Annotation of kernel/bsd/ufs/ffs/ffs_alloc.c, revision 1.1.1.1

1.1       root        1: /*
                      2:  * Copyright (c) 1999 Apple Computer, Inc. All rights reserved.
                      3:  *
                      4:  * @APPLE_LICENSE_HEADER_START@
                      5:  * 
                      6:  * Portions Copyright (c) 1999 Apple Computer, Inc.  All Rights
                      7:  * Reserved.  This file contains Original Code and/or Modifications of
                      8:  * Original Code as defined in and that are subject to the Apple Public
                      9:  * Source License Version 1.1 (the "License").  You may not use this file
                     10:  * except in compliance with the License.  Please obtain a copy of the
                     11:  * License at http://www.apple.com/publicsource and read it before using
                     12:  * this file.
                     13:  * 
                     14:  * The Original Code and all software distributed under the License are
                     15:  * distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, EITHER
                     16:  * EXPRESS OR IMPLIED, AND APPLE HEREBY DISCLAIMS ALL SUCH WARRANTIES,
                     17:  * INCLUDING WITHOUT LIMITATION, ANY WARRANTIES OF MERCHANTABILITY,
                     18:  * FITNESS FOR A PARTICULAR PURPOSE OR NON-INFRINGEMENT.  Please see the
                     19:  * License for the specific language governing rights and limitations
                     20:  * under the License.
                     21:  * 
                     22:  * @APPLE_LICENSE_HEADER_END@
                     23:  */
                     24: 
                     25: /* Copyright (c) 1995 NeXT Computer, Inc. All Rights Reserved */
                     26: /*
                     27:  * Copyright (c) 1982, 1986, 1989, 1993
                     28:  *     The Regents of the University of California.  All rights reserved.
                     29:  *
                     30:  * Redistribution and use in source and binary forms, with or without
                     31:  * modification, are permitted provided that the following conditions
                     32:  * are met:
                     33:  * 1. Redistributions of source code must retain the above copyright
                     34:  *    notice, this list of conditions and the following disclaimer.
                     35:  * 2. Redistributions in binary form must reproduce the above copyright
                     36:  *    notice, this list of conditions and the following disclaimer in the
                     37:  *    documentation and/or other materials provided with the distribution.
                     38:  * 3. All advertising materials mentioning features or use of this software
                     39:  *    must display the following acknowledgement:
                     40:  *     This product includes software developed by the University of
                     41:  *     California, Berkeley and its contributors.
                     42:  * 4. Neither the name of the University nor the names of its contributors
                     43:  *    may be used to endorse or promote products derived from this software
                     44:  *    without specific prior written permission.
                     45:  *
                     46:  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
                     47:  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
                     48:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
                     49:  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
                     50:  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
                     51:  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
                     52:  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
                     53:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
                     54:  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
                     55:  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
                     56:  * SUCH DAMAGE.
                     57:  *
                     58:  *     @(#)ffs_alloc.c 8.18 (Berkeley) 5/26/95
                     59:  */
                     60: #include <rev_endian_fs.h>
                     61: #include <vm/vm_pager.h>
                     62: #include <vm/vnode_pager.h>
                     63: 
                     64: #include <sys/param.h>
                     65: #include <sys/systm.h>
                     66: #include <sys/buf.h>
                     67: #include <sys/proc.h>
                     68: #include <sys/vnode.h>
                     69: #include <sys/mount.h>
                     70: #include <sys/kernel.h>
                     71: #include <sys/syslog.h>
                     72: 
                     73: #include <sys/vm.h>
                     74: 
                     75: #include <ufs/ufs/quota.h>
                     76: #include <ufs/ufs/inode.h>
                     77: 
                     78: #include <ufs/ffs/fs.h>
                     79: #include <ufs/ffs/ffs_extern.h>
                     80: 
                     81: #if REV_ENDIAN_FS
                     82: #include <ufs/ufs/ufs_byte_order.h>
                     83: #include <architecture/byte_order.h>
                     84: #endif /* REV_ENDIAN_FS */
                     85: 
                     86: extern u_long nextgennumber;
                     87: 
                     88: static ufs_daddr_t ffs_alloccg __P((struct inode *, int, ufs_daddr_t, int));
                     89: static ufs_daddr_t ffs_alloccgblk __P((struct fs *, struct cg *, ufs_daddr_t));
                     90: static ufs_daddr_t ffs_clusteralloc __P((struct inode *, int, ufs_daddr_t,
                     91:            int));
                     92: static ino_t   ffs_dirpref __P((struct fs *));
                     93: static ufs_daddr_t ffs_fragextend __P((struct inode *, int, long, int, int));
                     94: static void    ffs_fserr __P((struct fs *, u_int, char *));
                     95: static u_long  ffs_hashalloc
                     96:                    __P((struct inode *, int, long, int, u_int32_t (*)()));
                     97: static ino_t   ffs_nodealloccg __P((struct inode *, int, ufs_daddr_t, int));
                     98: static ufs_daddr_t ffs_mapsearch __P((struct fs *, struct cg *, ufs_daddr_t,
                     99:            int));
                    100: 
                    101: /*
                    102:  * Allocate a block in the file system.
                    103:  * 
                    104:  * The size of the requested block is given, which must be some
                    105:  * multiple of fs_fsize and <= fs_bsize.
                    106:  * A preference may be optionally specified. If a preference is given
                    107:  * the following hierarchy is used to allocate a block:
                    108:  *   1) allocate the requested block.
                    109:  *   2) allocate a rotationally optimal block in the same cylinder.
                    110:  *   3) allocate a block in the same cylinder group.
                    111:  *   4) quadradically rehash into other cylinder groups, until an
                    112:  *      available block is located.
                    113:  * If no block preference is given the following heirarchy is used
                    114:  * to allocate a block:
                    115:  *   1) allocate a block in the cylinder group that contains the
                    116:  *      inode for the file.
                    117:  *   2) quadradically rehash into other cylinder groups, until an
                    118:  *      available block is located.
                    119:  */
                    120: ffs_alloc(ip, lbn, bpref, size, cred, bnp)
                    121:        register struct inode *ip;
                    122:        ufs_daddr_t lbn, bpref;
                    123:        int size;
                    124:        struct ucred *cred;
                    125:        ufs_daddr_t *bnp;
                    126: {
                    127:        register struct fs *fs;
                    128:        ufs_daddr_t bno;
                    129:        int cg, error;
                    130: #if NeXT
                    131:        int devBlockSize=0;
                    132: #endif /* NeXT */
                    133:        *bnp = 0;
                    134:        fs = ip->i_fs;
                    135: #if DIAGNOSTIC
                    136:        if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) {
                    137:                printf("dev = 0x%x, bsize = %d, size = %d, fs = %s\n",
                    138:                    ip->i_dev, fs->fs_bsize, size, fs->fs_fsmnt);
                    139:                panic("ffs_alloc: bad size");
                    140:        }
                    141:        if (cred == NOCRED)
                    142:                panic("ffs_alloc: missing credential\n");
                    143: #endif /* DIAGNOSTIC */
                    144:        if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0)
                    145:                goto nospace;
                    146:        if (cred->cr_uid != 0 && freespace(fs, fs->fs_minfree) <= 0)
                    147:                goto nospace;
                    148: #ifdef NeXT
                    149:        VOP_DEVBLOCKSIZE(ip->i_devvp,&devBlockSize);
                    150: #endif /* NeXT */
                    151: #if QUOTA
                    152: #ifdef NeXT
                    153:         if (error = chkdq(ip, (long)btodb(size, devBlockSize), cred, 0))
                    154: #else
                    155:        if (error = chkdq(ip, (long)btodb(size), cred, 0))
                    156: #endif /* NeXT */
                    157:                return (error);
                    158: #endif /* QUOTA */
                    159:        if (bpref >= fs->fs_size)
                    160:                bpref = 0;
                    161:        if (bpref == 0)
                    162:                cg = ino_to_cg(fs, ip->i_number);
                    163:        else
                    164:                cg = dtog(fs, bpref);
                    165:        bno = (ufs_daddr_t)ffs_hashalloc(ip, cg, (long)bpref, size,
                    166:            (u_int32_t (*)())ffs_alloccg);
                    167:        if (bno > 0) {
                    168: #ifdef NeXT
                    169:                ip->i_blocks += btodb(size, devBlockSize);
                    170: #else
                    171:                ip->i_blocks += btodb(size);
                    172: #endif /* NeXT */
                    173:                ip->i_flag |= IN_CHANGE | IN_UPDATE;
                    174:                *bnp = bno;
                    175:                return (0);
                    176:        }
                    177: #if QUOTA
                    178:        /*
                    179:         * Restore user's disk quota because allocation failed.
                    180:         */
                    181: #ifdef NeXT
                    182:        (void) chkdq(ip, (long)-btodb(size, devBlockSize), cred, FORCE);
                    183: #else
                    184:        (void) chkdq(ip, (long)-btodb(size), cred, FORCE);
                    185: #endif /* NeXT */
                    186: #endif /* QUOTA */
                    187: nospace:
                    188:        ffs_fserr(fs, cred->cr_uid, "file system full");
                    189:        uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
                    190:        return (ENOSPC);
                    191: }
                    192: 
                    193: /*
                    194:  * Reallocate a fragment to a bigger size
                    195:  *
                    196:  * The number and size of the old block is given, and a preference
                    197:  * and new size is also specified. The allocator attempts to extend
                    198:  * the original block. Failing that, the regular block allocator is
                    199:  * invoked to get an appropriate block.
                    200:  */
                    201: ffs_realloccg(ip, lbprev, bpref, osize, nsize, cred, bpp)
                    202:        register struct inode *ip;
                    203:        ufs_daddr_t lbprev;
                    204:        ufs_daddr_t bpref;
                    205:        int osize, nsize;
                    206:        struct ucred *cred;
                    207:        struct buf **bpp;
                    208: {
                    209:        register struct fs *fs;
                    210:        struct buf *bp;
                    211:        int cg, request, error;
                    212:        ufs_daddr_t bprev, bno;
                    213: #ifdef NeXT
                    214:        int devBlockSize=0;
                    215: #endif
                    216:        
                    217:        *bpp = 0;
                    218:        fs = ip->i_fs;
                    219: #if DIAGNOSTIC
                    220:        if ((u_int)osize > fs->fs_bsize || fragoff(fs, osize) != 0 ||
                    221:            (u_int)nsize > fs->fs_bsize || fragoff(fs, nsize) != 0) {
                    222:                printf(
                    223:                    "dev = 0x%x, bsize = %d, osize = %d, nsize = %d, fs = %s\n",
                    224:                    ip->i_dev, fs->fs_bsize, osize, nsize, fs->fs_fsmnt);
                    225:                panic("ffs_realloccg: bad size");
                    226:        }
                    227:        if (cred == NOCRED)
                    228:                panic("ffs_realloccg: missing credential\n");
                    229: #endif /* DIAGNOSTIC */
                    230:        if (cred->cr_uid != 0 && freespace(fs, fs->fs_minfree) <= 0)
                    231:                goto nospace;
                    232:        if ((bprev = ip->i_db[lbprev]) == 0) {
                    233:                printf("dev = 0x%x, bsize = %d, bprev = %d, fs = %s\n",
                    234:                    ip->i_dev, fs->fs_bsize, bprev, fs->fs_fsmnt);
                    235:                panic("ffs_realloccg: bad bprev");
                    236:        }
                    237:        /*
                    238:         * Allocate the extra space in the buffer.
