macroize cylinder group array references to allow dynamic tables

[unix-history] / usr / src / sys / ufs / lfs / lfs_alloc.c

diff --git a/usr/src/sys/ufs/lfs/lfs_alloc.c b/usr/src/sys/ufs/lfs/lfs_alloc.c
index d51505b..04ca4b1 100644 (file)
--- a/usr/src/sys/ufs/lfs/lfs_alloc.c
+++ b/usr/src/sys/ufs/lfs/lfs_alloc.c
@@ -1,32 +1,55 @@
-/* Copyright (c) 1981 Regents of the University of California */
-
-static char vers[] = "@(#)lfs_alloc.c 1.7 %G%";
-
-/*     alloc.c 4.8     81/03/08        */
+/*
+ * Copyright (c) 1982, 1986 Regents of the University of California.
+ * All rights reserved.  The Berkeley software License Agreement
+ * specifies the terms and conditions for redistribution.
+ *
+ *     @(#)lfs_alloc.c 7.5 (Berkeley) %G%
+ */

 
 

-#include "../h/param.h"
-#include "../h/systm.h"
-#include "../h/mount.h"
-#include "../h/fs.h"
-#include "../h/conf.h"
-#include "../h/buf.h"
-#include "../h/inode.h"
-#include "../h/dir.h"
-#include "../h/user.h"
+#include "param.h"
+#include "systm.h"
+#include "mount.h"
+#include "fs.h"
+#include "buf.h"
+#include "inode.h"
+#include "dir.h"
+#include "user.h"
+#include "quota.h"
+#include "kernel.h"
+#include "syslog.h"
+#include "cmap.h"

 
 

-extern long            hashalloc();
-extern long            ialloccg();
+extern u_long          hashalloc();
+extern ino_t           ialloccg();

 extern daddr_t         alloccg();
 extern daddr_t         alloccgblk();
 extern daddr_t         fragextend();
 extern daddr_t         blkpref();
 extern daddr_t         mapsearch();
 extern int             inside[], around[];
 extern daddr_t         alloccg();
 extern daddr_t         alloccgblk();
 extern daddr_t         fragextend();
 extern daddr_t         blkpref();
 extern daddr_t         mapsearch();
 extern int             inside[], around[];

-extern unsigned char   fragtbl[];
+extern unsigned char   *fragtbl[];

 
 

+/*
+ * Allocate a block in the file system.
+ * 
+ * The size of the requested block is given, which must be some
+ * multiple of fs_fsize and <= fs_bsize.
+ * A preference may be optionally specified. If a preference is given
+ * the following hierarchy is used to allocate a block:
+ *   1) allocate the requested block.
+ *   2) allocate a rotationally optimal block in the same cylinder.
+ *   3) allocate a block in the same cylinder group.
+ *   4) quadradically rehash into other cylinder groups, until an
+ *      available block is located.
+ * If no block preference is given the following heirarchy is used
+ * to allocate a block:
+ *   1) allocate a block in the cylinder group that contains the
+ *      inode for the file.
+ *   2) quadradically rehash into other cylinder groups, until an
+ *      available block is located.
+ */

 struct buf *
 struct buf *

-alloc(dev, ip, bpref, size)
-       dev_t dev;
+alloc(ip, bpref, size)

        register struct inode *ip;
        daddr_t bpref;
        int size;
        register struct inode *ip;
        daddr_t bpref;
        int size;


@@ -36,21 +59,38 @@ alloc(dev, ip, bpref, size)
        register struct buf *bp;
        int cg;
        
        register struct buf *bp;
        int cg;
        

-       if ((unsigned)size > BSIZE || size % FSIZE != 0)
+       fs = ip->i_fs;
+       if ((unsigned)size > fs->fs_bsize || fragoff(fs, size) != 0) {
+               printf("dev = 0x%x, bsize = %d, size = %d, fs = %s\n",
+                   ip->i_dev, fs->fs_bsize, size, fs->fs_fsmnt);

                panic("alloc: bad size");
                panic("alloc: bad size");

-       fs = getfs(dev);
-       if (size == BSIZE &&
-           (fs->fs_nbfree == 0 ||
-           (!suser() && fs->fs_nbfree < fs->fs_nbfree * minfree / 100)))
+       }
+       if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0)
+               goto nospace;
+       if (u.u_uid != 0 && freespace(fs, fs->fs_minfree) <= 0)

                goto nospace;
                goto nospace;

+#ifdef QUOTA
+       u.u_error = chkdq(ip, (long)btodb(size), 0);
+       if (u.u_error)
+               return (NULL);
+#endif
+       if (bpref >= fs->fs_size)
+               bpref = 0;

        if (bpref == 0)
        if (bpref == 0)

-               cg = itog(ip->i_number, fs);
+               cg = itog(fs, ip->i_number);

        else
        else

-               cg = dtog(bpref, fs);
-       bno = hashalloc(dev, fs, cg, (long)bpref, size, alloccg);
-       if (bno == 0)
+               cg = dtog(fs, bpref);
+       bno = (daddr_t)hashalloc(ip, cg, (long)bpref, size,
+               (u_long (*)())alloccg);
+       if (bno <= 0)

                goto nospace;
                goto nospace;

-       bp = getblk(dev, bno, size);
+       ip->i_blocks += btodb(size);
+       ip->i_flag |= IUPD|ICHG;
+#ifdef SECSIZE
+       bp = getblk(ip->i_dev, fsbtodb(fs, bno), size, fs->fs_dbsize);
+#else SECSIZE
+       bp = getblk(ip->i_dev, fsbtodb(fs, bno), size);
+#endif SECSIZE

        clrbuf(bp);
        return (bp);
 nospace:
        clrbuf(bp);
        return (bp);
 nospace:


