip->i_flag |= IN_CHANGE;
220 if ((error = ffs_update(vp, 1)))
222 for (i = 0; i < NXADDR; i++) {
225 ffs_blkfree(ump, fs, ip->i_devvp, oldblks[i],
226 sblksize(fs, osize, i), ip->i_number,
231 if ((flags & IO_NORMAL) == 0)
233 if (vp->v_type == VLNK &&
234 (ip->i_size < vp->v_mount->mnt_maxsymlinklen ||
238 panic("ffs_truncate: partial truncate of symlink");
240 bzero(SHORTLINK(ip), (u_int)ip->i_size);
242 DIP_SET(ip, i_size, 0);
243 ip->i_flag |= IN_CHANGE | IN_UPDATE;
246 return ffs_update(vp, 1);
248 if (ip->i_size == length) {
249 ip->i_flag |= IN_CHANGE | IN_UPDATE;
252 return ffs_update(vp, 0);
255 panic("ffs_truncate: read-only filesystem");
256 if ((ip->i_flags & SF_SNAPSHOT) != 0)
258 vp->v_lasta = vp->v_clen = vp->v_cstart = vp->v_lastw = 0;
261 * Lengthen the size of the file. We must ensure that the
262 * last byte of the file is allocated. Since the smallest
263 * value of osize is 0, length will be at least 1.
265 if (osize < length) {
266 vnode_pager_setsize(vp, length);
268 error = UFS_BALLOC(vp, length - 1, 1, cred, flags, &bp);
270 vnode_pager_setsize(vp, osize);
274 DIP_SET(ip, i_size, length);
275 if (bp->b_bufsize == fs->fs_bsize)
276 bp->b_flags |= B_CLUSTEROK;
281 ip->i_flag |= IN_CHANGE | IN_UPDATE;
282 return ffs_update(vp, 1);
284 if (DOINGSOFTDEP(vp)) {
285 if (softdeptrunc == 0 && journaltrunc == 0) {
287 * If a file is only partially truncated, then
288 * we have to clean up the data structures
289 * describing the allocation past the truncation
290 * point. Finding and deallocating those structures
291 * is a lot of work. Since partial truncation occurs
292 * rarely, we solve the problem by syncing the file
293 * so that it will have no data structures left.
295 if ((error = ffs_syncvnode(vp, MNT_WAIT)) != 0)
298 flags = IO_NORMAL | (needextclean ? IO_EXT: 0);
300 softdep_journal_freeblocks(ip, cred, length,
303 softdep_setup_freeblocks(ip, length, flags);
304 ASSERT_VOP_LOCKED(vp, "ffs_truncate1");
305 if (journaltrunc == 0) {
306 ip->i_flag |= IN_CHANGE | IN_UPDATE;
307 error = ffs_update(vp, 0);
313 * Shorten the size of the file. If the file is not being
314 * truncated to a block boundary, the contents of the
315 * partial block following the end of the file must be
316 * zero'ed in case it ever becomes accessible again because
317 * of subsequent file growth. Directories however are not
318 * zero'ed as they should grow back initialized to empty.
320 offset = blkoff(fs, length);
323 DIP_SET(ip, i_size, length);
325 lbn = lblkno(fs, length);
327 error = UFS_BALLOC(vp, length - 1, 1, cred, flags, &bp);
331 * When we are doing soft updates and the UFS_BALLOC
332 * above fills in a direct block hole with a full sized
333 * block that will be truncated down to a fragment below,
334 * we must flush out the block dependency with an FSYNC
335 * so that we do not get a soft updates inconsistency
336 * when we create the fragment below.
338 if (DOINGSOFTDEP(vp) && lbn < NDADDR &&
339 fragroundup(fs, blkoff(fs, length)) < fs->fs_bsize &&
340 (error = ffs_syncvnode(vp, MNT_WAIT)) != 0)
343 DIP_SET(ip, i_size, length);
344 size = blksize(fs, ip, lbn);
345 if (vp->v_type != VDIR)
346 bzero((char *)bp->b_data + offset,
347 (u_int)(size - offset));
348 /* Kirk's code has reallocbuf(bp, size, 1) here */
350 if (bp->b_bufsize == fs->fs_bsize)
351 bp->b_flags |= B_CLUSTEROK;
358 * Calculate index into inode's block list of
359 * last direct and indirect blocks (if any)
360 * which we want to keep. Lastblock is -1 when
361 * the file is truncated to 0.