                    239:         */
                    240:        if (error = bread(ITOV(ip), lbprev, osize, NOCRED, &bp)) {
                    241:                brelse(bp);
                    242:                return (error);
                    243:        }
                    244: #ifdef NeXT
                    245:        VOP_DEVBLOCKSIZE(ip->i_devvp,&devBlockSize);
                    246: #endif /* NeXT */
                    247: 
                    248: #if QUOTA
                    249: #ifdef NeXT
                    250:        if (error = chkdq(ip, (long)btodb(nsize - osize, devBlockSize), cred, 0))
                    251: #else
                    252:        if (error = chkdq(ip, (long)btodb(nsize - osize), cred, 0))
                    253: #endif /* NeXT */
                    254:        {
                    255:                brelse(bp);
                    256:                return (error);
                    257:        }
                    258: #endif /* QUOTA */
                    259:        /*
                    260:         * Check for extension in the existing location.
                    261:         */
                    262:        cg = dtog(fs, bprev);
                    263:        if (bno = ffs_fragextend(ip, cg, (long)bprev, osize, nsize)) {
                    264:                if (bp->b_blkno != fsbtodb(fs, bno))
                    265:                        panic("bad blockno");
                    266: #ifdef NeXT
                    267:                ip->i_blocks += btodb(nsize - osize, devBlockSize);
                    268: #else
                    269:                ip->i_blocks += btodb(nsize - osize);
                    270: #endif /* NeXT */
                    271:                ip->i_flag |= IN_CHANGE | IN_UPDATE;
                    272:                allocbuf(bp, nsize);
                    273:                bp->b_flags |= B_DONE;
                    274:                bzero((char *)bp->b_data + osize, (u_int)nsize - osize);
                    275:                *bpp = bp;
                    276:                return (0);
                    277:        }
                    278:        /*
                    279:         * Allocate a new disk location.
                    280:         */
                    281:        if (bpref >= fs->fs_size)
                    282:                bpref = 0;
                    283:        switch ((int)fs->fs_optim) {
                    284:        case FS_OPTSPACE:
                    285:                /*
                    286:                 * Allocate an exact sized fragment. Although this makes 
                    287:                 * best use of space, we will waste time relocating it if 
                    288:                 * the file continues to grow. If the fragmentation is
                    289:                 * less than half of the minimum free reserve, we choose
                    290:                 * to begin optimizing for time.
                    291:                 */
                    292:                request = nsize;
                    293:                if (fs->fs_minfree < 5 ||
                    294:                    fs->fs_cstotal.cs_nffree >
                    295:                    fs->fs_dsize * fs->fs_minfree / (2 * 100))
                    296:                        break;
                    297:                log(LOG_NOTICE, "%s: optimization changed from SPACE to TIME\n",
                    298:                        fs->fs_fsmnt);
                    299:                fs->fs_optim = FS_OPTTIME;
                    300:                break;
                    301:        case FS_OPTTIME:
                    302:                /*
                    303:                 * At this point we have discovered a file that is trying to
                    304:                 * grow a small fragment to a larger fragment. To save time,
                    305:                 * we allocate a full sized block, then free the unused portion.
                    306:                 * If the file continues to grow, the `ffs_fragextend' call
                    307:                 * above will be able to grow it in place without further
                    308:                 * copying. If aberrant programs cause disk fragmentation to
                    309:                 * grow within 2% of the free reserve, we choose to begin
                    310:                 * optimizing for space.
                    311:                 */
                    312:                request = fs->fs_bsize;
                    313:                if (fs->fs_cstotal.cs_nffree <
                    314:                    fs->fs_dsize * (fs->fs_minfree - 2) / 100)
                    315:                        break;
                    316:                log(LOG_NOTICE, "%s: optimization changed from TIME to SPACE\n",
                    317:                        fs->fs_fsmnt);
                    318:                fs->fs_optim = FS_OPTSPACE;
                    319:                break;
                    320:        default:
                    321:                printf("dev = 0x%x, optim = %d, fs = %s\n",
                    322:                    ip->i_dev, fs->fs_optim, fs->fs_fsmnt);
                    323:                panic("ffs_realloccg: bad optim");
                    324:                /* NOTREACHED */
                    325:        }
                    326:        bno = (ufs_daddr_t)ffs_hashalloc(ip, cg, (long)bpref, request,
                    327:            (u_int32_t (*)())ffs_alloccg);
                    328:        if (bno > 0) {
                    329:                bp->b_blkno = fsbtodb(fs, bno);
                    330: #if MACH
                    331:                if (ITOV(ip)->v_vm_info != 0)
                    332:                        (void) vnode_uncache(ITOV(ip));
                    333: #endif
                    334:                ffs_blkfree(ip, bprev, (long)osize);
                    335:                if (nsize < request)
                    336:                        ffs_blkfree(ip, bno + numfrags(fs, nsize),
                    337:                            (long)(request - nsize));
                    338: #ifdef NeXT
                    339:                ip->i_blocks += btodb(nsize - osize, devBlockSize);
                    340: #else
                    341:                ip->i_blocks += btodb(nsize - osize);
                    342: #endif /* NeXT */
                    343:                ip->i_flag |= IN_CHANGE | IN_UPDATE;
                    344:                allocbuf(bp, nsize);
                    345:                bp->b_flags |= B_DONE;
                    346:                bzero((char *)bp->b_data + osize, (u_int)nsize - osize);
                    347:                *bpp = bp;
                    348:                return (0);
                    349:        }
                    350: #if QUOTA
                    351:        /*
                    352:         * Restore user's disk quota because allocation failed.
                    353:         */
                    354: #ifdef NeXT
                    355:        (void) chkdq(ip, (long)-btodb(nsize - osize, devBlockSize), cred, FORCE);
                    356: #else
                    357:        (void) chkdq(ip, (long)-btodb(nsize - osize), cred, FORCE);
                    358: #endif /* NeXT */
                    359: #endif /* QUOTA */
                    360:        brelse(bp);
                    361: nospace:
                    362:        /*
                    363:         * no space available
                    364:         */
                    365:        ffs_fserr(fs, cred->cr_uid, "file system full");
                    366:        uprintf("\n%s: write failed, file system is full\n", fs->fs_fsmnt);
                    367:        return (ENOSPC);
                    368: }
                    369: 
                    370: /*
                    371:  * Reallocate a sequence of blocks into a contiguous sequence of blocks.
                    372:  *
                    373:  * The vnode and an array of buffer pointers for a range of sequential
                    374:  * logical blocks to be made contiguous is given. The allocator attempts
                    375:  * to find a range of sequential blocks starting as close as possible to
                    376:  * an fs_rotdelay offset from the end of the allocation for the logical
                    377:  * block immediately preceeding the current range. If successful, the
                    378:  * physical block numbers in the buffer pointers and in the inode are
                    379:  * changed to reflect the new allocation. If unsuccessful, the allocation
                    380:  * is left unchanged. The success in doing the reallocation is returned.
                    381:  * Note that the error return is not reflected back to the user. Rather
                    382:  * the previous block allocation will be used.
                    383:  */
                    384: int doasyncfree = 1;
                    385: int doreallocblks = 1;
                    386: int prtrealloc = 0;
                    387: 
                    388: int
                    389: ffs_reallocblks(ap)
                    390:        struct vop_reallocblks_args /* {
                    391:                struct vnode *a_vp;
                    392:                struct cluster_save *a_buflist;
                    393:        } */ *ap;
                    394: {
                    395:        struct fs *fs;
                    396:        struct inode *ip;
                    397:        struct vnode *vp;
                    398:        struct buf *sbp, *ebp;
                    399:        ufs_daddr_t *bap, *sbap, *ebap;
                    400:        struct cluster_save *buflist;
                    401:        ufs_daddr_t start_lbn, end_lbn, soff, eoff, newblk, blkno;
                    402:        struct indir start_ap[NIADDR + 1], end_ap[NIADDR + 1], *idp;
                    403:        int i, len, start_lvl, end_lvl, pref, ssize;
                    404: #if    REV_ENDIAN_FS
                    405:        int rev_endian=0;
                    406: #endif /* REV_ENDIAN_FS */
                    407: 
                    408:        if (doreallocblks == 0)
                    409:                return (ENOSPC);
                    410:        vp = ap->a_vp;
                    411: #if    REV_ENDIAN_FS
                    412:        rev_endian = vp->v_mount->mnt_flag & MNT_REVEND;
                    413: #endif /* REV_ENDIAN_FS */
                    414: 
                    415:        ip = VTOI(vp);
                    416:        fs = ip->i_fs;
                    417:        if (fs->fs_contigsumsize <= 0)
                    418:                return (ENOSPC);
                    419:        buflist = ap->a_buflist;
                    420:        len = buflist->bs_nchildren;
                    421:        start_lbn = buflist->bs_children[0]->b_lblkno;
                    422:        end_lbn = start_lbn + len - 1;
                    423: #if DIAGNOSTIC
                    424:        for (i = 0; i < len; i++)
                    425:                if (!ffs_checkblk(ip,
                    426:                   dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
                    427:                        panic("ffs_reallocblks: unallocated block 1");
                    428:        for (i = 1; i < len; i++)
                    429:                if (buflist->bs_children[i]->b_lblkno != start_lbn + i)
                    430:                        panic("ffs_reallocblks: non-logical cluster");
                    431:        blkno = buflist->bs_children[0]->b_blkno;
                    432:        ssize = fsbtodb(fs, fs->fs_frag);
                    433:        for (i = 1; i < len - 1; i++)
                    434:                if (buflist->bs_children[i]->b_blkno != blkno + (i * ssize))
                    435:                        panic("ffs_reallocblks: non-physical cluster %d", i);
                    436: #endif
                    437:        /*
                    438:         * If the latest allocation is in a new cylinder group, assume that
                    439:         * the filesystem has decided to move and do not force it back to
                    440:         * the previous cylinder group.
                    441:         */
                    442:        if (dtog(fs, dbtofsb(fs, buflist->bs_children[0]->b_blkno)) !=
                    443:            dtog(fs, dbtofsb(fs, buflist->bs_children[len - 1]->b_blkno)))
                    444:                return (ENOSPC);
                    445:        if (ufs_getlbns(vp, start_lbn, start_ap, &start_lvl) ||
                    446:            ufs_getlbns(vp, end_lbn, end_ap, &end_lvl))
                    447:                return (ENOSPC);
                    448:        /*
                    449:         * Get the starting offset and block map for the first block.
                    450:         */
                    451:        if (start_lvl == 0) {
                    452:                sbap = &ip->i_db[0];
                    453:                soff = start_lbn;
                    454:        } else {
                    455:                idp = &start_ap[start_lvl - 1];
                    456:                if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &sbp)) {
                    457:                        brelse(sbp);
                    458:                        return (ENOSPC);
                    459:                }
                    460:                sbap = (ufs_daddr_t *)sbp->b_data;
                    461:                soff = idp->in_off;
                    462:        }
                    463:        /*
                    464:         * Find the preferred location for the cluster.