@@ -60,50 +100,153 @@ nospace:
        return (NULL);
 }
 
        return (NULL);
 }
 

+/*
+ * Reallocate a fragment to a bigger size
+ *
+ * The number and size of the old block is given, and a preference
+ * and new size is also specified. The allocator attempts to extend
+ * the original block. Failing that, the regular block allocator is
+ * invoked to get an appropriate block.
+ */

 struct buf *
 struct buf *

-realloccg(dev, ip, bprev, bpref, osize, nsize)
-       dev_t dev;
+realloccg(ip, bprev, bpref, osize, nsize)

        register struct inode *ip;
        daddr_t bprev, bpref;
        int osize, nsize;
 {
        register struct inode *ip;
        daddr_t bprev, bpref;
        int osize, nsize;
 {

-       daddr_t bno;

        register struct fs *fs;
        register struct buf *bp, *obp;
        register struct fs *fs;
        register struct buf *bp, *obp;

-       caddr_t cp;
-       int cg;
+       int cg, request;
+       daddr_t bno, bn;
+       int i, count;

        
        

-       if ((unsigned)osize > BSIZE || osize % FSIZE != 0 ||
-           (unsigned)nsize > BSIZE || nsize % FSIZE != 0)
+       fs = ip->i_fs;
+       if ((unsigned)osize > fs->fs_bsize || fragoff(fs, osize) != 0 ||
+           (unsigned)nsize > fs->fs_bsize || fragoff(fs, nsize) != 0) {
+               printf("dev = 0x%x, bsize = %d, osize = %d, nsize = %d, fs = %s\n",
+                   ip->i_dev, fs->fs_bsize, osize, nsize, fs->fs_fsmnt);

                panic("realloccg: bad size");
                panic("realloccg: bad size");

-       fs = getfs(dev);
-       if (bprev == 0)
+       }
+       if (u.u_uid != 0 && freespace(fs, fs->fs_minfree) <= 0)
+               goto nospace;
+       if (bprev == 0) {
+               printf("dev = 0x%x, bsize = %d, bprev = %d, fs = %s\n",
+                   ip->i_dev, fs->fs_bsize, bprev, fs->fs_fsmnt);

                panic("realloccg: bad bprev");
                panic("realloccg: bad bprev");

-       else
-               cg = dtog(bprev, fs);
-       bno = fragextend(dev, fs, cg, (long)bprev, osize, nsize);
+       }
+#ifdef QUOTA
+       u.u_error = chkdq(ip, (long)btodb(nsize - osize), 0);
+       if (u.u_error)
+               return (NULL);
+#endif
+       cg = dtog(fs, bprev);
+       bno = fragextend(ip, cg, (long)bprev, osize, nsize);

        if (bno != 0) {
        if (bno != 0) {

-               bp = bread(dev, bno, osize);
-               bp->b_bcount = nsize;
-               blkclr(bp->b_un.b_addr + osize, nsize - osize);
+               do {
+#ifdef SECSIZE
+                       bp = bread(ip->i_dev, fsbtodb(fs, bno), osize,
+                           fs->fs_dbsize);
+#else SECSIZE
+                       bp = bread(ip->i_dev, fsbtodb(fs, bno), osize);
+#endif SECSIZE
+                       if (bp->b_flags & B_ERROR) {
+                               brelse(bp);
+                               return (NULL);
+                       }
+               } while (brealloc(bp, nsize) == 0);
+               bp->b_flags |= B_DONE;
+               bzero(bp->b_un.b_addr + osize, (unsigned)nsize - osize);
+               ip->i_blocks += btodb(nsize - osize);
+               ip->i_flag |= IUPD|ICHG;

                return (bp);
        }
                return (bp);
        }

-       bno = hashalloc(dev, fs, cg, (long)bpref, nsize, alloccg);
-       if (bno != 0) {
+       if (bpref >= fs->fs_size)
+               bpref = 0;
+       switch ((int)fs->fs_optim) {
+       case FS_OPTSPACE:
+               /*
+                * Allocate an exact sized fragment. Although this makes 
+                * best use of space, we will waste time relocating it if 
+                * the file continues to grow. If the fragmentation is
+                * less than half of the minimum free reserve, we choose
+                * to begin optimizing for time.
+                */
+               request = nsize;
+               if (fs->fs_minfree < 5 ||
+                   fs->fs_cstotal.cs_nffree >
+                   fs->fs_dsize * fs->fs_minfree / (2 * 100))
+                       break;
+               log(LOG_NOTICE, "%s: optimization changed from SPACE to TIME\n",
+                       fs->fs_fsmnt);
+               fs->fs_optim = FS_OPTTIME;
+               break;
+       case FS_OPTTIME:

                /*
                /*

-                * make a new copy
+                * At this point we have discovered a file that is trying
+                * to grow a small fragment to a larger fragment. To save
+                * time, we allocate a full sized block, then free the 
+                * unused portion. If the file continues to grow, the 
+                * `fragextend' call above will be able to grow it in place
+                * without further copying. If aberrant programs cause
+                * disk fragmentation to grow within 2% of the free reserve,
+                * we choose to begin optimizing for space.

