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);
96 * Ensure that uid and gid are correct. This is a temporary
97 * fix until fsck has been changed to do the update.
99 if (fs->fs_magic == FS_UFS1_MAGIC && /* XXX */
100 fs->fs_old_inodefmt < FS_44INODEFMT) { /* XXX */
101 ip->i_din1->di_ouid = ip->i_uid; /* XXX */
102 ip->i_din1->di_ogid = ip->i_gid; /* XXX */
104 error = bread(ip->i_devvp, fsbtodb(fs, ino_to_fsba(fs, ip->i_number)),
105 (int)fs->fs_bsize, NOCRED, &bp);
110 if (DOINGSOFTDEP(vp))
111 softdep_update_inodeblock(ip, bp, waitfor);
112 else if (ip->i_effnlink != ip->i_nlink)
113 panic("ffs_update: bad link cnt");
114 if (ip->i_ump->um_fstype == UFS1)
115 *((struct ufs1_dinode *)bp->b_data +
116 ino_to_fsbo(fs, ip->i_number)) = *ip->i_din1;
118 *((struct ufs2_dinode *)bp->b_data +
119 ino_to_fsbo(fs, ip->i_number)) = *ip->i_din2;
120 if (waitfor && !DOINGASYNC(vp)) {
122 } else if (vm_page_count_severe() || buf_dirty_count_severe()) {
125 if (bp->b_bufsize == fs->fs_bsize)
126 bp->b_flags |= B_CLUSTEROK;
132 #define SINGLE 0 /* index of single indirect block */
133 #define DOUBLE 1 /* index of double indirect block */
134 #define TRIPLE 2 /* index of triple indirect block */
136 * Truncate the inode ip to at most length size, freeing the
140 ffs_truncate(vp, length, flags, cred, td)
148 ufs2_daddr_t bn, lbn, lastblock, lastiblock[NIADDR], indir_lbn[NIADDR];
149 ufs2_daddr_t oldblks[NDADDR + NIADDR], newblks[NDADDR + NIADDR];
150 ufs2_daddr_t count, blocksreleased = 0, datablocks;
154 struct ufsmount *ump;
155 int needextclean, softdepslowdown, extblocks;
156 int offset, size, level, nblocks;
157 int i, error, allerror;
165 ASSERT_VOP_LOCKED(vp, "ffs_truncate");
170 * Historically clients did not have to specify which data
171 * they were truncating. So, if not specified, we assume
172 * traditional behavior, e.g., just the normal data.
174 if ((flags & (IO_EXT | IO_NORMAL)) == 0)
177 * If we are truncating the extended-attributes, and cannot
178 * do it with soft updates, then do it slowly here. If we are
179 * truncating both the extended attributes and the file contents
180 * (e.g., the file is being unlinked), then pick it off with
181 * soft updates below.
184 softdepslowdown = DOINGSOFTDEP(vp) && softdep_slowdown(vp);
186 datablocks = DIP(ip, i_blocks);
187 if (fs->fs_magic == FS_UFS2_MAGIC && ip->i_din2->di_extsize > 0) {
188 extblocks = btodb(fragroundup(fs, ip->i_din2->di_extsize));
189 datablocks -= extblocks;
191 if ((flags & IO_EXT) && extblocks > 0) {
192 if (DOINGSOFTDEP(vp) && softdepslowdown == 0 && length == 0) {
193 if ((flags & IO_NORMAL) == 0) {
194 softdep_setup_freeblocks(ip, length, IO_EXT);
200 panic("ffs_truncate: partial trunc of extdata");
201 if ((error = ffs_syncvnode(vp, MNT_WAIT)) != 0)
203 osize = ip->i_din2->di_extsize;
204 ip->i_din2->di_blocks -= extblocks;
206 (void) chkdq(ip, -extblocks, NOCRED, 0);
208 vinvalbuf(vp, V_ALT, 0, 0);
210 OFF_TO_IDX(lblktosize(fs, -extblocks)), 0);
211 ip->i_din2->di_extsize = 0;
212 for (i = 0; i < NXADDR; i++) {
213 oldblks[i] = ip->i_din2->di_extb[i];
214 ip->i_din2->di_extb[i] = 0;
216 ip->i_flag |= IN_CHANGE;
217 if ((error = ffs_update(vp, 1)))
