1 /* $NetBSD: ffs_alloc.c,v 1.14 2004/06/20 22:20:18 jmc Exp $ */
2 /* From: NetBSD: ffs_alloc.c,v 1.50 2001/09/06 02:16:01 lukem Exp */
5 * Copyright (c) 2002 Networks Associates Technology, Inc.
8 * This software was developed for the FreeBSD Project by Marshall
9 * Kirk McKusick and Network Associates Laboratories, the Security
10 * Research Division of Network Associates, Inc. under DARPA/SPAWAR
11 * contract N66001-01-C-8035 ("CBOSS"), as part of the DARPA CHATS
14 * Copyright (c) 1982, 1986, 1989, 1993
15 * The Regents of the University of California. All rights reserved.
17 * Redistribution and use in source and binary forms, with or without
18 * modification, are permitted provided that the following conditions
20 * 1. Redistributions of source code must retain the above copyright
21 * notice, this list of conditions and the following disclaimer.
22 * 2. Redistributions in binary form must reproduce the above copyright
23 * notice, this list of conditions and the following disclaimer in the
24 * documentation and/or other materials provided with the distribution.
25 * 3. Neither the name of the University nor the names of its contributors
26 * may be used to endorse or promote products derived from this software
27 * without specific prior written permission.
29 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
30 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
31 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
32 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
33 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
34 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
35 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
36 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
37 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
38 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
41 * @(#)ffs_alloc.c 8.19 (Berkeley) 7/13/95
44 #include <sys/cdefs.h>
45 #if defined(__RCSID) && !defined(__lint)
46 __RCSID("$NetBSD: ffs_alloc.c,v 1.14 2004/06/20 22:20:18 jmc Exp $");
49 #include <sys/param.h>
56 #include <ufs/ufs/dinode.h>
57 #include <ufs/ffs/fs.h>
59 #include "ffs/ufs_bswap.h"
61 #include "ffs/ufs_inode.h"
62 #include "ffs/ffs_extern.h"
64 static int scanc(u_int, const u_char *, const u_char *, int);
66 static daddr_t ffs_alloccg(struct inode *, int, daddr_t, int);
67 static daddr_t ffs_alloccgblk(struct inode *, struct buf *, daddr_t);
68 static daddr_t ffs_hashalloc(struct inode *, int, daddr_t, int,
69 daddr_t (*)(struct inode *, int, daddr_t, int));
70 static int32_t ffs_mapsearch(struct fs *, struct cg *, daddr_t, int);
73 * Allocate a block in the file system.
75 * The size of the requested block is given, which must be some
76 * multiple of fs_fsize and <= fs_bsize.
77 * A preference may be optionally specified. If a preference is given
78 * the following hierarchy is used to allocate a block:
79 * 1) allocate the requested block.
80 * 2) allocate a rotationally optimal block in the same cylinder.
81 * 3) allocate a block in the same cylinder group.
82 * 4) quadradically rehash into other cylinder groups, until an
83 * available block is located.
84 * If no block preference is given the following hierarchy is used
85 * to allocate a block:
86 * 1) allocate a block in the cylinder group that contains the
88 * 2) quadradically rehash into other cylinder groups, until an
89 * available block is located.
92 ffs_alloc(struct inode *ip, daddr_t lbn, daddr_t bpref, int size,
95 struct fs *fs = ip->i_fs;
100 if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) {
101 errx(1, "ffs_alloc: bad size: bsize %d size %d",
104 if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0)
106 if (bpref >= fs->fs_size)
109 cg = ino_to_cg(fs, ip->i_number);
111 cg = dtog(fs, bpref);
112 bno = ffs_hashalloc(ip, cg, bpref, size, ffs_alloccg);
114 if (ip->i_fs->fs_magic == FS_UFS1_MAGIC)
115 ip->i_ffs1_blocks += size / DEV_BSIZE;
117 ip->i_ffs2_blocks += size / DEV_BSIZE;
126 * Select the desired position for the next block in a file. The file is
127 * logically divided into sections. The first section is composed of the
128 * direct blocks. Each additional section contains fs_maxbpg blocks.