                 */
                 */

-               obp = bread(dev, bprev, osize);
-               bp = getblk(dev, bno, nsize);
-               cp = bp->b_un.b_addr;
-               bp->b_un.b_addr = obp->b_un.b_addr;
-               obp->b_un.b_addr = cp;
-               obp->b_flags |= B_INVAL;
+               request = fs->fs_bsize;
+               if (fs->fs_cstotal.cs_nffree <
+                   fs->fs_dsize * (fs->fs_minfree - 2) / 100)
+                       break;
+               log(LOG_NOTICE, "%s: optimization changed from TIME to SPACE\n",
+                       fs->fs_fsmnt);
+               fs->fs_optim = FS_OPTSPACE;
+               break;
+       default:
+               printf("dev = 0x%x, optim = %d, fs = %s\n",
+                   ip->i_dev, fs->fs_optim, fs->fs_fsmnt);
+               panic("realloccg: bad optim");
+               /* NOTREACHED */
+       }
+       bno = (daddr_t)hashalloc(ip, cg, (long)bpref, request,
+               (u_long (*)())alloccg);
+       if (bno > 0) {
+#ifdef SECSIZE
+               obp = bread(ip->i_dev, fsbtodb(fs, bprev), osize,
+                   fs->fs_dbsize);
+#else SECSIZE
+               obp = bread(ip->i_dev, fsbtodb(fs, bprev), osize);
+#endif SECSIZE
+               if (obp->b_flags & B_ERROR) {
+                       brelse(obp);
+                       return (NULL);
+               }
+               bn = fsbtodb(fs, bno);
+#ifdef SECSIZE
+               bp = getblk(ip->i_dev, bn, nsize, fs->fs_dbsize);
+#else SECSIZE
+               bp = getblk(ip->i_dev, bn, nsize);
+#endif SECSIZE
+               bcopy(obp->b_un.b_addr, bp->b_un.b_addr, (u_int)osize);
+               count = howmany(osize, CLBYTES);
+               for (i = 0; i < count; i++)
+#ifdef SECSIZE
+                       munhash(ip->i_dev, bn + i * CLBYTES / fs->fs_dbsize);
+#else SECSIZE
+                       munhash(ip->i_dev, bn + i * CLBYTES / DEV_BSIZE);
+#endif SECSIZE
+               bzero(bp->b_un.b_addr + osize, (unsigned)nsize - osize);
+               if (obp->b_flags & B_DELWRI) {
+                       obp->b_flags &= ~B_DELWRI;
+                       u.u_ru.ru_oublock--;            /* delete charge */
+               }

                brelse(obp);
                brelse(obp);

-               fre(dev, bprev, osize);
-               blkclr(bp->b_un.b_addr + osize, nsize - osize);
-               return(bp);
+               blkfree(ip, bprev, (off_t)osize);
+               if (nsize < request)
+                       blkfree(ip, bno + numfrags(fs, nsize),
+                               (off_t)(request - nsize));
+               ip->i_blocks += btodb(nsize - osize);
+               ip->i_flag |= IUPD|ICHG;
+               return (bp);

        }
        }

+nospace:

        /*
         * no space available
         */
        /*
         * no space available
         */


@@ -113,101 +256,209 @@ realloccg(dev, ip, bprev, bpref, osize, nsize)
        return (NULL);
 }
 
        return (NULL);
 }
 

+/*
+ * Allocate an inode in the file system.
+ * 
+ * A preference may be optionally specified. If a preference is given
+ * the following hierarchy is used to allocate an inode:
+ *   1) allocate the requested inode.
+ *   2) allocate an inode in the same cylinder group.
+ *   3) quadradically rehash into other cylinder groups, until an
+ *      available inode is located.
+ * If no inode preference is given the following heirarchy is used
+ * to allocate an inode:
+ *   1) allocate an inode in cylinder group 0.
+ *   2) quadradically rehash into other cylinder groups, until an
+ *      available inode is located.
+ */

 struct inode *
 struct inode *

-ialloc(dev, ipref, mode)
-       dev_t dev;
+ialloc(pip, ipref, mode)
+       register struct inode *pip;

        ino_t ipref;
        int mode;
 {
        ino_t ipref;
        int mode;
 {

-       daddr_t ino;
+       ino_t ino;

        register struct fs *fs;
        register struct inode *ip;
        int cg;
        
        register struct fs *fs;
        register struct inode *ip;
        int cg;
        

-       fs = getfs(dev);
-       if (fs->fs_nifree == 0)
+       fs = pip->i_fs;
+       if (fs->fs_cstotal.cs_nifree == 0)

                goto noinodes;
                goto noinodes;

-       cg = itog(ipref, fs);
-       ino = hashalloc(dev, fs, cg, (long)ipref, mode, ialloccg);
+#ifdef QUOTA
+       u.u_error = chkiq(pip->i_dev, (struct inode *)NULL, u.u_uid, 0);
+       if (u.u_error)
+               return (NULL);
+#endif
+       if (ipref >= fs->fs_ncg * fs->fs_ipg)
+               ipref = 0;
+       cg = itog(fs, ipref);
+       ino = (ino_t)hashalloc(pip, cg, (long)ipref, mode, ialloccg);

        if (ino == 0)
                goto noinodes;
        if (ino == 0)
                goto noinodes;

-       ip = iget(dev, ino);
+       ip = iget(pip->i_dev, pip->i_fs, ino);

        if (ip == NULL) {
        if (ip == NULL) {

-               ifree(dev, ino);
+               ifree(pip, ino, 0);

                return (NULL);
        }
                return (NULL);
        }

-       if (ip->i_mode)
+       if (ip->i_mode) {
+               printf("mode = 0%o, inum = %d, fs = %s\n",
+                   ip->i_mode, ip->i_number, fs->fs_fsmnt);

                panic("ialloc: dup alloc");
                panic("ialloc: dup alloc");

+       }
+       if (ip->i_blocks) {                             /* XXX */
+               printf("free inode %s/%d had %d blocks\n",
+                   fs->fs_fsmnt, ino, ip->i_blocks);
+               ip->i_blocks = 0;
+       }

        return (ip);
 noinodes:
        fserr(fs, "out of inodes");
        return (ip);
 noinodes:
        fserr(fs, "out of inodes");

-       uprintf("\n%s: create failed, no inodes free\n", fs->fs_fsmnt);
+       uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt);

        u.u_error = ENOSPC;
        return (NULL);
 }
 
 /*
        u.u_error = ENOSPC;
        return (NULL);
 }
 
 /*

- * find a cylinder to place a directory
+ * Find a cylinder to place a directory.
+ *
+ * The policy implemented by this algorithm is to select from
+ * among those cylinder groups with above the average number of
+ * free inodes, the one with the smallest number of directories.