                    465:         */
                    466:        pref = ffs_blkpref(ip, start_lbn, soff, sbap);
                    467:        /*
                    468:         * If the block range spans two block maps, get the second map.
                    469:         */
                    470:        if (end_lvl == 0 || (idp = &end_ap[end_lvl - 1])->in_off + 1 >= len) {
                    471:                ssize = len;
                    472:        } else {
                    473: #if DIAGNOSTIC
                    474:                if (start_lvl && start_ap[start_lvl-1].in_lbn == idp->in_lbn)
                    475:                        panic("ffs_reallocblk: start == end");
                    476: #endif
                    477:                ssize = len - (idp->in_off + 1);
                    478:                if (bread(vp, idp->in_lbn, (int)fs->fs_bsize, NOCRED, &ebp))
                    479:                        goto fail;
                    480:                ebap = (ufs_daddr_t *)ebp->b_data;
                    481:        }
                    482:        /*
                    483:         * Search the block map looking for an allocation of the desired size.
                    484:         */
                    485:        if ((newblk = (ufs_daddr_t)ffs_hashalloc(ip, dtog(fs, pref), (long)pref,
                    486:            len, (u_int32_t (*)())ffs_clusteralloc)) == 0)
                    487:                goto fail;
                    488:        /*
                    489:         * We have found a new contiguous block.
                    490:         *
                    491:         * First we have to replace the old block pointers with the new
                    492:         * block pointers in the inode and indirect blocks associated
                    493:         * with the file.
                    494:         */
                    495: #ifdef DEBUG
                    496:        if (prtrealloc)
                    497:                printf("realloc: ino %d, lbns %d-%d\n\told:", ip->i_number,
                    498:                    start_lbn, end_lbn);
                    499: #endif
                    500:        blkno = newblk;
                    501:        for (bap = &sbap[soff], i = 0; i < len; i++, blkno += fs->fs_frag) {
                    502:                if (i == ssize)
                    503:                        bap = ebap;
                    504: #if DIAGNOSTIC
                    505:                if (!ffs_checkblk(ip,
                    506:                   dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
                    507:                        panic("ffs_reallocblks: unallocated block 2");
                    508: #if REV_ENDIAN_FS
                    509:                if (rev_endian && !(bap >= &ip->i_db[0] && bap <= &ip->i_db[NDADDR-1])) {
                    510:                        if (dbtofsb(fs, buflist->bs_children[i]->b_blkno) != NXSwapLong(*bap))
                    511:                                panic("ffs_reallocblks: alloc mismatch");
                    512:                } else {
                    513: #endif /* REV_ENDIAN_FS */
                    514:                        if (dbtofsb(fs, buflist->bs_children[i]->b_blkno) != *bap)
                    515:                                panic("ffs_reallocblks: alloc mismatch");
                    516: #if REV_ENDIAN_FS
                    517:                }
                    518: #endif /* REV_ENDIAN_FS */
                    519: #endif
                    520: #ifdef DEBUG
                    521: #if REV_ENDIAN_FS
                    522:                if (rev_endian && !(bap >= &ip->i_db[0] && bap <= &ip->i_db[NDADDR-1])) {
                    523:                        if (prtrealloc)
                    524:                                printf(" %d,", NXSwapLong(*bap));
                    525:                }else {
                    526: #endif /* REV_ENDIAN_FS */
                    527:                        if (prtrealloc)
                    528:                                printf(" %d,", *bap);
                    529: #if REV_ENDIAN_FS
                    530:                }
                    531: #endif /* REV_ENDIAN_FS */
                    532: #endif
                    533: #if REV_ENDIAN_FS
                    534:                if (rev_endian && !(bap >= &ip->i_db[0] && bap <= &ip->i_db[NDADDR-1])) {
                    535:                        /* An indirect block need to swap as
                    536:                        *  it is going to be written out directly
                    537:                        */
                    538:                        *bap++ = NXSwapLong(blkno);
                    539:                }
                    540:                else {
                    541:                        /* Direct block; going to be written out
                    542:                        * by a VOP_UPDATE; which takes care of swapping 
                    543:                        */
                    544: #endif /* REV_ENDIAN_FS */
                    545:                        *bap++ = blkno;
                    546: #if REV_ENDIAN_FS
                    547:                }
                    548: #endif /* REV_ENDIAN_FS */
                    549: 
                    550:        }
                    551:        /*
                    552:         * Next we must write out the modified inode and indirect blocks.
                    553:         * For strict correctness, the writes should be synchronous since
                    554:         * the old block values may have been written to disk. In practise
                    555:         * they are almost never written, but if we are concerned about 
                    556:         * strict correctness, the `doasyncfree' flag should be set to zero.
                    557:         *
                    558:         * The test on `doasyncfree' should be changed to test a flag
                    559:         * that shows whether the associated buffers and inodes have
                    560:         * been written. The flag should be set when the cluster is
                    561:         * started and cleared whenever the buffer or inode is flushed.
                    562:         * We can then check below to see if it is set, and do the
                    563:         * synchronous write only when it has been cleared.
                    564:         */
                    565:        if (sbap != &ip->i_db[0]) {
                    566:                if (doasyncfree)
                    567:                        bdwrite(sbp);
                    568:                else
                    569:                        bwrite(sbp);
                    570:        } else {
                    571:                ip->i_flag |= IN_CHANGE | IN_UPDATE;
                    572:                if (!doasyncfree)
                    573:                        VOP_UPDATE(vp, &time, &time, MNT_WAIT);
                    574:        }
                    575:        if (ssize < len)
                    576:                if (doasyncfree)
                    577:                        bdwrite(ebp);
                    578:                else
                    579:                        bwrite(ebp);
                    580:        /*
                    581:         * Last, free the old blocks and assign the new blocks to the buffers.
                    582:         */
                    583: #ifdef DEBUG
                    584:        if (prtrealloc)
                    585:                printf("\n\tnew:");
                    586: #endif
                    587:        for (blkno = newblk, i = 0; i < len; i++, blkno += fs->fs_frag) {
                    588:                ffs_blkfree(ip, dbtofsb(fs, buflist->bs_children[i]->b_blkno),
                    589:                    fs->fs_bsize);
                    590:                buflist->bs_children[i]->b_blkno = fsbtodb(fs, blkno);
                    591: #if DIAGNOSTIC
                    592:                if (!ffs_checkblk(ip,
                    593:                   dbtofsb(fs, buflist->bs_children[i]->b_blkno), fs->fs_bsize))
                    594:                        panic("ffs_reallocblks: unallocated block 3");
                    595:                if (prtrealloc)
                    596:                        printf(" %d,", blkno);
                    597: #endif
                    598:        }
                    599: #ifdef DEBUG
                    600:        if (prtrealloc) {
                    601:                prtrealloc--;
                    602:                printf("\n");
                    603:        }
                    604: #endif
                    605:        return (0);
                    606: 
                    607: fail:
                    608:        if (ssize < len)
                    609:                brelse(ebp);
                    610:        if (sbap != &ip->i_db[0])
                    611:                brelse(sbp);
                    612:        return (ENOSPC);
                    613: }
                    614: 
                    615: /*
                    616:  * Allocate an inode in the file system.
                    617:  * 
                    618:  * If allocating a directory, use ffs_dirpref to select the inode.
                    619:  * If allocating in a directory, the following hierarchy is followed:
                    620:  *   1) allocate the preferred inode.
                    621:  *   2) allocate an inode in the same cylinder group.
                    622:  *   3) quadradically rehash into other cylinder groups, until an
                    623:  *      available inode is located.
                    624:  * If no inode preference is given the following heirarchy is used
                    625:  * to allocate an inode:
                    626:  *   1) allocate an inode in cylinder group 0.
                    627:  *   2) quadradically rehash into other cylinder groups, until an
                    628:  *      available inode is located.
                    629:  */
                    630: int
                    631: ffs_valloc(ap)
                    632:        struct vop_valloc_args /* {
                    633:                struct vnode *a_pvp;
                    634:                int a_mode;
                    635:                struct ucred *a_cred;
                    636:                struct vnode **a_vpp;
                    637:        } */ *ap;
                    638: {
                    639:        register struct vnode *pvp = ap->a_pvp;
                    640:        register struct inode *pip;
                    641:        register struct fs *fs;
                    642:        register struct inode *ip;
                    643:        mode_t mode = ap->a_mode;
                    644:        ino_t ino, ipref;
                    645:        int cg, error;
                    646:        
                    647:        *ap->a_vpp = NULL;
                    648:        pip = VTOI(pvp);
                    649:        fs = pip->i_fs;
                    650:        if (fs->fs_cstotal.cs_nifree == 0)
                    651:                goto noinodes;
                    652: 
                    653:        if ((mode & IFMT) == IFDIR)
                    654:                ipref = ffs_dirpref(fs);
                    655:        else
                    656:                ipref = pip->i_number;
                    657:        if (ipref >= fs->fs_ncg * fs->fs_ipg)
                    658:                ipref = 0;
                    659:        cg = ino_to_cg(fs, ipref);
                    660:        ino = (ino_t)ffs_hashalloc(pip, cg, (long)ipref, mode, ffs_nodealloccg);
                    661:        if (ino == 0)
                    662:                goto noinodes;
                    663:        error = VFS_VGET(pvp->v_mount, ino, ap->a_vpp);
                    664:        if (error) {
                    665:                VOP_VFREE(pvp, ino, mode);
                    666:                return (error);
                    667:        }
                    668:        ip = VTOI(*ap->a_vpp);
                    669:        if (ip->i_mode) {
                    670:                printf("mode = 0%o, inum = %d, fs = %s\n",
                    671:                    ip->i_mode, ip->i_number, fs->fs_fsmnt);
                    672:                panic("ffs_valloc: dup alloc");
                    673:        }
                    674:        if (ip->i_blocks) {                             /* XXX */
                    675:                printf("free inode %s/%d had %d blocks\n",
                    676:                    fs->fs_fsmnt, ino, ip->i_blocks);
                    677:                ip->i_blocks = 0;
                    678:        }
                    679:        ip->i_flags = 0;
                    680:        /*
                    681:         * Set up a new generation number for this inode.
                    682:         */
                    683:        if (++nextgennumber < (u_long)time.tv_sec)
                    684:                nextgennumber = time.tv_sec;
                    685:        ip->i_gen = nextgennumber;
                    686:        return (0);
                    687: noinodes:
                    688:        ffs_fserr(fs, ap->a_cred->cr_uid, "out of inodes");
                    689:        uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt);
                    690:        return (ENOSPC);
                    691: }
                    692: 
                    693: /*
                    694:  * Find a cylinder to place a directory.
                    695:  *
                    696:  * The policy implemented by this algorithm is to select from
                    697:  * among those cylinder groups with above the average number of
                    698:  * free inodes, the one with the smallest number of directories.