219 for (i = 0; i < NXADDR; i++) {
222 ffs_blkfree(ump, fs, ip->i_devvp, oldblks[i],
223 sblksize(fs, osize, i), ip->i_number);
227 if ((flags & IO_NORMAL) == 0)
229 if (length > fs->fs_maxfilesize)
231 if (vp->v_type == VLNK &&
232 (ip->i_size < vp->v_mount->mnt_maxsymlinklen ||
236 panic("ffs_truncate: partial truncate of symlink");
238 bzero(SHORTLINK(ip), (u_int)ip->i_size);
240 DIP_SET(ip, i_size, 0);
241 ip->i_flag |= IN_CHANGE | IN_UPDATE;
243 softdep_setup_freeblocks(ip, length, IO_EXT);
244 return (ffs_update(vp, 1));
246 if (ip->i_size == length) {
247 ip->i_flag |= IN_CHANGE | IN_UPDATE;
249 softdep_setup_freeblocks(ip, length, IO_EXT);
250 return (ffs_update(vp, 0));
253 panic("ffs_truncate: read-only filesystem");
255 error = getinoquota(ip);
259 if ((ip->i_flags & SF_SNAPSHOT) != 0)
261 vp->v_lasta = vp->v_clen = vp->v_cstart = vp->v_lastw = 0;
262 if (DOINGSOFTDEP(vp)) {
263 if (length > 0 || softdepslowdown) {
265 * If a file is only partially truncated, then
266 * we have to clean up the data structures
267 * describing the allocation past the truncation
268 * point. Finding and deallocating those structures
269 * is a lot of work. Since partial truncation occurs
270 * rarely, we solve the problem by syncing the file
271 * so that it will have no data structures left.
273 if ((error = ffs_syncvnode(vp, MNT_WAIT)) != 0)
276 if (ip->i_flag & IN_SPACECOUNTED)
277 fs->fs_pendingblocks -= datablocks;
281 (void) chkdq(ip, -datablocks, NOCRED, 0);
283 softdep_setup_freeblocks(ip, length, needextclean ?
284 IO_EXT | IO_NORMAL : IO_NORMAL);
285 ASSERT_VOP_LOCKED(vp, "ffs_truncate1");
286 vinvalbuf(vp, needextclean ? 0 : V_NORMAL, 0, 0);
288 vn_pages_remove(vp, 0,
289 OFF_TO_IDX(lblktosize(fs, -extblocks)));
290 vnode_pager_setsize(vp, 0);
291 ip->i_flag |= IN_CHANGE | IN_UPDATE;
292 return (ffs_update(vp, 0));
297 * Lengthen the size of the file. We must ensure that the
298 * last byte of the file is allocated. Since the smallest
299 * value of osize is 0, length will be at least 1.
301 if (osize < length) {
302 vnode_pager_setsize(vp, length);
304 error = UFS_BALLOC(vp, length - 1, 1, cred, flags, &bp);
306 vnode_pager_setsize(vp, osize);
310 DIP_SET(ip, i_size, length);
311 if (bp->b_bufsize == fs->fs_bsize)
312 bp->b_flags |= B_CLUSTEROK;
317 ip->i_flag |= IN_CHANGE | IN_UPDATE;
318 return (ffs_update(vp, 1));
321 * Shorten the size of the file. If the file is not being
322 * truncated to a block boundary, the contents of the
323 * partial block following the end of the file must be
324 * zero'ed in case it ever becomes accessible again because
325 * of subsequent file growth. Directories however are not
326 * zero'ed as they should grow back initialized to empty.
328 offset = blkoff(fs, length);
331 DIP_SET(ip, i_size, length);
333 lbn = lblkno(fs, length);
335 error = UFS_BALLOC(vp, length - 1, 1, cred, flags, &bp);
340 * When we are doing soft updates and the UFS_BALLOC
341 * above fills in a direct block hole with a full sized
342 * block that will be truncated down to a fragment below,
343 * we must flush out the block dependency with an FSYNC
344 * so that we do not get a soft updates inconsistency
345 * when we create the fragment below.