  */
  */

-dirpref(dev)
-       dev_t dev;
-{
+ino_t
+dirpref(fs)

        register struct fs *fs;
        register struct fs *fs;

+{

        int cg, minndir, mincg, avgifree;
 
        int cg, minndir, mincg, avgifree;
 

-       fs = getfs(dev);
-       avgifree = fs->fs_nifree / fs->fs_ncg;
+       avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg;

        minndir = fs->fs_ipg;
        mincg = 0;
        for (cg = 0; cg < fs->fs_ncg; cg++)
        minndir = fs->fs_ipg;
        mincg = 0;
        for (cg = 0; cg < fs->fs_ncg; cg++)

-               if (fs->fs_cs(cg).cs_ndir < minndir &&
-                   fs->fs_cs(cg).cs_nifree >= avgifree) {
+               if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
+                   fs->fs_cs(fs, cg).cs_nifree >= avgifree) {

                        mincg = cg;
                        mincg = cg;

-                       minndir = fs->fs_cs(cg).cs_ndir;
+                       minndir = fs->fs_cs(fs, cg).cs_ndir;

                }
                }

-       return (fs->fs_ipg * mincg);
+       return ((ino_t)(fs->fs_ipg * mincg));

 }
 
 /*
 }
 
 /*

- * select a cylinder to place a large block of data
+ * Select the desired position for the next block in a file.  The file is
+ * logically divided into sections. The first section is composed of the
+ * direct blocks. Each additional section contains fs_maxbpg blocks.
+ * 
+ * If no blocks have been allocated in the first section, the policy is to
+ * request a block in the same cylinder group as the inode that describes
+ * the file. If no blocks have been allocated in any other section, the
+ * policy is to place the section in a cylinder group with a greater than
+ * average number of free blocks.  An appropriate cylinder group is found
+ * by using a rotor that sweeps the cylinder groups. When a new group of
+ * blocks is needed, the sweep begins in the cylinder group following the
+ * cylinder group from which the previous allocation was made. The sweep
+ * continues until a cylinder group with greater than the average number
+ * of free blocks is found. If the allocation is for the first block in an
+ * indirect block, the information on the previous allocation is unavailable;
+ * here a best guess is made based upon the logical block number being
+ * allocated.
+ * 
+ * If a section is already partially allocated, the policy is to
+ * contiguously allocate fs_maxcontig blocks.  The end of one of these
+ * contiguous blocks and the beginning of the next is physically separated
+ * so that the disk head will be in transit between them for at least
+ * fs_rotdelay milliseconds.  This is to allow time for the processor to
+ * schedule another I/O transfer.

  */
  */

-blkpref(dev)
-       dev_t dev;
+daddr_t
+blkpref(ip, lbn, indx, bap)
+       struct inode *ip;
+       daddr_t lbn;
+       int indx;
+       daddr_t *bap;

 {
        register struct fs *fs;
 {
        register struct fs *fs;

-       int cg, avgbfree;
+       register int cg;
+       int avgbfree, startcg;
+       daddr_t nextblk;

 
 

-       fs = getfs(dev);
-       avgbfree = fs->fs_nbfree / fs->fs_ncg;
-       for (cg = fs->fs_cgrotor + 1; cg < fs->fs_ncg; cg++)
-               if (fs->fs_cs(cg).cs_nbfree >= avgbfree) {
-                       fs->fs_cgrotor = cg;
-                       return (fs->fs_fpg * cg + FRAG);
-               }
-       for (cg = 0; cg <= fs->fs_cgrotor; cg++)
-               if (fs->fs_cs(cg).cs_nbfree >= avgbfree) {
-                       fs->fs_cgrotor = cg;
-                       return (fs->fs_fpg * cg + FRAG);
+       fs = ip->i_fs;
+       if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
+               if (lbn < NDADDR) {
+                       cg = itog(fs, ip->i_number);
+                       return (fs->fs_fpg * cg + fs->fs_frag);

                }
                }

-       return (0);
+               /*
+                * Find a cylinder with greater than average number of
+                * unused data blocks.
+                */
+               if (indx == 0 || bap[indx - 1] == 0)
+                       startcg = itog(fs, ip->i_number) + lbn / fs->fs_maxbpg;
+               else
+                       startcg = dtog(fs, bap[indx - 1]) + 1;
+               startcg %= fs->fs_ncg;
+               avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
+               for (cg = startcg; cg < fs->fs_ncg; cg++)
+                       if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
+                               fs->fs_cgrotor = cg;
+                               return (fs->fs_fpg * cg + fs->fs_frag);
+                       }
+               for (cg = 0; cg <= startcg; cg++)
+                       if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
+                               fs->fs_cgrotor = cg;
+                               return (fs->fs_fpg * cg + fs->fs_frag);
+                       }
+               return (NULL);
+       }
+       /*
+        * One or more previous blocks have been laid out. If less
+        * than fs_maxcontig previous blocks are contiguous, the
+        * next block is requested contiguously, otherwise it is
+        * requested rotationally delayed by fs_rotdelay milliseconds.
+        */
+       nextblk = bap[indx - 1] + fs->fs_frag;
+       if (indx > fs->fs_maxcontig &&
+           bap[indx - fs->fs_maxcontig] + blkstofrags(fs, fs->fs_maxcontig)
+           != nextblk)
+               return (nextblk);
+       if (fs->fs_rotdelay != 0)
+               /*
+                * Here we convert ms of delay to frags as:
+                * (frags) = (ms) * (rev/sec) * (sect/rev) /
+                *      ((sect/frag) * (ms/sec))
+                * then round up to the next block.
+                */
+               nextblk += roundup(fs->fs_rotdelay * fs->fs_rps * fs->fs_nsect /
+                   (NSPF(fs) * 1000), fs->fs_frag);
+       return (nextblk);