                    699:  */
                    700: static ino_t
                    701: ffs_dirpref(fs)
                    702:        register struct fs *fs;
                    703: {
                    704:        int cg, minndir, mincg, avgifree;
                    705: 
                    706:        avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg;
                    707:        minndir = fs->fs_ipg;
                    708:        mincg = 0;
                    709:        for (cg = 0; cg < fs->fs_ncg; cg++)
                    710:                if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
                    711:                    fs->fs_cs(fs, cg).cs_nifree >= avgifree) {
                    712:                        mincg = cg;
                    713:                        minndir = fs->fs_cs(fs, cg).cs_ndir;
                    714:                }
                    715:        return ((ino_t)(fs->fs_ipg * mincg));
                    716: }
                    717: 
                    718: /*
                    719:  * Select the desired position for the next block in a file.  The file is
                    720:  * logically divided into sections. The first section is composed of the
                    721:  * direct blocks. Each additional section contains fs_maxbpg blocks.
                    722:  * 
                    723:  * If no blocks have been allocated in the first section, the policy is to
                    724:  * request a block in the same cylinder group as the inode that describes
                    725:  * the file. If no blocks have been allocated in any other section, the
                    726:  * policy is to place the section in a cylinder group with a greater than
                    727:  * average number of free blocks.  An appropriate cylinder group is found
                    728:  * by using a rotor that sweeps the cylinder groups. When a new group of
                    729:  * blocks is needed, the sweep begins in the cylinder group following the
                    730:  * cylinder group from which the previous allocation was made. The sweep
                    731:  * continues until a cylinder group with greater than the average number
                    732:  * of free blocks is found. If the allocation is for the first block in an
                    733:  * indirect block, the information on the previous allocation is unavailable;
                    734:  * here a best guess is made based upon the logical block number being
                    735:  * allocated.
                    736:  * 
                    737:  * If a section is already partially allocated, the policy is to
                    738:  * contiguously allocate fs_maxcontig blocks.  The end of one of these
                    739:  * contiguous blocks and the beginning of the next is physically separated
                    740:  * so that the disk head will be in transit between them for at least
                    741:  * fs_rotdelay milliseconds.  This is to allow time for the processor to
                    742:  * schedule another I/O transfer.
                    743:  */
                    744: ufs_daddr_t
                    745: ffs_blkpref(ip, lbn, indx, bap)
                    746:        struct inode *ip;
                    747:        ufs_daddr_t lbn;
                    748:        int indx;
                    749:        ufs_daddr_t *bap;
                    750: {
                    751:        register struct fs *fs;
                    752:        register int cg;
                    753:        int avgbfree, startcg;
                    754:        ufs_daddr_t nextblk;
                    755: #if    REV_ENDIAN_FS
                    756:        daddr_t prev=0;
                    757:        struct vnode *vp=ITOV(ip);
                    758:        struct mount *mp=vp->v_mount;
                    759:        int rev_endian=(mp->mnt_flag & MNT_REVEND);
                    760: #endif /* REV_ENDIAN_FS */
                    761: 
                    762:        fs = ip->i_fs;
                    763: #if    REV_ENDIAN_FS
                    764:        if (indx && bap) {
                    765:        if (rev_endian) {
                    766:                if (bap != &ip->i_db[0])
                    767:                        prev = NXSwapLong(bap[indx - 1]);
                    768:                else
                    769:                        prev = bap[indx - 1];
                    770:        } else prev = bap[indx - 1];
                    771:        }
                    772:        if (indx % fs->fs_maxbpg == 0 || prev == 0)
                    773: #else  /* REV_ENDIAN_FS */
                    774:        if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) 
                    775: #endif /* REV_ENDIAN_FS */
                    776:        {
                    777:                if (lbn < NDADDR) {
                    778:                        cg = ino_to_cg(fs, ip->i_number);
                    779:                        return (fs->fs_fpg * cg + fs->fs_frag);
                    780:                }
                    781:                /*
                    782:                 * Find a cylinder with greater than average number of
                    783:                 * unused data blocks.
                    784:                 */
                    785: #if    REV_ENDIAN_FS
                    786:                if (indx == 0 || prev == 0)
                    787: #else  /* REV_ENDIAN_FS */
                    788:                if (indx == 0 || bap[indx - 1] == 0)
                    789: #endif /* REV_ENDIAN_FS */
                    790:                        startcg =
                    791:                            ino_to_cg(fs, ip->i_number) + lbn / fs->fs_maxbpg;
                    792:                else
                    793: #if    REV_ENDIAN_FS
                    794:                        startcg = dtog(fs, prev) + 1;
                    795: #else  /* REV_ENDIAN_FS */
                    796:                        startcg = dtog(fs, bap[indx - 1]) + 1;
                    797: #endif /* REV_ENDIAN_FS */
                    798:                startcg %= fs->fs_ncg;
                    799:                avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
                    800:                for (cg = startcg; cg < fs->fs_ncg; cg++)
                    801:                        if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
                    802:                                fs->fs_cgrotor = cg;
                    803:                                return (fs->fs_fpg * cg + fs->fs_frag);
                    804:                        }
                    805:                for (cg = 0; cg <= startcg; cg++)
                    806:                        if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
                    807:                                fs->fs_cgrotor = cg;
                    808:                                return (fs->fs_fpg * cg + fs->fs_frag);
                    809:                        }
                    810:                return (NULL);
                    811:        }
                    812:        /*
                    813:         * One or more previous blocks have been laid out. If less
                    814:         * than fs_maxcontig previous blocks are contiguous, the
                    815:         * next block is requested contiguously, otherwise it is
                    816:         * requested rotationally delayed by fs_rotdelay milliseconds.
                    817:         */
                    818: #if    REV_ENDIAN_FS
                    819:        if (rev_endian) {
                    820:                nextblk = prev + fs->fs_frag;
                    821:                if (indx < fs->fs_maxcontig) {
                    822:                        return (nextblk);
                    823:                }
                    824:                if (bap != &ip->i_db[0])
                    825:                        prev = NXSwapLong(bap[indx - fs->fs_maxcontig]);
                    826:                else
                    827:                        prev = bap[indx - fs->fs_maxcontig];
                    828:                if (prev + blkstofrags(fs, fs->fs_maxcontig) != nextblk)
                    829:                        return (nextblk);
                    830:        } else {
                    831: #endif /* REV_ENDIAN_FS */
                    832:        nextblk = bap[indx - 1] + fs->fs_frag;
                    833:        if (indx < fs->fs_maxcontig || bap[indx - fs->fs_maxcontig] +
                    834:            blkstofrags(fs, fs->fs_maxcontig) != nextblk)
                    835:                return (nextblk);
                    836: #if REV_ENDIAN_FS
                    837:        }
                    838: #endif /* REV_ENDIAN_FS */
                    839:        if (fs->fs_rotdelay != 0)
                    840:                /*
                    841:                 * Here we convert ms of delay to frags as:
                    842:                 * (frags) = (ms) * (rev/sec) * (sect/rev) /
                    843:                 *      ((sect/frag) * (ms/sec))
                    844:                 * then round up to the next block.
                    845:                 */
                    846:                nextblk += roundup(fs->fs_rotdelay * fs->fs_rps * fs->fs_nsect /
                    847:                    (NSPF(fs) * 1000), fs->fs_frag);
                    848:        return (nextblk);
                    849: }
                    850: 
                    851: /*
                    852:  * Implement the cylinder overflow algorithm.
                    853:  *
                    854:  * The policy implemented by this algorithm is:
                    855:  *   1) allocate the block in its requested cylinder group.
                    856:  *   2) quadradically rehash on the cylinder group number.
                    857:  *   3) brute force search for a free block.
                    858:  */
                    859: /*VARARGS5*/
                    860: static u_long
                    861: ffs_hashalloc(ip, cg, pref, size, allocator)
                    862:        struct inode *ip;
                    863:        int cg;
                    864:        long pref;
                    865:        int size;       /* size for data blocks, mode for inodes */
                    866:        u_int32_t (*allocator)();
                    867: {
                    868:        register struct fs *fs;
                    869:        long result;
                    870:        int i, icg = cg;
                    871: 
                    872:        fs = ip->i_fs;
                    873:        /*
                    874:         * 1: preferred cylinder group
                    875:         */
                    876:        result = (*allocator)(ip, cg, pref, size);
                    877:        if (result)
                    878:                return (result);
                    879:        /*
                    880:         * 2: quadratic rehash
                    881:         */
                    882:        for (i = 1; i < fs->fs_ncg; i *= 2) {
                    883:                cg += i;
                    884:                if (cg >= fs->fs_ncg)
                    885:                        cg -= fs->fs_ncg;
                    886:                result = (*allocator)(ip, cg, 0, size);
                    887:                if (result)
                    888:                        return (result);
                    889:        }
                    890:        /*
                    891:         * 3: brute force search
                    892:         * Note that we start at i == 2, since 0 was checked initially,
                    893:         * and 1 is always checked in the quadratic rehash.
                    894:         */
                    895:        cg = (icg + 2) % fs->fs_ncg;
                    896:        for (i = 2; i < fs->fs_ncg; i++) {
                    897:                result = (*allocator)(ip, cg, 0, size);
                    898:                if (result)
                    899:                        return (result);
                    900:                cg++;
                    901:                if (cg == fs->fs_ncg)
                    902:                        cg = 0;
                    903:        }
                    904:        return (NULL);
                    905: }
                    906: 
                    907: /*
                    908:  * Determine whether a fragment can be extended.
                    909:  *
                    910:  * Check to see if the necessary fragments are available, and 
                    911:  * if they are, allocate them.
                    912:  */
                    913: static ufs_daddr_t
                    914: ffs_fragextend(ip, cg, bprev, osize, nsize)
                    915:        struct inode *ip;
                    916:        int cg;
                    917:        long bprev;
                    918:        int osize, nsize;
                    919: {
                    920:        register struct fs *fs;
                    921:        register struct cg *cgp;
                    922:        struct buf *bp;
                    923:        long bno;
                    924:        int frags, bbase;
                    925:        int i, error;
                    926: #if REV_ENDIAN_FS
                    927:        struct vnode *vp=ITOV(ip);
                    928:        struct mount *mp=vp->v_mount;
                    929:        int rev_endian=(mp->mnt_flag & MNT_REVEND);
                    930: #endif /* REV_ENDIAN_FS */
                    931: 
                    932:        fs = ip->i_fs;
                    933:        if (fs->fs_cs(fs, cg).cs_nffree < numfrags(fs, nsize - osize))
                    934:                return (NULL);
                    935:        frags = numfrags(fs, nsize);
                    936:        bbase = fragnum(fs, bprev);
                    937:        if (bbase > fragnum(fs, (bprev + frags - 1))) {
                    938:                /* cannot extend across a block boundary */
                    939:                return (NULL);
                    940:        }
                    941:        error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
                    942:                (int)fs->fs_cgsize, NOCRED, &bp);
                    943:        if (error) {
                    944:                brelse(bp);
                    945:                return (NULL);
                    946:        }
                    947:        cgp = (struct cg *)bp->b_data;
                    948: #if REV_ENDIAN_FS
                    949:        if (rev_endian) {
                    950:                byte_swap_cgin(cgp, fs);
                    951:        }
                    952: #endif /* REV_ENDIAN_FS */
                    953: 
                    954:        if (!cg_chkmagic(cgp)) {
                    955: #if REV_ENDIAN_FS
                    956:                if (rev_endian)
                    957:                        byte_swap_cgout(cgp,fs);
                    958: #endif /* REV_ENDIAN_FS */
                    959:                brelse(bp);
                    960:                return (NULL);
                    961:        }
                    962:        cgp->cg_time = time.tv_sec;
                    963:        bno = dtogd(fs, bprev);
                    964:        for (i = numfrags(fs, osize); i < frags; i++)
                    965:                if (isclr(cg_blksfree(cgp), bno + i)) {
                    966: #if REV_ENDIAN_FS
                    967:                        if (rev_endian)
                    968:                                byte_swap_cgout(cgp,fs);
                    969: #endif /* REV_ENDIAN_FS */
                    970:                        brelse(bp);
                    971:                        return (NULL);
                    972:                }
                    973:        /*
                    974:         * the current fragment can be extended
                    975:         * deduct the count on fragment being extended into
                    976:         * increase the count on the remaining fragment (if any)
                    977:         * allocate the extended piece
                    978:         */
                    979:        for (i = frags; i < fs->fs_frag - bbase; i++)
                    980:                if (isclr(cg_blksfree(cgp), bno + i))
                    981:                        break;
                    982:        cgp->cg_frsum[i - numfrags(fs, osize)]--;
                    983:        if (i != frags)
                    984:                cgp->cg_frsum[i - frags]++;
                    985:        for (i = numfrags(fs, osize); i < frags; i++) {
                    986:                clrbit(cg_blksfree(cgp), bno + i);
                    987:                cgp->cg_cs.cs_nffree--;
                    988:                fs->fs_cstotal.cs_nffree--;
                    989:                fs->fs_cs(fs, cg).cs_nffree--;
                    990:        }
                    991:        fs->fs_fmod = 1;
                    992: #if REV_ENDIAN_FS
                    993:        if (rev_endian)
                    994:                byte_swap_cgout(cgp,fs);
                    995: #endif /* REV_ENDIAN_FS */
                    996:        bdwrite(bp);
                    997:        return (bprev);
                    998: }
                    999: 
                   1000: /*
                   1001:  * Determine whether a block can be allocated.