130 * If no blocks have been allocated in the first section, the policy is to
131 * request a block in the same cylinder group as the inode that describes
132 * the file. If no blocks have been allocated in any other section, the
133 * policy is to place the section in a cylinder group with a greater than
134 * average number of free blocks. An appropriate cylinder group is found
135 * by using a rotor that sweeps the cylinder groups. When a new group of
136 * blocks is needed, the sweep begins in the cylinder group following the
137 * cylinder group from which the previous allocation was made. The sweep
138 * continues until a cylinder group with greater than the average number
139 * of free blocks is found. If the allocation is for the first block in an
140 * indirect block, the information on the previous allocation is unavailable;
141 * here a best guess is made based upon the logical block number being
144 * If a section is already partially allocated, the policy is to
145 * contiguously allocate fs_maxcontig blocks. The end of one of these
146 * contiguous blocks and the beginning of the next is physically separated
147 * so that the disk head will be in transit between them for at least
148 * fs_rotdelay milliseconds. This is to allow time for the processor to
149 * schedule another I/O transfer.
153 ffs_blkpref_ufs1(struct inode *ip, daddr_t lbn, int indx, int32_t *bap)
157 int avgbfree, startcg;
160 if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
161 if (lbn < NDADDR + NINDIR(fs)) {
162 cg = ino_to_cg(fs, ip->i_number);
163 return (fs->fs_fpg * cg + fs->fs_frag);
166 * Find a cylinder with greater than average number of
167 * unused data blocks.
169 if (indx == 0 || bap[indx - 1] == 0)
171 ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg;
174 ufs_rw32(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + 1);
175 startcg %= fs->fs_ncg;
176 avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
177 for (cg = startcg; cg < fs->fs_ncg; cg++)
178 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree)
179 return (fs->fs_fpg * cg + fs->fs_frag);
180 for (cg = 0; cg <= startcg; cg++)
181 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree)
182 return (fs->fs_fpg * cg + fs->fs_frag);
186 * We just always try to lay things out contiguously.
188 return ufs_rw32(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + fs->fs_frag;
192 ffs_blkpref_ufs2(ip, lbn, indx, bap)
200 int avgbfree, startcg;
203 if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
204 if (lbn < NDADDR + NINDIR(fs)) {
205 cg = ino_to_cg(fs, ip->i_number);
206 return (fs->fs_fpg * cg + fs->fs_frag);
209 * Find a cylinder with greater than average number of
210 * unused data blocks.
212 if (indx == 0 || bap[indx - 1] == 0)
214 ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg;
217 ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + 1);
218 startcg %= fs->fs_ncg;
219 avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
220 for (cg = startcg; cg < fs->fs_ncg; cg++)
221 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
222 return (fs->fs_fpg * cg + fs->fs_frag);
224 for (cg = 0; cg < startcg; cg++)
225 if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
226 return (fs->fs_fpg * cg + fs->fs_frag);
231 * We just always try to lay things out contiguously.
233 return ufs_rw64(bap[indx - 1], UFS_FSNEEDSWAP(fs)) + fs->fs_frag;
237 * Implement the cylinder overflow algorithm.
239 * The policy implemented by this algorithm is:
240 * 1) allocate the block in its requested cylinder group.
241 * 2) quadradically rehash on the cylinder group number.
242 * 3) brute force search for a free block.
244 * `size': size for data blocks, mode for inodes
248 ffs_hashalloc(struct inode *ip, int cg, daddr_t pref, int size,
249 daddr_t (*allocator)(struct inode *, int, daddr_t, int))
257 * 1: preferred cylinder group
259 result = (*allocator)(ip, cg, pref, size);
263 * 2: quadratic rehash
265 for (i = 1; i < fs->fs_ncg; i *= 2) {
267 if (cg >= fs->fs_ncg)
269 result = (*allocator)(ip, cg, 0, size);
274 * 3: brute force search
275 * Note that we start at i == 2, since 0 was checked initially,
276 * and 1 is always checked in the quadratic rehash.
278 cg = (icg + 2) % fs->fs_ncg;
279 for (i = 2; i < fs->fs_ncg; i++) {
280 result = (*allocator)(ip, cg, 0, size);
284 if (cg == fs->fs_ncg)
291 * Determine whether a block can be allocated.