 }
 
 }
 

-long
-hashalloc(dev, fs, cg, pref, size, allocator)
-       dev_t dev;
-       register struct fs *fs;
+/*
+ * Implement the cylinder overflow algorithm.
+ *
+ * The policy implemented by this algorithm is:
+ *   1) allocate the block in its requested cylinder group.
+ *   2) quadradically rehash on the cylinder group number.
+ *   3) brute force search for a free block.
+ */
+/*VARARGS5*/
+u_long
+hashalloc(ip, cg, pref, size, allocator)
+       struct inode *ip;

        int cg;
        long pref;
        int size;       /* size for data blocks, mode for inodes */
        int cg;
        long pref;
        int size;       /* size for data blocks, mode for inodes */

-       long (*allocator)();
+       u_long (*allocator)();

 {
 {

+       register struct fs *fs;

        long result;
        int i, icg = cg;
 
        long result;
        int i, icg = cg;
 

+       fs = ip->i_fs;

        /*
         * 1: preferred cylinder group
         */
        /*
         * 1: preferred cylinder group
         */

-       result = (*allocator)(dev, fs, cg, pref, size);
+       result = (*allocator)(ip, cg, pref, size);

        if (result)
                return (result);
        /*
        if (result)
                return (result);
        /*


@@ -217,104 +468,135 @@ hashalloc(dev, fs, cg, pref, size, allocator)
                cg += i;
                if (cg >= fs->fs_ncg)
                        cg -= fs->fs_ncg;
                cg += i;
                if (cg >= fs->fs_ncg)
                        cg -= fs->fs_ncg;

-               result = (*allocator)(dev, fs, cg, 0, size);
+               result = (*allocator)(ip, cg, 0, size);

                if (result)
                        return (result);
        }
        /*
         * 3: brute force search
                if (result)
                        return (result);
        }
        /*
         * 3: brute force search

+        * Note that we start at i == 2, since 0 was checked initially,
+        * and 1 is always checked in the quadratic rehash.

         */
         */

-       cg = icg;
-       for (i = 0; i < fs->fs_ncg; i++) {
-               result = (*allocator)(dev, fs, cg, 0, size);
+       cg = (icg + 2) % fs->fs_ncg;
+       for (i = 2; i < fs->fs_ncg; i++) {
+               result = (*allocator)(ip, cg, 0, size);

                if (result)
                        return (result);
                cg++;
                if (cg == fs->fs_ncg)
                        cg = 0;
        }
                if (result)
                        return (result);
                cg++;
                if (cg == fs->fs_ncg)
                        cg = 0;
        }

-       return (0);
+       return (NULL);

 }
 
 }
 

+/*
+ * Determine whether a fragment can be extended.
+ *
+ * Check to see if the necessary fragments are available, and 
+ * if they are, allocate them.
+ */

 daddr_t
 daddr_t

-fragextend(dev, fs, cg, bprev, osize, nsize)
-       dev_t dev;
-       register struct fs *fs;
+fragextend(ip, cg, bprev, osize, nsize)
+       struct inode *ip;

        int cg;
        long bprev;
        int osize, nsize;
 {
        int cg;
        long bprev;
        int osize, nsize;
 {

+       register struct fs *fs;

        register struct buf *bp;
        register struct cg *cgp;
        long bno;
        int frags, bbase;
        int i;
 
        register struct buf *bp;
        register struct cg *cgp;
        long bno;
        int frags, bbase;
        int i;
 

-       frags = nsize / FSIZE;
-       bbase = bprev % FRAG;
-       if (bbase > (bprev + frags - 1) % FRAG) {
-               /* cannot extend across a block boundry */
-               return (0);
-       }
-       bp = bread(dev, cgtod(cg, fs), BSIZE);
-       if (bp->b_flags & B_ERROR)
-               return (0);
+       fs = ip->i_fs;
+       if (fs->fs_cs(fs, cg).cs_nffree < numfrags(fs, nsize - osize))
+               return (NULL);
+       frags = numfrags(fs, nsize);
+       bbase = fragnum(fs, bprev);
+       if (bbase > fragnum(fs, (bprev + frags - 1))) {
+               /* cannot extend across a block boundary */
+               return (NULL);
+       }
+#ifdef SECSIZE
+       bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize,
+           fs->fs_dbsize);
+#else SECSIZE
+       bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize);
+#endif SECSIZE

        cgp = bp->b_un.b_cg;
        cgp = bp->b_un.b_cg;

-       bno = bprev % fs->fs_fpg;
-       for (i = osize / FSIZE; i < frags; i++) {
-               if (isclr(cgp->cg_free, bno + i))
-                       break;
+       if (bp->b_flags & B_ERROR || !cg_chkmagic(cgp)) {
+               brelse(bp);
+               return (NULL);

        }
        }

-       if (i == frags) {
-               /*
-                * the current fragment can be extended
-                * deduct the count on fragment being extended into
-                * increase the count on the remaining fragment (if any)
-                * allocate the extended piece
-                */
-               for (i = frags; i < FRAG - bbase; i++)
-                       if (isclr(cgp->cg_free, bno + i))
-                               break;
-               cgp->cg_frsum[i - osize / FSIZE]--;
-               if (i != frags)
-                       cgp->cg_frsum[i - frags]++;
-               for (i = osize / FSIZE; i < frags; i++) {
-                       clrbit(cgp->cg_free, bno + i);
-                       cgp->cg_nffree--;
-                       fs->fs_nffree--;
+       cgp->cg_time = time.tv_sec;
+       bno = dtogd(fs, bprev);
+       for (i = numfrags(fs, osize); i < frags; i++)
+               if (isclr(cg_blksfree(cgp), bno + i)) {
+                       brelse(bp);
+                       return (NULL);