363 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
364 lastiblock[SINGLE] = lastblock - NDADDR;
365 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
366 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
367 nblocks = btodb(fs->fs_bsize);
369 * Update file and block pointers on disk before we start freeing
370 * blocks. If we crash before free'ing blocks below, the blocks
371 * will be returned to the free list. lastiblock values are also
372 * normalized to -1 for calls to ffs_indirtrunc below.
374 for (level = TRIPLE; level >= SINGLE; level--) {
375 oldblks[NDADDR + level] = DIP(ip, i_ib[level]);
376 if (lastiblock[level] < 0) {
377 DIP_SET(ip, i_ib[level], 0);
378 lastiblock[level] = -1;
381 for (i = 0; i < NDADDR; i++) {
382 oldblks[i] = DIP(ip, i_db[i]);
384 DIP_SET(ip, i_db[i], 0);
386 ip->i_flag |= IN_CHANGE | IN_UPDATE;
387 allerror = ffs_update(vp, 1);
390 * Having written the new inode to disk, save its new configuration
391 * and put back the old block pointers long enough to process them.
392 * Note that we save the new block configuration so we can check it
395 for (i = 0; i < NDADDR; i++) {
396 newblks[i] = DIP(ip, i_db[i]);
397 DIP_SET(ip, i_db[i], oldblks[i]);
399 for (i = 0; i < NIADDR; i++) {
400 newblks[NDADDR + i] = DIP(ip, i_ib[i]);
401 DIP_SET(ip, i_ib[i], oldblks[NDADDR + i]);
404 DIP_SET(ip, i_size, osize);
406 error = vtruncbuf(vp, cred, td, length, fs->fs_bsize);
407 if (error && (allerror == 0))
411 * Indirect blocks first.
413 indir_lbn[SINGLE] = -NDADDR;
414 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
415 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
416 for (level = TRIPLE; level >= SINGLE; level--) {
417 bn = DIP(ip, i_ib[level]);
419 error = ffs_indirtrunc(ip, indir_lbn[level],
420 fsbtodb(fs, bn), lastiblock[level], level, &count);
423 blocksreleased += count;
424 if (lastiblock[level] < 0) {
425 DIP_SET(ip, i_ib[level], 0);
426 ffs_blkfree(ump, fs, ip->i_devvp, bn,
427 fs->fs_bsize, ip->i_number,
429 blocksreleased += nblocks;
432 if (lastiblock[level] >= 0)
437 * All whole direct blocks or frags.
439 for (i = NDADDR - 1; i > lastblock; i--) {
442 bn = DIP(ip, i_db[i]);
445 DIP_SET(ip, i_db[i], 0);
446 bsize = blksize(fs, ip, i);
447 ffs_blkfree(ump, fs, ip->i_devvp, bn, bsize, ip->i_number,
449 blocksreleased += btodb(bsize);
455 * Finally, look for a change in size of the
456 * last direct block; release any frags.
458 bn = DIP(ip, i_db[lastblock]);
460 long oldspace, newspace;
463 * Calculate amount of space we're giving
464 * back as old block size minus new block size.
466 oldspace = blksize(fs, ip, lastblock);
468 DIP_SET(ip, i_size, length);
469 newspace = blksize(fs, ip, lastblock);
471 panic("ffs_truncate: newspace");
472 if (oldspace - newspace > 0) {
474 * Block number of space to be free'd is
475 * the old block # plus the number of frags
476 * required for the storage we're keeping.
478 bn += numfrags(fs, newspace);
479 ffs_blkfree(ump, fs, ip->i_devvp, bn,
480 oldspace - newspace, ip->i_number, vp->v_type, NULL);
481 blocksreleased += btodb(oldspace - newspace);
486 for (level = SINGLE; level <= TRIPLE; level++)
487 if (newblks[NDADDR + level] != DIP(ip, i_ib[level]))
488 panic("ffs_truncate1");
489 for (i = 0; i < NDADDR; i++)
490 if (newblks[i] != DIP(ip, i_db[i]))
491 panic("ffs_truncate2");
494 (fs->fs_magic != FS_UFS2_MAGIC || ip->i_din2->di_extsize == 0) &&
495 (bo->bo_dirty.bv_cnt > 0 || bo->bo_clean.bv_cnt > 0))
496 panic("ffs_truncate3");
498 #endif /* INVARIANTS */
500 * Put back the real size.