                   1002:  *
                   1003:  * Check to see if a block of the appropriate size is available,
                   1004:  * and if it is, allocate it.
                   1005:  */
                   1006: static ufs_daddr_t
                   1007: ffs_alloccg(ip, cg, bpref, size)
                   1008:        struct inode *ip;
                   1009:        int cg;
                   1010:        ufs_daddr_t bpref;
                   1011:        int size;
                   1012: {
                   1013:        register struct fs *fs;
                   1014:        register struct cg *cgp;
                   1015:        struct buf *bp;
                   1016:        register int i;
                   1017:        int error, bno, frags, allocsiz;
                   1018: #if REV_ENDIAN_FS
                   1019:        struct vnode *vp=ITOV(ip);
                   1020:        struct mount *mp=vp->v_mount;
                   1021:        int rev_endian=(mp->mnt_flag & MNT_REVEND);
                   1022: #endif /* REV_ENDIAN_FS */
                   1023: 
                   1024:        fs = ip->i_fs;
                   1025:        if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize)
                   1026:                return (NULL);
                   1027:        error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
                   1028:                (int)fs->fs_cgsize, NOCRED, &bp);
                   1029:        if (error) {
                   1030:                brelse(bp);
                   1031:                return (NULL);
                   1032:        }
                   1033:        cgp = (struct cg *)bp->b_data;
                   1034: #if REV_ENDIAN_FS
                   1035:        if (rev_endian)
                   1036:                byte_swap_cgin(cgp,fs);
                   1037: #endif /* REV_ENDIAN_FS */
                   1038:        if (!cg_chkmagic(cgp) ||
                   1039:            (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize)) {
                   1040: #if REV_ENDIAN_FS
                   1041:                if (rev_endian)
                   1042:                        byte_swap_cgout(cgp,fs);
                   1043: #endif /* REV_ENDIAN_FS */
                   1044:                brelse(bp);
                   1045:                return (NULL);
                   1046:        }
                   1047:        cgp->cg_time = time.tv_sec;
                   1048:        if (size == fs->fs_bsize) {
                   1049:                bno = ffs_alloccgblk(fs, cgp, bpref);
                   1050: #if REV_ENDIAN_FS
                   1051:                if (rev_endian)
                   1052:                        byte_swap_cgout(cgp,fs);
                   1053: #endif /* REV_ENDIAN_FS */
                   1054:                bdwrite(bp);
                   1055:                return (bno);
                   1056:        }
                   1057:        /*
                   1058:         * check to see if any fragments are already available
                   1059:         * allocsiz is the size which will be allocated, hacking
                   1060:         * it down to a smaller size if necessary
                   1061:         */
                   1062:        frags = numfrags(fs, size);
                   1063:        for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++)
                   1064:                if (cgp->cg_frsum[allocsiz] != 0)
                   1065:                        break;
                   1066:        if (allocsiz == fs->fs_frag) {
                   1067:                /*
                   1068:                 * no fragments were available, so a block will be 
                   1069:                 * allocated, and hacked up
                   1070:                 */
                   1071:                if (cgp->cg_cs.cs_nbfree == 0) {
                   1072: #if    REV_ENDIAN_FS
                   1073:                        if (rev_endian)
                   1074:                                byte_swap_cgout(cgp,fs);
                   1075: #endif /* REV_ENDIAN_FS */
                   1076:                        brelse(bp);
                   1077:                        return (NULL);
                   1078:                }
                   1079:                bno = ffs_alloccgblk(fs, cgp, bpref);
                   1080:                bpref = dtogd(fs, bno);
                   1081:                for (i = frags; i < fs->fs_frag; i++)
                   1082:                        setbit(cg_blksfree(cgp), bpref + i);
                   1083:                i = fs->fs_frag - frags;
                   1084:                cgp->cg_cs.cs_nffree += i;
                   1085:                fs->fs_cstotal.cs_nffree += i;
                   1086:                fs->fs_cs(fs, cg).cs_nffree += i;
                   1087:                fs->fs_fmod = 1;
                   1088:                cgp->cg_frsum[i]++;
                   1089: #if    REV_ENDIAN_FS
                   1090:                if (rev_endian)
                   1091:                        byte_swap_cgout(cgp,fs);
                   1092: #endif /* REV_ENDIAN_FS */
                   1093:                bdwrite(bp);
                   1094:                return (bno);
                   1095:        }
                   1096:        bno = ffs_mapsearch(fs, cgp, bpref, allocsiz);
                   1097:        if (bno < 0) {
                   1098: #if    REV_ENDIAN_FS
                   1099:                if (rev_endian)
                   1100:                        byte_swap_cgout(cgp,fs);
                   1101: #endif /* REV_ENDIAN_FS */
                   1102:                brelse(bp);
                   1103:                return (NULL);
                   1104:        }
                   1105:        for (i = 0; i < frags; i++)
                   1106:                clrbit(cg_blksfree(cgp), bno + i);
                   1107:        cgp->cg_cs.cs_nffree -= frags;
                   1108:        fs->fs_cstotal.cs_nffree -= frags;
                   1109:        fs->fs_cs(fs, cg).cs_nffree -= frags;
                   1110:        fs->fs_fmod = 1;
                   1111:        cgp->cg_frsum[allocsiz]--;
                   1112:        if (frags != allocsiz)
                   1113:                cgp->cg_frsum[allocsiz - frags]++;
                   1114: #if    REV_ENDIAN_FS
                   1115:        if (rev_endian)
                   1116:                byte_swap_cgout(cgp,fs);
                   1117: #endif /* REV_ENDIAN_FS */
                   1118:        bdwrite(bp);
                   1119:        return (cg * fs->fs_fpg + bno);
                   1120: }
                   1121: 
                   1122: /*
                   1123:  * Allocate a block in a cylinder group.
                   1124:  *
                   1125:  * This algorithm implements the following policy:
                   1126:  *   1) allocate the requested block.
                   1127:  *   2) allocate a rotationally optimal block in the same cylinder.
                   1128:  *   3) allocate the next available block on the block rotor for the
                   1129:  *      specified cylinder group.
                   1130:  * Note that this routine only allocates fs_bsize blocks; these
                   1131:  * blocks may be fragmented by the routine that allocates them.
                   1132:  */
                   1133: static ufs_daddr_t
                   1134: ffs_alloccgblk(fs, cgp, bpref)
                   1135:        register struct fs *fs;
                   1136:        register struct cg *cgp;
                   1137:        ufs_daddr_t bpref;
                   1138: {
                   1139:        ufs_daddr_t bno, blkno;
                   1140:        int cylno, pos, delta;
                   1141:        short *cylbp;
                   1142:        register int i;
                   1143: 
                   1144:        if (bpref == 0 || dtog(fs, bpref) != cgp->cg_cgx) {
                   1145:                bpref = cgp->cg_rotor;
                   1146:                goto norot;
                   1147:        }
                   1148:        bpref = blknum(fs, bpref);
                   1149:        bpref = dtogd(fs, bpref);
                   1150:        /*
                   1151:         * if the requested block is available, use it
                   1152:         */
                   1153:        if (ffs_isblock(fs, cg_blksfree(cgp), fragstoblks(fs, bpref))) {
                   1154:                bno = bpref;
                   1155:                goto gotit;
                   1156:        }
                   1157:        if (fs->fs_nrpos <= 1 || fs->fs_cpc == 0) {
                   1158:                /*
                   1159:                 * Block layout information is not available.
                   1160:                 * Leaving bpref unchanged means we take the
                   1161:                 * next available free block following the one 
                   1162:                 * we just allocated. Hopefully this will at
                   1163:                 * least hit a track cache on drives of unknown
                   1164:                 * geometry (e.g. SCSI).
                   1165:                 */
                   1166:                goto norot;
                   1167:        }
                   1168:        /*
                   1169:         * check for a block available on the same cylinder
                   1170:         */
                   1171:        cylno = cbtocylno(fs, bpref);
                   1172:        if (cg_blktot(cgp)[cylno] == 0)
                   1173:                goto norot;
                   1174:        /*
                   1175:         * check the summary information to see if a block is 
                   1176:         * available in the requested cylinder starting at the
                   1177:         * requested rotational position and proceeding around.
                   1178:         */
                   1179:        cylbp = cg_blks(fs, cgp, cylno);
                   1180:        pos = cbtorpos(fs, bpref);
                   1181:        for (i = pos; i < fs->fs_nrpos; i++)
                   1182:                if (cylbp[i] > 0)
                   1183:                        break;
                   1184:        if (i == fs->fs_nrpos)
                   1185:                for (i = 0; i < pos; i++)
                   1186:                        if (cylbp[i] > 0)
                   1187:                                break;
                   1188:        if (cylbp[i] > 0) {
                   1189:                /*
                   1190:                 * found a rotational position, now find the actual
                   1191:                 * block. A panic if none is actually there.