293 * Check to see if a block of the appropriate size is available,
294 * and if it is, allocate it.
297 ffs_alloccg(struct inode *ip, int cg, daddr_t bpref, int size)
302 int error, frags, allocsiz, i;
303 struct fs *fs = ip->i_fs;
304 const int needswap = UFS_FSNEEDSWAP(fs);
306 if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize)
308 error = bread(ip->i_fd, ip->i_fs, fsbtodb(fs, cgtod(fs, cg)),
309 (int)fs->fs_cgsize, &bp);
314 cgp = (struct cg *)bp->b_data;
315 if (!cg_chkmagic_swap(cgp, needswap) ||
316 (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize)) {
320 if (size == fs->fs_bsize) {
321 bno = ffs_alloccgblk(ip, bp, bpref);
326 * check to see if any fragments are already available
327 * allocsiz is the size which will be allocated, hacking
328 * it down to a smaller size if necessary
330 frags = numfrags(fs, size);
331 for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++)
332 if (cgp->cg_frsum[allocsiz] != 0)
334 if (allocsiz == fs->fs_frag) {
336 * no fragments were available, so a block will be
337 * allocated, and hacked up
339 if (cgp->cg_cs.cs_nbfree == 0) {
343 bno = ffs_alloccgblk(ip, bp, bpref);
344 bpref = dtogd(fs, bno);
345 for (i = frags; i < fs->fs_frag; i++)
346 setbit(cg_blksfree_swap(cgp, needswap), bpref + i);
347 i = fs->fs_frag - frags;
348 ufs_add32(cgp->cg_cs.cs_nffree, i, needswap);
349 fs->fs_cstotal.cs_nffree += i;
350 fs->fs_cs(fs, cg).cs_nffree += i;
352 ufs_add32(cgp->cg_frsum[i], 1, needswap);
356 bno = ffs_mapsearch(fs, cgp, bpref, allocsiz);
357 for (i = 0; i < frags; i++)
358 clrbit(cg_blksfree_swap(cgp, needswap), bno + i);
359 ufs_add32(cgp->cg_cs.cs_nffree, -frags, needswap);
360 fs->fs_cstotal.cs_nffree -= frags;
361 fs->fs_cs(fs, cg).cs_nffree -= frags;
363 ufs_add32(cgp->cg_frsum[allocsiz], -1, needswap);
364 if (frags != allocsiz)
365 ufs_add32(cgp->cg_frsum[allocsiz - frags], 1, needswap);
366 blkno = cg * fs->fs_fpg + bno;
372 * Allocate a block in a cylinder group.
374 * This algorithm implements the following policy:
375 * 1) allocate the requested block.
376 * 2) allocate a rotationally optimal block in the same cylinder.
377 * 3) allocate the next available block on the block rotor for the
378 * specified cylinder group.
379 * Note that this routine only allocates fs_bsize blocks; these
380 * blocks may be fragmented by the routine that allocates them.
383 ffs_alloccgblk(struct inode *ip, struct buf *bp, daddr_t bpref)
388 struct fs *fs = ip->i_fs;
389 const int needswap = UFS_FSNEEDSWAP(fs);
392 cgp = (struct cg *)bp->b_data;
393 blksfree = cg_blksfree_swap(cgp, needswap);
394 if (bpref == 0 || dtog(fs, bpref) != ufs_rw32(cgp->cg_cgx, needswap)) {
395 bpref = ufs_rw32(cgp->cg_rotor, needswap);
397 bpref = blknum(fs, bpref);
398 bno = dtogd(fs, bpref);
400 * if the requested block is available, use it
402 if (ffs_isblock(fs, blksfree, fragstoblks(fs, bno)))
406 * Take the next available one in this cylinder group.
408 bno = ffs_mapsearch(fs, cgp, bpref, (int)fs->fs_frag);
411 cgp->cg_rotor = ufs_rw32(bno, needswap);
413 blkno = fragstoblks(fs, bno);
414 ffs_clrblock(fs, blksfree, (long)blkno);
415 ffs_clusteracct(fs, cgp, blkno, -1);
416 ufs_add32(cgp->cg_cs.cs_nbfree, -1, needswap);
417 fs->fs_cstotal.cs_nbfree--;
418 fs->fs_cs(fs, ufs_rw32(cgp->cg_cgx, needswap)).cs_nbfree--;
420 blkno = ufs_rw32(cgp->cg_cgx, needswap) * fs->fs_fpg + bno;
425 * Free a block or fragment.