347 if (DOINGSOFTDEP(vp) && lbn < NDADDR &&
348 fragroundup(fs, blkoff(fs, length)) < fs->fs_bsize &&
349 (error = ffs_syncvnode(vp, MNT_WAIT)) != 0)
352 DIP_SET(ip, i_size, length);
353 size = blksize(fs, ip, lbn);
354 if (vp->v_type != VDIR)
355 bzero((char *)bp->b_data + offset,
356 (u_int)(size - offset));
357 /* Kirk's code has reallocbuf(bp, size, 1) here */
359 if (bp->b_bufsize == fs->fs_bsize)
360 bp->b_flags |= B_CLUSTEROK;
367 * Calculate index into inode's block list of
368 * last direct and indirect blocks (if any)
369 * which we want to keep. Lastblock is -1 when
370 * the file is truncated to 0.
372 lastblock = lblkno(fs, length + fs->fs_bsize - 1) - 1;
373 lastiblock[SINGLE] = lastblock - NDADDR;
374 lastiblock[DOUBLE] = lastiblock[SINGLE] - NINDIR(fs);
375 lastiblock[TRIPLE] = lastiblock[DOUBLE] - NINDIR(fs) * NINDIR(fs);
376 nblocks = btodb(fs->fs_bsize);
378 * Update file and block pointers on disk before we start freeing
379 * blocks. If we crash before free'ing blocks below, the blocks
380 * will be returned to the free list. lastiblock values are also
381 * normalized to -1 for calls to ffs_indirtrunc below.
383 for (level = TRIPLE; level >= SINGLE; level--) {
384 oldblks[NDADDR + level] = DIP(ip, i_ib[level]);
385 if (lastiblock[level] < 0) {
386 DIP_SET(ip, i_ib[level], 0);
387 lastiblock[level] = -1;
390 for (i = 0; i < NDADDR; i++) {
391 oldblks[i] = DIP(ip, i_db[i]);
393 DIP_SET(ip, i_db[i], 0);
395 ip->i_flag |= IN_CHANGE | IN_UPDATE;
396 allerror = ffs_update(vp, 1);
399 * Having written the new inode to disk, save its new configuration
400 * and put back the old block pointers long enough to process them.
401 * Note that we save the new block configuration so we can check it
404 for (i = 0; i < NDADDR; i++) {
405 newblks[i] = DIP(ip, i_db[i]);
406 DIP_SET(ip, i_db[i], oldblks[i]);
408 for (i = 0; i < NIADDR; i++) {
409 newblks[NDADDR + i] = DIP(ip, i_ib[i]);
410 DIP_SET(ip, i_ib[i], oldblks[NDADDR + i]);
413 DIP_SET(ip, i_size, osize);
415 error = vtruncbuf(vp, cred, td, length, fs->fs_bsize);
416 if (error && (allerror == 0))
420 * Indirect blocks first.
422 indir_lbn[SINGLE] = -NDADDR;
423 indir_lbn[DOUBLE] = indir_lbn[SINGLE] - NINDIR(fs) - 1;
424 indir_lbn[TRIPLE] = indir_lbn[DOUBLE] - NINDIR(fs) * NINDIR(fs) - 1;
425 for (level = TRIPLE; level >= SINGLE; level--) {
426 bn = DIP(ip, i_ib[level]);
428 error = ffs_indirtrunc(ip, indir_lbn[level],
429 fsbtodb(fs, bn), lastiblock[level], level, &count);
432 blocksreleased += count;
433 if (lastiblock[level] < 0) {
434 DIP_SET(ip, i_ib[level], 0);
435 ffs_blkfree(ump, fs, ip->i_devvp, bn,
436 fs->fs_bsize, ip->i_number);
437 blocksreleased += nblocks;
440 if (lastiblock[level] >= 0)
445 * All whole direct blocks or frags.
447 for (i = NDADDR - 1; i > lastblock; i--) {
450 bn = DIP(ip, i_db[i]);
453 DIP_SET(ip, i_db[i], 0);
454 bsize = blksize(fs, ip, i);
455 ffs_blkfree(ump, fs, ip->i_devvp, bn, bsize, ip->i_number);
456 blocksreleased += btodb(bsize);
462 * Finally, look for a change in size of the
463 * last direct block; release any frags.
465 bn = DIP(ip, i_db[lastblock]);
467 long oldspace, newspace;
470 * Calculate amount of space we're giving
471 * back as old block size minus new block size.
473 oldspace = blksize(fs, ip, lastblock);
475 DIP_SET(ip, i_size, length);
476 newspace = blksize(fs, ip, lastblock);
478 panic("ffs_truncate: newspace");
479 if (oldspace - newspace > 0) {
481 * Block number of space to be free'd is
482 * the old block # plus the number of frags
483 * required for the storage we're keeping.