                   1192:                 */
                   1193:                pos = cylno % fs->fs_cpc;
                   1194:                bno = (cylno - pos) * fs->fs_spc / NSPB(fs);
                   1195:                if (fs_postbl(fs, pos)[i] == -1) {
                   1196:                        printf("pos = %d, i = %d, fs = %s\n",
                   1197:                            pos, i, fs->fs_fsmnt);
                   1198:                        panic("ffs_alloccgblk: cyl groups corrupted");
                   1199:                }
                   1200:                for (i = fs_postbl(fs, pos)[i];; ) {
                   1201:                        if (ffs_isblock(fs, cg_blksfree(cgp), bno + i)) {
                   1202:                                bno = blkstofrags(fs, (bno + i));
                   1203:                                goto gotit;
                   1204:                        }
                   1205:                        delta = fs_rotbl(fs)[i];
                   1206:                        if (delta <= 0 ||
                   1207:                            delta + i > fragstoblks(fs, fs->fs_fpg))
                   1208:                                break;
                   1209:                        i += delta;
                   1210:                }
                   1211:                printf("pos = %d, i = %d, fs = %s\n", pos, i, fs->fs_fsmnt);
                   1212:                panic("ffs_alloccgblk: can't find blk in cyl");
                   1213:        }
                   1214: norot:
                   1215:        /*
                   1216:         * no blocks in the requested cylinder, so take next
                   1217:         * available one in this cylinder group.
                   1218:         */
                   1219:        bno = ffs_mapsearch(fs, cgp, bpref, (int)fs->fs_frag);
                   1220:        if (bno < 0)
                   1221:                return (NULL);
                   1222:        cgp->cg_rotor = bno;
                   1223: gotit:
                   1224:        blkno = fragstoblks(fs, bno);
                   1225:        ffs_clrblock(fs, cg_blksfree(cgp), (long)blkno);
                   1226:        ffs_clusteracct(fs, cgp, blkno, -1);
                   1227:        cgp->cg_cs.cs_nbfree--;
                   1228:        fs->fs_cstotal.cs_nbfree--;
                   1229:        fs->fs_cs(fs, cgp->cg_cgx).cs_nbfree--;
                   1230:        cylno = cbtocylno(fs, bno);
                   1231:        cg_blks(fs, cgp, cylno)[cbtorpos(fs, bno)]--;
                   1232:        cg_blktot(cgp)[cylno]--;
                   1233:        fs->fs_fmod = 1;
                   1234:        return (cgp->cg_cgx * fs->fs_fpg + bno);
                   1235: }
                   1236: 
                   1237: /*
                   1238:  * Determine whether a cluster can be allocated.
                   1239:  *
                   1240:  * We do not currently check for optimal rotational layout if there
                   1241:  * are multiple choices in the same cylinder group. Instead we just
                   1242:  * take the first one that we find following bpref.
                   1243:  */
                   1244: static ufs_daddr_t
                   1245: ffs_clusteralloc(ip, cg, bpref, len)
                   1246:        struct inode *ip;
                   1247:        int cg;
                   1248:        ufs_daddr_t bpref;
                   1249:        int len;
                   1250: {
                   1251:        register struct fs *fs;
                   1252:        register struct cg *cgp;
                   1253:        struct buf *bp;
                   1254:        int i, got, run, bno, bit, map;
                   1255:        u_char *mapp;
                   1256:        int32_t *lp;
                   1257: #if REV_ENDIAN_FS
                   1258:        struct vnode *vp=ITOV(ip);
                   1259:        struct mount *mp=vp->v_mount;
                   1260:        int rev_endian=(mp->mnt_flag & MNT_REVEND);
                   1261: #endif /* REV_ENDIAN_FS */
                   1262: 
                   1263:        fs = ip->i_fs;
                   1264:        if (fs->fs_maxcluster[cg] < len)
                   1265:                return (NULL);
                   1266:        if (bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize,
                   1267:            NOCRED, &bp))
                   1268:                goto fail;
                   1269:        cgp = (struct cg *)bp->b_data;
                   1270: #if    REV_ENDIAN_FS
                   1271:        if (rev_endian)
                   1272:                byte_swap_cgin(cgp,fs);
                   1273: #endif /* REV_ENDIAN_FS */
                   1274:        if (!cg_chkmagic(cgp)) {
                   1275: #if    REV_ENDIAN_FS
                   1276:                if (rev_endian)
                   1277:                        byte_swap_cgout(cgp,fs);
                   1278: #endif /* REV_ENDIAN_FS */
                   1279:                goto fail;
                   1280:        }
                   1281:        /*
                   1282:         * Check to see if a cluster of the needed size (or bigger) is
                   1283:         * available in this cylinder group.
                   1284:         */
                   1285:        lp = &cg_clustersum(cgp)[len];
                   1286:        for (i = len; i <= fs->fs_contigsumsize; i++)
                   1287:                if (*lp++ > 0)
                   1288:                        break;
                   1289:        if (i > fs->fs_contigsumsize) {
                   1290:                /*
                   1291:                 * This is the first time looking for a cluster in this
                   1292:                 * cylinder group. Update the cluster summary information
                   1293:                 * to reflect the true maximum sized cluster so that
                   1294:                 * future cluster allocation requests can avoid reading
                   1295:                 * the cylinder group map only to find no clusters.
                   1296:                 */
                   1297:                lp = &cg_clustersum(cgp)[len - 1];
                   1298:                for (i = len - 1; i > 0; i--)
                   1299:                        if (*lp-- > 0)
                   1300:                                break;
                   1301:                fs->fs_maxcluster[cg] = i;
                   1302: #if    REV_ENDIAN_FS
                   1303:                if (rev_endian)
                   1304:                        byte_swap_cgout(cgp,fs);
                   1305: #endif /* REV_ENDIAN_FS */
                   1306:                goto fail;
                   1307:        }
                   1308:        /*
                   1309:         * Search the cluster map to find a big enough cluster.
                   1310:         * We take the first one that we find, even if it is larger
                   1311:         * than we need as we prefer to get one close to the previous
                   1312:         * block allocation. We do not search before the current
                   1313:         * preference point as we do not want to allocate a block
                   1314:         * that is allocated before the previous one (as we will
                   1315:         * then have to wait for another pass of the elevator
                   1316:         * algorithm before it will be read). We prefer to fail and
                   1317:         * be recalled to try an allocation in the next cylinder group.
                   1318:         */
                   1319:        if (dtog(fs, bpref) != cg)
                   1320:                bpref = 0;
                   1321:        else
                   1322:                bpref = fragstoblks(fs, dtogd(fs, blknum(fs, bpref)));
                   1323:        mapp = &cg_clustersfree(cgp)[bpref / NBBY];
                   1324:        map = *mapp++;
                   1325:        bit = 1 << (bpref % NBBY);
                   1326:        for (run = 0, got = bpref; got < cgp->cg_nclusterblks; got++) {
                   1327:                if ((map & bit) == 0) {
                   1328:                        run = 0;
                   1329:                } else {
                   1330:                        run++;
                   1331:                        if (run == len)
                   1332:                                break;
                   1333:                }
                   1334:                if ((got & (NBBY - 1)) != (NBBY - 1)) {
                   1335:                        bit <<= 1;
                   1336:                } else {
                   1337:                        map = *mapp++;
                   1338:                        bit = 1;
                   1339:                }
                   1340:        }
                   1341:        if (got == cgp->cg_nclusterblks) {
                   1342: #if    REV_ENDIAN_FS
                   1343:                if (rev_endian)
                   1344:                        byte_swap_cgout(cgp,fs);
                   1345: #endif /* REV_ENDIAN_FS */
                   1346:                goto fail;
                   1347:        }
                   1348:        /*
                   1349:         * Allocate the cluster that we have found.
                   1350:         */
                   1351:        for (i = 1; i <= len; i++)
                   1352:                if (!ffs_isblock(fs, cg_blksfree(cgp), got - run + i))
                   1353:                        panic("ffs_clusteralloc: map mismatch");
                   1354:        bno = cg * fs->fs_fpg + blkstofrags(fs, got - run + 1);
                   1355:        if (dtog(fs, bno) != cg)
                   1356:                panic("ffs_clusteralloc: allocated out of group");
                   1357:        len = blkstofrags(fs, len);
                   1358:        for (i = 0; i < len; i += fs->fs_frag)
                   1359:                if ((got = ffs_alloccgblk(fs, cgp, bno + i)) != bno + i)
                   1360:                        panic("ffs_clusteralloc: lost block");
                   1361: #if    REV_ENDIAN_FS
                   1362:        if (rev_endian)
                   1363:                byte_swap_cgout(cgp,fs);
                   1364: #endif /* REV_ENDIAN_FS */
                   1365:        bdwrite(bp);
                   1366:        return (bno);
                   1367: 
                   1368: fail:
                   1369:        brelse(bp);
                   1370:        return (0);
                   1371: }
                   1372: 
                   1373: /*
                   1374:  * Determine whether an inode can be allocated.
                   1375:  *
                   1376:  * Check to see if an inode is available, and if it is,
                   1377:  * allocate it using the following policy:
                   1378:  *   1) allocate the requested inode.
                   1379:  *   2) allocate the next available inode after the requested
                   1380:  *      inode in the specified cylinder group.
                   1381:  */
                   1382: static ino_t
                   1383: ffs_nodealloccg(ip, cg, ipref, mode)
                   1384:        struct inode *ip;
                   1385:        int cg;
                   1386:        ufs_daddr_t ipref;
                   1387:        int mode;
                   1388: {
                   1389:        register struct fs *fs;
                   1390:        register struct cg *cgp;
                   1391:        struct buf *bp;
                   1392:        int error, start, len, loc, map, i;
                   1393: #if REV_ENDIAN_FS
                   1394:        struct vnode *vp=ITOV(ip);
                   1395:        struct mount *mp=vp->v_mount;
                   1396:        int rev_endian=(mp->mnt_flag & MNT_REVEND);
                   1397: #endif /* REV_ENDIAN_FS */
                   1398: 
                   1399:        fs = ip->i_fs;
                   1400:        if (fs->fs_cs(fs, cg).cs_nifree == 0)
                   1401:                return (NULL);
                   1402:        error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
                   1403:                (int)fs->fs_cgsize, NOCRED, &bp);
                   1404:        if (error) {
                   1405:                brelse(bp);
                   1406:                return (NULL);
                   1407:        }
                   1408:        cgp = (struct cg *)bp->b_data;
                   1409: #if REV_ENDIAN_FS
                   1410:        if (rev_endian)
                   1411:                byte_swap_cgin(cgp,fs);
                   1412: #endif /* REV_ENDIAN_FS */
                   1413:        if (!cg_chkmagic(cgp) || cgp->cg_cs.cs_nifree == 0) {
                   1414: #if REV_ENDIAN_FS
                   1415:                if (rev_endian)
                   1416:                        byte_swap_cgout(cgp,fs);
                   1417: #endif /* REV_ENDIAN_FS */
                   1418:                brelse(bp);
                   1419:                return (NULL);
                   1420:        }
                   1421: 
                   1422:        cgp->cg_time = time.tv_sec;
                   1423:        if (ipref) {
                   1424:                ipref %= fs->fs_ipg;
                   1425:                if (isclr(cg_inosused(cgp), ipref))
                   1426:                        goto gotit;
                   1427:        }
                   1428:        start = cgp->cg_irotor / NBBY;
                   1429:        len = howmany(fs->fs_ipg - cgp->cg_irotor, NBBY);
                   1430:        loc = skpc(0xff, len, &cg_inosused(cgp)[start]);
                   1431:        if (loc == 0) {
                   1432:                len = start + 1;
                   1433:                start = 0;
                   1434:                loc = skpc(0xff, len, &cg_inosused(cgp)[0]);
                   1435:                if (loc == 0) {
                   1436:                        printf("cg = %d, irotor = %d, fs = %s\n",
                   1437:                            cg, cgp->cg_irotor, fs->fs_fsmnt);
                   1438:                        panic("ffs_nodealloccg: map corrupted");
                   1439:                        /* NOTREACHED */
                   1440:                }
                   1441:        }
                   1442:        i = start + len - loc;
                   1443:        map = cg_inosused(cgp)[i];
                   1444:        ipref = i * NBBY;
                   1445:        for (i = 1; i < (1 << NBBY); i <<= 1, ipref++) {
                   1446:                if ((map & i) == 0) {
                   1447:                        cgp->cg_irotor = ipref;
                   1448:                        goto gotit;
                   1449:                }
                   1450:        }
                   1451:        printf("fs = %s\n", fs->fs_fsmnt);
                   1452:        panic("ffs_nodealloccg: block not in map");
                   1453:        /* NOTREACHED */
                   1454: gotit:
                   1455:        setbit(cg_inosused(cgp), ipref);
                   1456:        cgp->cg_cs.cs_nifree--;
                   1457:        fs->fs_cstotal.cs_nifree--;
                   1458:        fs->fs_cs(fs, cg).cs_nifree--;
                   1459:        fs->fs_fmod = 1;
                   1460:        if ((mode & IFMT) == IFDIR) {
                   1461:                cgp->cg_cs.cs_ndir++;
                   1462:                fs->fs_cstotal.cs_ndir++;
                   1463:                fs->fs_cs(fs, cg).cs_ndir++;
                   1464:        }
                   1465: #if REV_ENDIAN_FS
                   1466:        if (rev_endian)
                   1467:                byte_swap_cgout(cgp,fs);
                   1468: #endif /* REV_ENDIAN_FS */
                   1469:        bdwrite(bp);
                   1470:        return (cg * fs->fs_ipg + ipref);
                   1471: }
                   1472: 
                   1473: /*
                   1474:  * Free a block or fragment.