427 * The specified block or fragment is placed back in the
428 * free map. If a fragment is deallocated, a possible
429 * block reassembly is checked.
432 ffs_blkfree(struct inode *ip, daddr_t bno, long size)
436 int32_t fragno, cgbno;
437 int i, error, cg, blk, frags, bbase;
438 struct fs *fs = ip->i_fs;
439 const int needswap = UFS_FSNEEDSWAP(fs);
441 if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0 ||
442 fragnum(fs, bno) + numfrags(fs, size) > fs->fs_frag) {
443 errx(1, "blkfree: bad size: bno %lld bsize %d size %ld",
444 (long long)bno, fs->fs_bsize, size);
447 if (bno >= fs->fs_size) {
448 warnx("bad block %lld, ino %d", (long long)bno, ip->i_number);
451 error = bread(ip->i_fd, ip->i_fs, fsbtodb(fs, cgtod(fs, cg)),
452 (int)fs->fs_cgsize, &bp);
457 cgp = (struct cg *)bp->b_data;
458 if (!cg_chkmagic_swap(cgp, needswap)) {
462 cgbno = dtogd(fs, bno);
463 if (size == fs->fs_bsize) {
464 fragno = fragstoblks(fs, cgbno);
465 if (!ffs_isfreeblock(fs, cg_blksfree_swap(cgp, needswap), fragno)) {
466 errx(1, "blkfree: freeing free block %lld",
469 ffs_setblock(fs, cg_blksfree_swap(cgp, needswap), fragno);
470 ffs_clusteracct(fs, cgp, fragno, 1);
471 ufs_add32(cgp->cg_cs.cs_nbfree, 1, needswap);
472 fs->fs_cstotal.cs_nbfree++;
473 fs->fs_cs(fs, cg).cs_nbfree++;
475 bbase = cgbno - fragnum(fs, cgbno);
477 * decrement the counts associated with the old frags
479 blk = blkmap(fs, cg_blksfree_swap(cgp, needswap), bbase);
480 ffs_fragacct_swap(fs, blk, cgp->cg_frsum, -1, needswap);
482 * deallocate the fragment
484 frags = numfrags(fs, size);
485 for (i = 0; i < frags; i++) {
486 if (isset(cg_blksfree_swap(cgp, needswap), cgbno + i)) {
487 errx(1, "blkfree: freeing free frag: block %lld",
488 (long long)(cgbno + i));
490 setbit(cg_blksfree_swap(cgp, needswap), cgbno + i);
492 ufs_add32(cgp->cg_cs.cs_nffree, i, needswap);
493 fs->fs_cstotal.cs_nffree += i;
494 fs->fs_cs(fs, cg).cs_nffree += i;
496 * add back in counts associated with the new frags
498 blk = blkmap(fs, cg_blksfree_swap(cgp, needswap), bbase);
499 ffs_fragacct_swap(fs, blk, cgp->cg_frsum, 1, needswap);
501 * if a complete block has been reassembled, account for it
503 fragno = fragstoblks(fs, bbase);
504 if (ffs_isblock(fs, cg_blksfree_swap(cgp, needswap), fragno)) {
505 ufs_add32(cgp->cg_cs.cs_nffree, -fs->fs_frag, needswap);
506 fs->fs_cstotal.cs_nffree -= fs->fs_frag;
507 fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
508 ffs_clusteracct(fs, cgp, fragno, 1);
509 ufs_add32(cgp->cg_cs.cs_nbfree, 1, needswap);
510 fs->fs_cstotal.cs_nbfree++;
511 fs->fs_cs(fs, cg).cs_nbfree++;
520 scanc(u_int size, const u_char *cp, const u_char table[], int mask)
522 const u_char *end = &cp[size];
524 while (cp < end && (table[*cp] & mask) == 0)
530 * Find a block of the specified size in the specified cylinder group.