503 DIP_SET(ip, i_size, length);
504 DIP_SET(ip, i_blocks, DIP(ip, i_blocks) - blocksreleased);
506 if (DIP(ip, i_blocks) < 0) /* sanity */
507 DIP_SET(ip, i_blocks, 0);
508 ip->i_flag |= IN_CHANGE;
510 (void) chkdq(ip, -blocksreleased, NOCRED, 0);
516 softdep_journal_freeblocks(ip, cred, length, IO_EXT);
518 softdep_setup_freeblocks(ip, length, IO_EXT);
519 return ffs_update(vp, MNT_WAIT);
523 * Release blocks associated with the inode ip and stored in the indirect
524 * block bn. Blocks are free'd in LIFO order up to (but not including)
525 * lastbn. If level is greater than SINGLE, the block is an indirect block
526 * and recursive calls to indirtrunc must be used to cleanse other indirect
530 ffs_indirtrunc(ip, lbn, dbn, lastbn, level, countp)
532 ufs2_daddr_t lbn, lastbn;
535 ufs2_daddr_t *countp;
538 struct fs *fs = ip->i_fs;
541 int i, nblocks, error = 0, allerror = 0;
542 ufs2_daddr_t nb, nlbn, last;
543 ufs2_daddr_t blkcount, factor, blocksreleased = 0;
544 ufs1_daddr_t *bap1 = NULL;
545 ufs2_daddr_t *bap2 = NULL;
546 # define BAP(ip, i) (((ip)->i_ump->um_fstype == UFS1) ? bap1[i] : bap2[i])
549 * Calculate index in current block of last
550 * block to be kept. -1 indicates the entire
551 * block so we need not calculate the index.
553 factor = lbn_offset(fs, level);
557 nblocks = btodb(fs->fs_bsize);
559 * Get buffer of block pointers, zero those entries corresponding
560 * to blocks to be free'd, and update on disk copy first. Since
561 * double(triple) indirect before single(double) indirect, calls
562 * to bmap on these blocks will fail. However, we already have
563 * the on disk address, so we have to set the b_blkno field
564 * explicitly instead of letting bread do everything for us.
567 bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0, 0);
568 if ((bp->b_flags & B_CACHE) == 0) {
569 curthread->td_ru.ru_inblock++; /* pay for read */
570 bp->b_iocmd = BIO_READ;
571 bp->b_flags &= ~B_INVAL;
572 bp->b_ioflags &= ~BIO_ERROR;
573 if (bp->b_bcount > bp->b_bufsize)
574 panic("ffs_indirtrunc: bad buffer size");
576 vfs_busy_pages(bp, 0);
577 bp->b_iooffset = dbtob(bp->b_blkno);
587 if (ip->i_ump->um_fstype == UFS1)
588 bap1 = (ufs1_daddr_t *)bp->b_data;
590 bap2 = (ufs2_daddr_t *)bp->b_data;
592 copy = malloc(fs->fs_bsize, M_TEMP, M_WAITOK);
593 bcopy((caddr_t)bp->b_data, copy, (u_int)fs->fs_bsize);
594 for (i = last + 1; i < NINDIR(fs); i++)
595 if (ip->i_ump->um_fstype == UFS1)
599 if (DOINGASYNC(vp)) {
606 if (ip->i_ump->um_fstype == UFS1)
607 bap1 = (ufs1_daddr_t *)copy;
609 bap2 = (ufs2_daddr_t *)copy;
613 * Recursively free totally unused blocks.
615 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
616 i--, nlbn += factor) {
620 if (level > SINGLE) {
621 if ((error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
622 (ufs2_daddr_t)-1, level - 1, &blkcount)) != 0)
624 blocksreleased += blkcount;
626 ffs_blkfree(ip->i_ump, fs, ip->i_devvp, nb, fs->fs_bsize,
627 ip->i_number, vp->v_type, NULL);
628 blocksreleased += nblocks;
632 * Recursively free last partial block.
634 if (level > SINGLE && lastbn >= 0) {
635 last = lastbn % factor;
638 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
639 last, level - 1, &blkcount);
642 blocksreleased += blkcount;
648 bp->b_flags |= B_INVAL | B_NOCACHE;
652 *countp = blocksreleased;
657 ffs_rdonly(struct inode *ip)
660 return (ip->i_ump->um_fs->fs_ronly != 0);