485 bn += numfrags(fs, newspace);
486 ffs_blkfree(ump, fs, ip->i_devvp, bn,
487 oldspace - newspace, ip->i_number);
488 blocksreleased += btodb(oldspace - newspace);
493 for (level = SINGLE; level <= TRIPLE; level++)
494 if (newblks[NDADDR + level] != DIP(ip, i_ib[level]))
495 panic("ffs_truncate1");
496 for (i = 0; i < NDADDR; i++)
497 if (newblks[i] != DIP(ip, i_db[i]))
498 panic("ffs_truncate2");
501 (fs->fs_magic != FS_UFS2_MAGIC || ip->i_din2->di_extsize == 0) &&
502 (bo->bo_dirty.bv_cnt > 0 || bo->bo_clean.bv_cnt > 0))
503 panic("ffs_truncate3");
505 #endif /* INVARIANTS */
507 * Put back the real size.
510 DIP_SET(ip, i_size, length);
511 DIP_SET(ip, i_blocks, DIP(ip, i_blocks) - blocksreleased);
513 if (DIP(ip, i_blocks) < 0) /* sanity */
514 DIP_SET(ip, i_blocks, 0);
515 ip->i_flag |= IN_CHANGE;
517 (void) chkdq(ip, -blocksreleased, NOCRED, 0);
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.
554 for (i = SINGLE; i < level; i++)
555 factor *= NINDIR(fs);
559 nblocks = btodb(fs->fs_bsize);
561 * Get buffer of block pointers, zero those entries corresponding
562 * to blocks to be free'd, and update on disk copy first. Since
563 * double(triple) indirect before single(double) indirect, calls
564 * to bmap on these blocks will fail. However, we already have
565 * the on disk address, so we have to set the b_blkno field
566 * explicitly instead of letting bread do everything for us.
569 bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0, 0);
570 if ((bp->b_flags & B_CACHE) == 0) {
571 curthread->td_ru.ru_inblock++; /* pay for read */
572 bp->b_iocmd = BIO_READ;
573 bp->b_flags &= ~B_INVAL;
574 bp->b_ioflags &= ~BIO_ERROR;
575 if (bp->b_bcount > bp->b_bufsize)
576 panic("ffs_indirtrunc: bad buffer size");
578 vfs_busy_pages(bp, 0);
579 bp->b_iooffset = dbtob(bp->b_blkno);
589 if (ip->i_ump->um_fstype == UFS1)
590 bap1 = (ufs1_daddr_t *)bp->b_data;
592 bap2 = (ufs2_daddr_t *)bp->b_data;
594 copy = malloc(fs->fs_bsize, M_TEMP, M_WAITOK);
595 bcopy((caddr_t)bp->b_data, copy, (u_int)fs->fs_bsize);
596 for (i = last + 1; i < NINDIR(fs); i++)
597 if (ip->i_ump->um_fstype == UFS1)
601 if (DOINGASYNC(vp)) {
608 if (ip->i_ump->um_fstype == UFS1)
609 bap1 = (ufs1_daddr_t *)copy;
611 bap2 = (ufs2_daddr_t *)copy;
615 * Recursively free totally unused blocks.
617 for (i = NINDIR(fs) - 1, nlbn = lbn + 1 - i * factor; i > last;
618 i--, nlbn += factor) {
622 if (level > SINGLE) {
623 if ((error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
624 (ufs2_daddr_t)-1, level - 1, &blkcount)) != 0)
626 blocksreleased += blkcount;
628 ffs_blkfree(ip->i_ump, fs, ip->i_devvp, nb, fs->fs_bsize,
630 blocksreleased += nblocks;
634 * Recursively free last partial block.
636 if (level > SINGLE && lastbn >= 0) {
637 last = lastbn % factor;
640 error = ffs_indirtrunc(ip, nlbn, fsbtodb(fs, nb),
641 last, level - 1, &blkcount);
644 blocksreleased += blkcount;
650 bp->b_flags |= B_INVAL | B_NOCACHE;
654 *countp = blocksreleased;
659 ffs_rdonly(struct inode *ip)
662 return (ip->i_ump->um_fs->fs_ronly != 0);