                   1475:  *
                   1476:  * The specified block or fragment is placed back in the
                   1477:  * free map. If a fragment is deallocated, a possible 
                   1478:  * block reassembly is checked.
                   1479:  */
                   1480: ffs_blkfree(ip, bno, size)
                   1481:        register struct inode *ip;
                   1482:        ufs_daddr_t bno;
                   1483:        long size;
                   1484: {
                   1485:        register struct fs *fs;
                   1486:        register struct cg *cgp;
                   1487:        struct buf *bp;
                   1488:        ufs_daddr_t blkno;
                   1489:        int i, error, cg, blk, frags, bbase;
                   1490: #if REV_ENDIAN_FS
                   1491:        struct vnode *vp=ITOV(ip);
                   1492:        struct mount *mp=vp->v_mount;
                   1493:        int rev_endian=(mp->mnt_flag & MNT_REVEND);
                   1494: #endif /* REV_ENDIAN_FS */
                   1495:        fs = ip->i_fs;
                   1496:        if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) {
                   1497:                printf("dev = 0x%x, bsize = %d, size = %d, fs = %s\n",
                   1498:                    ip->i_dev, fs->fs_bsize, size, fs->fs_fsmnt);
                   1499:                panic("blkfree: bad size");
                   1500:        }
                   1501:        cg = dtog(fs, bno);
                   1502:        if ((u_int)bno >= fs->fs_size) {
                   1503:                printf("bad block %d, ino %d\n", bno, ip->i_number);
                   1504:                ffs_fserr(fs, ip->i_uid, "bad block");
                   1505:                return;
                   1506:        }
                   1507:        error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
                   1508:                (int)fs->fs_cgsize, NOCRED, &bp);
                   1509:        if (error) {
                   1510:                brelse(bp);
                   1511:                return;
                   1512:        }
                   1513:        cgp = (struct cg *)bp->b_data;
                   1514: #if REV_ENDIAN_FS
                   1515:        if (rev_endian)
                   1516:                byte_swap_cgin(cgp,fs);
                   1517: #endif /* REV_ENDIAN_FS */
                   1518:        if (!cg_chkmagic(cgp)) {
                   1519: #if REV_ENDIAN_FS
                   1520:                if (rev_endian)
                   1521:                        byte_swap_cgout(cgp,fs);
                   1522: #endif /* REV_ENDIAN_FS */
                   1523:                brelse(bp);
                   1524:                return;
                   1525:        }
                   1526:        cgp->cg_time = time.tv_sec;
                   1527:        bno = dtogd(fs, bno);
                   1528:        if (size == fs->fs_bsize) {
                   1529:                blkno = fragstoblks(fs, bno);
                   1530:                if (ffs_isblock(fs, cg_blksfree(cgp), blkno)) {
                   1531:                        printf("dev = 0x%x, block = %d, fs = %s\n",
                   1532:                            ip->i_dev, bno, fs->fs_fsmnt);
                   1533:                        panic("blkfree: freeing free block");
                   1534:                }
                   1535:                ffs_setblock(fs, cg_blksfree(cgp), blkno);
                   1536:                ffs_clusteracct(fs, cgp, blkno, 1);
                   1537:                cgp->cg_cs.cs_nbfree++;
                   1538:                fs->fs_cstotal.cs_nbfree++;
                   1539:                fs->fs_cs(fs, cg).cs_nbfree++;
                   1540:                i = cbtocylno(fs, bno);
                   1541:                cg_blks(fs, cgp, i)[cbtorpos(fs, bno)]++;
                   1542:                cg_blktot(cgp)[i]++;
                   1543:        } else {
                   1544:                bbase = bno - fragnum(fs, bno);
                   1545:                /*
                   1546:                 * decrement the counts associated with the old frags
                   1547:                 */
                   1548:                blk = blkmap(fs, cg_blksfree(cgp), bbase);
                   1549:                ffs_fragacct(fs, blk, cgp->cg_frsum, -1);
                   1550:                /*
                   1551:                 * deallocate the fragment
                   1552:                 */
                   1553:                frags = numfrags(fs, size);
                   1554:                for (i = 0; i < frags; i++) {
                   1555:                        if (isset(cg_blksfree(cgp), bno + i)) {
                   1556:                                printf("dev = 0x%x, block = %d, fs = %s\n",
                   1557:                                    ip->i_dev, bno + i, fs->fs_fsmnt);
                   1558:                                panic("blkfree: freeing free frag");
                   1559:                        }
                   1560:                        setbit(cg_blksfree(cgp), bno + i);
                   1561:                }
                   1562:                cgp->cg_cs.cs_nffree += i;
                   1563:                fs->fs_cstotal.cs_nffree += i;
                   1564:                fs->fs_cs(fs, cg).cs_nffree += i;
                   1565:                /*
                   1566:                 * add back in counts associated with the new frags
                   1567:                 */
                   1568:                blk = blkmap(fs, cg_blksfree(cgp), bbase);
                   1569:                ffs_fragacct(fs, blk, cgp->cg_frsum, 1);
                   1570:                /*
                   1571:                 * if a complete block has been reassembled, account for it
                   1572:                 */
                   1573:                blkno = fragstoblks(fs, bbase);
                   1574:                if (ffs_isblock(fs, cg_blksfree(cgp), blkno)) {
                   1575:                        cgp->cg_cs.cs_nffree -= fs->fs_frag;
                   1576:                        fs->fs_cstotal.cs_nffree -= fs->fs_frag;
                   1577:                        fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
                   1578:                        ffs_clusteracct(fs, cgp, blkno, 1);
                   1579:                        cgp->cg_cs.cs_nbfree++;
                   1580:                        fs->fs_cstotal.cs_nbfree++;
                   1581:                        fs->fs_cs(fs, cg).cs_nbfree++;
                   1582:                        i = cbtocylno(fs, bbase);
                   1583:                        cg_blks(fs, cgp, i)[cbtorpos(fs, bbase)]++;
                   1584:                        cg_blktot(cgp)[i]++;
                   1585:                }
                   1586:        }
                   1587:        fs->fs_fmod = 1;
                   1588: #if REV_ENDIAN_FS
                   1589:        if (rev_endian)
                   1590:                byte_swap_cgout(cgp,fs);
                   1591: #endif /* REV_ENDIAN_FS */
                   1592:        bdwrite(bp);
                   1593: }
                   1594: 
                   1595: #if DIAGNOSTIC
                   1596: /*
                   1597:  * Verify allocation of a block or fragment. Returns true if block or
                   1598:  * fragment is allocated, false if it is free.
                   1599:  */
                   1600: ffs_checkblk(ip, bno, size)
                   1601:        struct inode *ip;
                   1602:        ufs_daddr_t bno;
                   1603:        long size;
                   1604: {
                   1605:        struct fs *fs;
                   1606:        struct cg *cgp;
                   1607:        struct buf *bp;
                   1608:        int i, error, frags, free;
                   1609: #if REV_ENDIAN_FS
                   1610:        struct vnode *vp=ITOV(ip);
                   1611:        struct mount *mp=vp->v_mount;
                   1612:        int rev_endian=(mp->mnt_flag & MNT_REVEND);
                   1613: #endif /* REV_ENDIAN_FS */
                   1614: 
                   1615:        fs = ip->i_fs;
                   1616:        if ((u_int)size > fs->fs_bsize || fragoff(fs, size) != 0) {
                   1617:                printf("bsize = %d, size = %d, fs = %s\n",
                   1618:                    fs->fs_bsize, size, fs->fs_fsmnt);
                   1619:                panic("checkblk: bad size");
                   1620:        }
                   1621:        if ((u_int)bno >= fs->fs_size)
                   1622:                panic("checkblk: bad block %d", bno);
                   1623:        error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, dtog(fs, bno))),
                   1624:                (int)fs->fs_cgsize, NOCRED, &bp);
                   1625:        if (error) {
                   1626:                brelse(bp);
                   1627:                return;
                   1628:        }
                   1629:        cgp = (struct cg *)bp->b_data;
                   1630: #if REV_ENDIAN_FS
                   1631:        if (rev_endian)
                   1632:                byte_swap_cgin(cgp,fs);
                   1633: #endif /* REV_ENDIAN_FS */
                   1634:        if (!cg_chkmagic(cgp)) {
                   1635: #if REV_ENDIAN_FS
                   1636:                if (rev_endian)
                   1637:                        byte_swap_cgout(cgp,fs);
                   1638: #endif /* REV_ENDIAN_FS */
                   1639:                brelse(bp);
                   1640:                return;
                   1641:        }
                   1642:        bno = dtogd(fs, bno);
                   1643:        if (size == fs->fs_bsize) {
                   1644:                free = ffs_isblock(fs, cg_blksfree(cgp), fragstoblks(fs, bno));
                   1645:        } else {
                   1646:                frags = numfrags(fs, size);
                   1647:                for (free = 0, i = 0; i < frags; i++)
                   1648:                        if (isset(cg_blksfree(cgp), bno + i))
                   1649:                                free++;
                   1650:                if (free != 0 && free != frags)
                   1651:                        panic("checkblk: partially free fragment");
                   1652:        }
                   1653: #if REV_ENDIAN_FS
                   1654:        if (rev_endian)
                   1655:                byte_swap_cgout(cgp,fs);
                   1656: #endif /* REV_ENDIAN_FS */
                   1657:        brelse(bp);
                   1658:        return (!free);
                   1659: }
                   1660: #endif /* DIAGNOSTIC */
                   1661: 
                   1662: /*
                   1663:  * Free an inode.