532 * It is a panic if a request is made to find a block if none are
536 ffs_mapsearch(struct fs *fs, struct cg *cgp, daddr_t bpref, int allocsiz)
539 int start, len, loc, i;
540 int blk, field, subfield, pos;
542 const int needswap = UFS_FSNEEDSWAP(fs);
545 * find the fragment by searching through the free block
546 * map for an appropriate bit pattern
549 start = dtogd(fs, bpref) / NBBY;
551 start = ufs_rw32(cgp->cg_frotor, needswap) / NBBY;
552 len = howmany(fs->fs_fpg, NBBY) - start;
555 loc = scanc((u_int)len,
556 (const u_char *)&cg_blksfree_swap(cgp, needswap)[start],
557 (const u_char *)fragtbl[fs->fs_frag],
558 (1 << (allocsiz - 1 + (fs->fs_frag % NBBY))));
562 loc = scanc((u_int)len,
563 (const u_char *)&cg_blksfree_swap(cgp, needswap)[0],
564 (const u_char *)fragtbl[fs->fs_frag],
565 (1 << (allocsiz - 1 + (fs->fs_frag % NBBY))));
568 "ffs_alloccg: map corrupted: start %d len %d offset %d %ld",
570 ufs_rw32(cgp->cg_freeoff, needswap),
571 (long)cg_blksfree_swap(cgp, needswap) - (long)cgp);
575 bno = (start + len - loc) * NBBY;
576 cgp->cg_frotor = ufs_rw32(bno, needswap);
578 * found the byte in the map
579 * sift through the bits to find the selected frag
581 for (i = bno + NBBY; bno < i; bno += fs->fs_frag) {
582 blk = blkmap(fs, cg_blksfree_swap(cgp, needswap), bno);
584 field = around[allocsiz];
585 subfield = inside[allocsiz];
586 for (pos = 0; pos <= fs->fs_frag - allocsiz; pos++) {
587 if ((blk & field) == subfield)
593 errx(1, "ffs_alloccg: block not in map: bno %lld", (long long)bno);
598 * Update the cluster map because of an allocation or free.
600 * Cnt == 1 means free; cnt == -1 means allocating.
603 ffs_clusteracct(struct fs *fs, struct cg *cgp, int32_t blkno, int cnt)
607 u_char *freemapp, *mapp;
608 int i, start, end, forw, back, map, bit;
609 const int needswap = UFS_FSNEEDSWAP(fs);
611 if (fs->fs_contigsumsize <= 0)
613 freemapp = cg_clustersfree_swap(cgp, needswap);
614 sump = cg_clustersum_swap(cgp, needswap);
616 * Allocate or clear the actual block.
619 setbit(freemapp, blkno);
621 clrbit(freemapp, blkno);
623 * Find the size of the cluster going forward.
626 end = start + fs->fs_contigsumsize;
627 if (end >= ufs_rw32(cgp->cg_nclusterblks, needswap))
628 end = ufs_rw32(cgp->cg_nclusterblks, needswap);
629 mapp = &freemapp[start / NBBY];
631 bit = 1 << (start % NBBY);
632 for (i = start; i < end; i++) {
633 if ((map & bit) == 0)
635 if ((i & (NBBY - 1)) != (NBBY - 1)) {
644 * Find the size of the cluster going backward.
647 end = start - fs->fs_contigsumsize;
650 mapp = &freemapp[start / NBBY];
652 bit = 1 << (start % NBBY);
653 for (i = start; i > end; i--) {
654 if ((map & bit) == 0)
656 if ((i & (NBBY - 1)) != 0) {
660 bit = 1 << (NBBY - 1);
665 * Account for old cluster and the possibly new forward and
669 if (i > fs->fs_contigsumsize)
670 i = fs->fs_contigsumsize;
671 ufs_add32(sump[i], cnt, needswap);
673 ufs_add32(sump[back], -cnt, needswap);
675 ufs_add32(sump[forw], -cnt, needswap);
678 * Update cluster summary information.
680 lp = &sump[fs->fs_contigsumsize];
681 for (i = fs->fs_contigsumsize; i > 0; i--)
682 if (ufs_rw32(*lp--, needswap) > 0)
684 fs->fs_maxcluster[ufs_rw32(cgp->cg_cgx, needswap)] = i;