                }
                }

-               fs->fs_fmod++;
-               bdwrite(bp);
-               return (bprev);
+       /*
+        * the current fragment can be extended
+        * deduct the count on fragment being extended into
+        * increase the count on the remaining fragment (if any)
+        * allocate the extended piece
+        */
+       for (i = frags; i < fs->fs_frag - bbase; i++)
+               if (isclr(cg_blksfree(cgp), bno + i))
+                       break;
+       cgp->cg_frsum[i - numfrags(fs, osize)]--;
+       if (i != frags)
+               cgp->cg_frsum[i - frags]++;
+       for (i = numfrags(fs, osize); i < frags; i++) {
+               clrbit(cg_blksfree(cgp), bno + i);
+               cgp->cg_cs.cs_nffree--;
+               fs->fs_cstotal.cs_nffree--;
+               fs->fs_cs(fs, cg).cs_nffree--;

        }
        }

-       brelse(bp);
-       return (0);
+       fs->fs_fmod++;
+       bdwrite(bp);
+       return (bprev);

 }
 
 }
 

+/*
+ * Determine whether a block can be allocated.
+ *
+ * Check to see if a block of the apprpriate size is available,
+ * and if it is, allocate it.
+ */

 daddr_t
 daddr_t

-alloccg(dev, fs, cg, bpref, size)
-       dev_t dev;
-       register struct fs *fs;
+alloccg(ip, cg, bpref, size)
+       struct inode *ip;

        int cg;
        daddr_t bpref;
        int size;
 {
        int cg;
        daddr_t bpref;
        int size;
 {

+       register struct fs *fs;

        register struct buf *bp;
        register struct cg *cgp;
        int bno, frags;
        int allocsiz;
        register int i;
 
        register struct buf *bp;
        register struct cg *cgp;
        int bno, frags;
        int allocsiz;
        register int i;
 

-       bp = bread(dev, cgtod(cg, fs), BSIZE);
-       if (bp->b_flags & B_ERROR)
-               return (0);
+       fs = ip->i_fs;
+       if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize)
+               return (NULL);
+#ifdef SECSIZE
+       bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize,
+           fs->fs_dbsize);
+#else SECSIZE
+       bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize);
+#endif SECSIZE

        cgp = bp->b_un.b_cg;
        cgp = bp->b_un.b_cg;

-       if (size == BSIZE) {
-               if (cgp->cg_nbfree == 0) {
-                       brelse(bp);
-                       return (0);
-               }
-               bno = alloccgblk(dev, fs, cgp, bpref);
+       if (bp->b_flags & B_ERROR || !cg_chkmagic(cgp) ||
+           (cgp->cg_cs.cs_nbfree == 0 && size == fs->fs_bsize)) {
+               brelse(bp);
+               return (NULL);
+       }
+       cgp->cg_time = time.tv_sec;
+       if (size == fs->fs_bsize) {
+               bno = alloccgblk(fs, cgp, bpref);

                bdwrite(bp);
                return (bno);
        }
                bdwrite(bp);
                return (bno);
        }


@@ -323,37 +605,43 @@ alloccg(dev, fs, cg, bpref, size)
         * allocsiz is the size which will be allocated, hacking
         * it down to a smaller size if necessary
         */
         * allocsiz is the size which will be allocated, hacking
         * it down to a smaller size if necessary
         */

-       frags = size / FSIZE;
-       for (allocsiz = frags; allocsiz < FRAG; allocsiz++)
+       frags = numfrags(fs, size);
+       for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++)

                if (cgp->cg_frsum[allocsiz] != 0)
                        break;
                if (cgp->cg_frsum[allocsiz] != 0)
                        break;

-       if (allocsiz == FRAG) {
+       if (allocsiz == fs->fs_frag) {

                /*
                 * no fragments were available, so a block will be 
                 * allocated, and hacked up
                 */
                /*
                 * no fragments were available, so a block will be 
                 * allocated, and hacked up
                 */

-               if (cgp->cg_nbfree == 0) {
+               if (cgp->cg_cs.cs_nbfree == 0) {

                        brelse(bp);
                        brelse(bp);

-                       return (0);
+                       return (NULL);

                }
                }

-               bno = alloccgblk(dev, fs, cgp, bpref);
-               bpref = bno % fs->fs_fpg;
-               for (i = frags; i < FRAG; i++)
-                       setbit(cgp->cg_free, bpref + i);
-               i = FRAG - frags;
-               cgp->cg_nffree += i;
-               fs->fs_nffree += i;
+               bno = alloccgblk(fs, cgp, bpref);
+               bpref = dtogd(fs, bno);
+               for (i = frags; i < fs->fs_frag; i++)
+                       setbit(cg_blksfree(cgp), bpref + i);
+               i = fs->fs_frag - frags;
+               cgp->cg_cs.cs_nffree += i;
+               fs->fs_cstotal.cs_nffree += i;
+               fs->fs_cs(fs, cg).cs_nffree += i;
+               fs->fs_fmod++;

                cgp->cg_frsum[i]++;
                bdwrite(bp);
                return (bno);
        }
        bno = mapsearch(fs, cgp, bpref, allocsiz);
                cgp->cg_frsum[i]++;
                bdwrite(bp);
                return (bno);
        }
        bno = mapsearch(fs, cgp, bpref, allocsiz);

-       if (bno == 0)
-               return (0);
+       if (bno < 0) {
+               brelse(bp);
+               return (NULL);
+       }

        for (i = 0; i < frags; i++)
        for (i = 0; i < frags; i++)

-               clrbit(cgp->cg_free, bno + i);
-       cgp->cg_nffree -= frags;
-       fs->fs_nffree -= frags;
+               clrbit(cg_blksfree(cgp), bno + i);
+       cgp->cg_cs.cs_nffree -= frags;
+       fs->fs_cstotal.cs_nffree -= frags;
+       fs->fs_cs(fs, cg).cs_nffree -= frags;
+       fs->fs_fmod++;

        cgp->cg_frsum[allocsiz]--;
        if (frags != allocsiz)
                cgp->cg_frsum[allocsiz - frags]++;
        cgp->cg_frsum[allocsiz]--;
        if (frags != allocsiz)
                cgp->cg_frsum[allocsiz - frags]++;