                   1664:  *
                   1665:  * The specified inode is placed back in the free map.
                   1666:  */
                   1667: int
                   1668: ffs_vfree(ap)
                   1669:        struct vop_vfree_args /* {
                   1670:                struct vnode *a_pvp;
                   1671:                ino_t a_ino;
                   1672:                int a_mode;
                   1673:        } */ *ap;
                   1674: {
                   1675:        register struct fs *fs;
                   1676:        register struct cg *cgp;
                   1677:        register struct inode *pip;
                   1678:        ino_t ino = ap->a_ino;
                   1679:        struct buf *bp;
                   1680:        int error, cg;
                   1681: #if REV_ENDIAN_FS
                   1682:        struct vnode *vp=ap->a_pvp;
                   1683:        struct mount *mp=vp->v_mount;
                   1684:        int rev_endian=(mp->mnt_flag & MNT_REVEND);
                   1685: #endif /* REV_ENDIAN_FS */
                   1686: 
                   1687:        pip = VTOI(ap->a_pvp);
                   1688:        fs = pip->i_fs;
                   1689:        if ((u_int)ino >= fs->fs_ipg * fs->fs_ncg)
                   1690:                panic("ifree: range: dev = 0x%x, ino = %d, fs = %s\n",
                   1691:                    pip->i_dev, ino, fs->fs_fsmnt);
                   1692:        cg = ino_to_cg(fs, ino);
                   1693:        error = bread(pip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
                   1694:                (int)fs->fs_cgsize, NOCRED, &bp);
                   1695:        if (error) {
                   1696:                brelse(bp);
                   1697:                return (0);
                   1698:        }
                   1699:        cgp = (struct cg *)bp->b_data;
                   1700: #if REV_ENDIAN_FS
                   1701:        if (rev_endian)
                   1702:                byte_swap_cgin(cgp,fs);
                   1703: #endif /* REV_ENDIAN_FS */
                   1704:        if (!cg_chkmagic(cgp)) {
                   1705: #if REV_ENDIAN_FS
                   1706:                if (rev_endian)
                   1707:                        byte_swap_cgout(cgp,fs);
                   1708: #endif /* REV_ENDIAN_FS */
                   1709:                brelse(bp);
                   1710:                return (0);
                   1711:        }
                   1712:        cgp->cg_time = time.tv_sec;
                   1713:        ino %= fs->fs_ipg;
                   1714:        if (isclr(cg_inosused(cgp), ino)) {
                   1715:                printf("dev = 0x%x, ino = %d, fs = %s\n",
                   1716:                    pip->i_dev, ino, fs->fs_fsmnt);
                   1717:                if (fs->fs_ronly == 0)
                   1718:                        panic("ifree: freeing free inode");
                   1719:        }
                   1720:        clrbit(cg_inosused(cgp), ino);
                   1721:        if (ino < cgp->cg_irotor)
                   1722:                cgp->cg_irotor = ino;
                   1723:        cgp->cg_cs.cs_nifree++;
                   1724:        fs->fs_cstotal.cs_nifree++;
                   1725:        fs->fs_cs(fs, cg).cs_nifree++;
                   1726:        if ((ap->a_mode & IFMT) == IFDIR) {
                   1727:                cgp->cg_cs.cs_ndir--;
                   1728:                fs->fs_cstotal.cs_ndir--;
                   1729:                fs->fs_cs(fs, cg).cs_ndir--;
                   1730:        }
                   1731:        fs->fs_fmod = 1;
                   1732: #if REV_ENDIAN_FS
                   1733:        if (rev_endian)
                   1734:                byte_swap_cgout(cgp,fs);
                   1735: #endif /* REV_ENDIAN_FS */
                   1736:        bdwrite(bp);
                   1737:        return (0);
                   1738: }
                   1739: 
                   1740: /*
                   1741:  * Find a block of the specified size in the specified cylinder group.
                   1742:  *
                   1743:  * It is a panic if a request is made to find a block if none are
                   1744:  * available.
                   1745:  */
                   1746: static ufs_daddr_t
                   1747: ffs_mapsearch(fs, cgp, bpref, allocsiz)
                   1748:        register struct fs *fs;
                   1749:        register struct cg *cgp;
                   1750:        ufs_daddr_t bpref;
                   1751:        int allocsiz;
                   1752: {
                   1753:        ufs_daddr_t bno;
                   1754:        int start, len, loc, i;
                   1755:        int blk, field, subfield, pos;
                   1756: 
                   1757:        /*
                   1758:         * find the fragment by searching through the free block
                   1759:         * map for an appropriate bit pattern
                   1760:         */
                   1761:        if (bpref)
                   1762:                start = dtogd(fs, bpref) / NBBY;
                   1763:        else
                   1764:                start = cgp->cg_frotor / NBBY;
                   1765:        len = howmany(fs->fs_fpg, NBBY) - start;
                   1766:        loc = scanc((u_int)len, (u_char *)&cg_blksfree(cgp)[start],
                   1767:                (u_char *)fragtbl[fs->fs_frag],
                   1768:                (u_char)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY))));
                   1769:        if (loc == 0) {
                   1770:                len = start + 1;
                   1771:                start = 0;
                   1772:                loc = scanc((u_int)len, (u_char *)&cg_blksfree(cgp)[0],
                   1773:                        (u_char *)fragtbl[fs->fs_frag],
                   1774:                        (u_char)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY))));
                   1775:                if (loc == 0) {
                   1776:                        printf("start = %d, len = %d, fs = %s\n",
                   1777:                            start, len, fs->fs_fsmnt);
                   1778:                        panic("ffs_alloccg: map corrupted");
                   1779:                        /* NOTREACHED */
                   1780:                }
                   1781:        }
                   1782:        bno = (start + len - loc) * NBBY;
                   1783:        cgp->cg_frotor = bno;
                   1784:        /*
                   1785:         * found the byte in the map
                   1786:         * sift through the bits to find the selected frag
                   1787:         */
                   1788:        for (i = bno + NBBY; bno < i; bno += fs->fs_frag) {
                   1789:                blk = blkmap(fs, cg_blksfree(cgp), bno);
                   1790:                blk <<= 1;
                   1791:                field = around[allocsiz];
                   1792:                subfield = inside[allocsiz];
                   1793:                for (pos = 0; pos <= fs->fs_frag - allocsiz; pos++) {
                   1794:                        if ((blk & field) == subfield)
                   1795:                                return (bno + pos);
                   1796:                        field <<= 1;
                   1797:                        subfield <<= 1;
                   1798:                }
                   1799:        }
                   1800:        printf("bno = %d, fs = %s\n", bno, fs->fs_fsmnt);
                   1801:        panic("ffs_alloccg: block not in map");
                   1802:        return (-1);
                   1803: }
                   1804: 
                   1805: /*
                   1806:  * Update the cluster map because of an allocation or free.
                   1807:  *
                   1808:  * Cnt == 1 means free; cnt == -1 means allocating.
                   1809:  */
                   1810: ffs_clusteracct(fs, cgp, blkno, cnt)
                   1811:        struct fs *fs;
                   1812:        struct cg *cgp;
                   1813:        ufs_daddr_t blkno;
                   1814:        int cnt;
                   1815: {
                   1816:        int32_t *sump;
                   1817:        int32_t *lp;
                   1818:        u_char *freemapp, *mapp;
                   1819:        int i, start, end, forw, back, map, bit;
                   1820: 
                   1821:        if (fs->fs_contigsumsize <= 0)
                   1822:                return;
                   1823:        freemapp = cg_clustersfree(cgp);
                   1824:        sump = cg_clustersum(cgp);
                   1825:        /*
                   1826:         * Allocate or clear the actual block.
                   1827:         */
                   1828:        if (cnt > 0)
                   1829:                setbit(freemapp, blkno);
                   1830:        else
                   1831:                clrbit(freemapp, blkno);
                   1832:        /*
                   1833:         * Find the size of the cluster going forward.
                   1834:         */
                   1835:        start = blkno + 1;
                   1836:        end = start + fs->fs_contigsumsize;
                   1837:        if (end >= cgp->cg_nclusterblks)
                   1838:                end = cgp->cg_nclusterblks;
                   1839:        mapp = &freemapp[start / NBBY];
                   1840:        map = *mapp++;
                   1841:        bit = 1 << (start % NBBY);
                   1842:        for (i = start; i < end; i++) {
                   1843:                if ((map & bit) == 0)
                   1844:                        break;
                   1845:                if ((i & (NBBY - 1)) != (NBBY - 1)) {
                   1846:                        bit <<= 1;
                   1847:                } else {
                   1848:                        map = *mapp++;
                   1849:                        bit = 1;
                   1850:                }
                   1851:        }
                   1852:        forw = i - start;
                   1853:        /*
                   1854:         * Find the size of the cluster going backward.
                   1855:         */
                   1856:        start = blkno - 1;
                   1857:        end = start - fs->fs_contigsumsize;
                   1858:        if (end < 0)
                   1859:                end = -1;
                   1860:        mapp = &freemapp[start / NBBY];
                   1861:        map = *mapp--;
                   1862:        bit = 1 << (start % NBBY);
                   1863:        for (i = start; i > end; i--) {
                   1864:                if ((map & bit) == 0)
                   1865:                        break;
                   1866:                if ((i & (NBBY - 1)) != 0) {
                   1867:                        bit >>= 1;
                   1868:                } else {
                   1869:                        map = *mapp--;
                   1870:                        bit = 1 << (NBBY - 1);
                   1871:                }
                   1872:        }
                   1873:        back = start - i;
                   1874:        /*
                   1875:         * Account for old cluster and the possibly new forward and
                   1876:         * back clusters.
                   1877:         */
                   1878:        i = back + forw + 1;
                   1879:        if (i > fs->fs_contigsumsize)
                   1880:                i = fs->fs_contigsumsize;
                   1881:        sump[i] += cnt;
                   1882:        if (back > 0)
                   1883:                sump[back] -= cnt;
                   1884:        if (forw > 0)
                   1885:                sump[forw] -= cnt;
                   1886:        /*
                   1887:         * Update cluster summary information.
                   1888:         */
                   1889:        lp = &sump[fs->fs_contigsumsize];
                   1890:        for (i = fs->fs_contigsumsize; i > 0; i--)
                   1891:                if (*lp-- > 0)
                   1892:                        break;
                   1893:        fs->fs_maxcluster[cgp->cg_cgx] = i;
                   1894: }
                   1895: 
                   1896: /*
                   1897:  * Fserr prints the name of a file system with an error diagnostic.
                   1898:  * 
                   1899:  * The form of the error message is:
                   1900:  *     fs: error message
                   1901:  */
                   1902: static void
                   1903: ffs_fserr(fs, uid, cp)
                   1904:        struct fs *fs;
                   1905:        u_int uid;
                   1906:        char *cp;
                   1907: {
                   1908: 
                   1909:        log(LOG_ERR, "uid %d on %s: %s\n", uid, fs->fs_fsmnt, cp);
                   1910: }

unix.superglobalmegacorp.com

This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.