@@ -361,128 +649,210 @@ alloccg(dev, fs, cg, bpref, size)
        return (cg * fs->fs_fpg + bno);
 }
 
        return (cg * fs->fs_fpg + bno);
 }
 

+/*
+ * Allocate a block in a cylinder group.
+ *
+ * This algorithm implements the following policy:
+ *   1) allocate the requested block.
+ *   2) allocate a rotationally optimal block in the same cylinder.
+ *   3) allocate the next available block on the block rotor for the
+ *      specified cylinder group.
+ * Note that this routine only allocates fs_bsize blocks; these
+ * blocks may be fragmented by the routine that allocates them.
+ */

 daddr_t
 daddr_t

-alloccgblk(dev, fs, cgp, bpref)
-       dev_t dev;
-       struct fs *fs;
+alloccgblk(fs, cgp, bpref)
+       register struct fs *fs;

        register struct cg *cgp;
        daddr_t bpref;
 {
        daddr_t bno;
        register struct cg *cgp;
        daddr_t bpref;
 {
        daddr_t bno;

-       int cylno, pos;
+       int cylno, pos, delta;

        short *cylbp;
        short *cylbp;

-       int i, j;
+       register int i;

 
        if (bpref == 0) {
                bpref = cgp->cg_rotor;
 
        if (bpref == 0) {
                bpref = cgp->cg_rotor;

-       } else {
-               bpref &= ~(FRAG - 1);
-               bpref %= fs->fs_fpg;
-               /*
-                * if the requested block is available, use it
-                */
-               if (isblock(cgp->cg_free, bpref/FRAG)) {
-                       bno = bpref;
-                       goto gotit;
-               }
+               goto norot;
+       }
+       bpref = blknum(fs, bpref);
+       bpref = dtogd(fs, bpref);
+       /*
+        * if the requested block is available, use it
+        */
+       if (isblock(fs, cg_blksfree(cgp), fragstoblks(fs, bpref))) {
+               bno = bpref;
+               goto gotit;
+       }
+       /*
+        * check for a block available on the same cylinder
+        */
+       cylno = cbtocylno(fs, bpref);
+       if (cg_blktot(cgp)[cylno] == 0)
+               goto norot;
+       if (fs->fs_cpc == 0) {

                /*
                /*

-                * check for a block available on the same cylinder
-                * beginning with one which is rotationally optimal
+                * block layout info is not available, so just have
+                * to take any block in this cylinder.

                 */
                 */

-               i = bpref * NSPF;
-               cylno = i / fs->fs_spc;
-               cylbp = cgp->cg_b[cylno];
-               pos = (i + (ROTDELAY == 0) ?
-                       0 : 1 + ROTDELAY * HZ * fs->fs_nsect / (NSPF * 1000)) %
-                       fs->fs_nsect * NRPOS / fs->fs_nsect;
-               for (i = pos; i < NRPOS; i++)
+               bpref = howmany(fs->fs_spc * cylno, NSPF(fs));
+               goto norot;
+       }
+       /*
+        * check the summary information to see if a block is 
+        * available in the requested cylinder starting at the
+        * requested rotational position and proceeding around.
+        */
+       cylbp = cg_blks(fs, cgp, cylno);
+       pos = cbtorpos(fs, bpref);
+       for (i = pos; i < fs->fs_nrpos; i++)
+               if (cylbp[i] > 0)
+                       break;
+       if (i == fs->fs_nrpos)
+               for (i = 0; i < pos; i++)

                        if (cylbp[i] > 0)
                                break;
                        if (cylbp[i] > 0)
                                break;

-               if (i == NRPOS)
-                       for (i = 0; i < pos; i++)
-                               if (cylbp[i] > 0)
-                                       break;
-               if (cylbp[i] > 0) {
-                       bpref = cylno * fs->fs_spc / (NSPF * FRAG);
-                       for (j = fs->fs_postbl[i]; j > -1; j = fs->fs_rotbl[j]) {
-                               if (isblock(cgp->cg_free, bpref + j)) {
-                                       bno = (bpref + j) * FRAG;
-                                       goto gotit;
-                               }
+       if (cylbp[i] > 0) {
+               /*
+                * found a rotational position, now find the actual
+                * block. A panic if none is actually there.
+                */
+               pos = cylno % fs->fs_cpc;
+               bno = (cylno - pos) * fs->fs_spc / NSPB(fs);
+               if (fs_postbl(fs, pos)[i] == -1) {
+                       printf("pos = %d, i = %d, fs = %s\n",
+                           pos, i, fs->fs_fsmnt);
+                       panic("alloccgblk: cyl groups corrupted");
+               }
+               for (i = fs_postbl(fs, pos)[i];; ) {
+                       if (isblock(fs, cg_blksfree(cgp), bno + i)) {
+                               bno = blkstofrags(fs, (bno + i));
+                               goto gotit;

                        }
                        }

-                       panic("alloccgblk: can't find blk in cyl");
+                       delta = fs_rotbl(fs)[i];
+                       if (delta <= 0 ||
+                           delta + i > fragstoblks(fs, fs->fs_fpg))
+                               break;
+                       i += delta;

                }
                }

+               printf("pos = %d, i = %d, fs = %s\n", pos, i, fs->fs_fsmnt);
+               panic("alloccgblk: can't find blk in cyl");

        }
        }

-       bno = mapsearch(fs, cgp, bpref, FRAG);
-       if (bno == 0)
-               return (0);
+norot:
+       /*
+        * no blocks in the requested cylinder, so take next
+        * available one in this cylinder group.
+        */
+       bno = mapsearch(fs, cgp, bpref, (int)fs->fs_frag);
+       if (bno < 0)
+               return (NULL);

        cgp->cg_rotor = bno;
 gotit:
        cgp->cg_rotor = bno;
 gotit:

-       clrblock(cgp->cg_free, bno/FRAG);
-       cgp->cg_nbfree--;
-       fs->fs_nbfree--;
-       fs->fs_cs(cgp->cg_cgx).cs_nbfree--;
-       i = bno * NSPF;
-       cgp->cg_b[i/fs->fs_spc][i%fs->fs_nsect*NRPOS/fs->fs_nsect]--;
+       clrblock(fs, cg_blksfree(cgp), (long)fragstoblks(fs, bno));
+       cgp->cg_cs.cs_nbfree--;
+       fs->fs_cstotal.cs_nbfree--;
+       fs->fs_cs(fs, cgp->cg_cgx).cs_nbfree--;
+       cylno = cbtocylno(fs, bno);
+       cg_blks(fs, cgp, cylno)[cbtorpos(fs, bno)]--;
+       cg_blktot(cgp)[cylno]--;

        fs->fs_fmod++;
        return (cgp->cg_cgx * fs->fs_fpg + bno);
 }
        fs->fs_fmod++;
        return (cgp->cg_cgx * fs->fs_fpg + bno);
 }

-       
-long
-ialloccg(dev, fs, cg, ipref, mode)
-       dev_t dev;
-       register struct fs *fs;
+
+/*
+ * Determine whether an inode can be allocated.
+ *
+ * Check to see if an inode is available, and if it is,
+ * allocate it using the following policy:
+ *   1) allocate the requested inode.
+ *   2) allocate the next available inode after the requested
+ *      inode in the specified cylinder group.
+ */
+ino_t
+ialloccg(ip, cg, ipref, mode)
+       struct inode *ip;

        int cg;
        daddr_t ipref;
        int mode;
 {
        int cg;
        daddr_t ipref;
        int mode;
 {

-       register struct buf *bp;
+       register struct fs *fs;

        register struct cg *cgp;
        register struct cg *cgp;

-       int i;
+       struct buf *bp;
+       int start, len, loc, map, i;

 
 

-       bp = bread(dev, cgtod(cg, fs), BSIZE);
-       if (bp->b_flags & B_ERROR)
-               return (0);
+       fs = ip->i_fs;
+       if (fs->fs_cs(fs, cg).cs_nifree == 0)
+               return (NULL);
+#ifdef SECSIZE
+       bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize,
+           fs->fs_dbsize);
+#else SECSIZE
+       bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize);
+#endif SECSIZE

        cgp = bp->b_un.b_cg;
        cgp = bp->b_un.b_cg;

-       if (cgp->cg_nifree == 0) {
+       if (bp->b_flags & B_ERROR || !cg_chkmagic(cgp) ||
+           cgp->cg_cs.cs_nifree == 0) {

                brelse(bp);
                brelse(bp);

-               return (0);
+               return (NULL);

        }
        }

+       cgp->cg_time = time.tv_sec;

        if (ipref) {
                ipref %= fs->fs_ipg;
        if (ipref) {
                ipref %= fs->fs_ipg;

-               if (isclr(cgp->cg_iused, ipref))
+               if (isclr(cg_inosused(cgp), ipref))

                        goto gotit;
                        goto gotit;

-       } else
-               ipref = cgp->cg_irotor;
-       for (i = 0; i < fs->fs_ipg; i++) {
-               ipref++;
-               if (ipref >= fs->fs_ipg)
-                       ipref = 0;
-               if (isclr(cgp->cg_iused, ipref)) {
+       }
+       start = cgp->cg_irotor / NBBY;
+       len = howmany(fs->fs_ipg - cgp->cg_irotor, NBBY);
+       loc = skpc(0xff, len, &cg_inosused(cgp)[start]);
+       if (loc == 0) {
+               len = start + 1;
+               start = 0;
+               loc = skpc(0xff, len, &cg_inosused(cgp)[0]);
+               if (loc == 0) {
+                       printf("cg = %s, irotor = %d, fs = %s\n",
+                           cg, cgp->cg_irotor, fs->fs_fsmnt);
+                       panic("ialloccg: map corrupted");
+                       /* NOTREACHED */
+               }
+       }
+       i = start + len - loc;
+       map = cg_inosused(cgp)[i];
+       ipref = i * NBBY;
+       for (i = 1; i < (1 << NBBY); i <<= 1, ipref++) {
+               if ((map & i) == 0) {

                        cgp->cg_irotor = ipref;
                        goto gotit;
                }
        }
                        cgp->cg_irotor = ipref;
                        goto gotit;
                }
        }

-       brelse(bp);
-       return (0);
+       printf("fs = %s\n", fs->fs_fsmnt);
+       panic("ialloccg: block not in map");
+       /* NOTREACHED */

 gotit:
 gotit:

-       setbit(cgp->cg_iused, ipref);
-       cgp->cg_nifree--;
-       fs->fs_nifree--;
-       fs->fs_cs(cg).cs_nifree--;
+       setbit(cg_inosused(cgp), ipref);
+       cgp->cg_cs.cs_nifree--;
+       fs->fs_cstotal.cs_nifree--;
+       fs->fs_cs(fs, cg).cs_nifree--;

        fs->fs_fmod++;
        if ((mode & IFMT) == IFDIR) {
        fs->fs_fmod++;
        if ((mode & IFMT) == IFDIR) {

-               cgp->cg_ndir++;
-               fs->fs_cs(cg).cs_ndir++;
+               cgp->cg_cs.cs_ndir++;
+               fs->fs_cstotal.cs_ndir++;
+               fs->fs_cs(fs, cg).cs_ndir++;

        }
        bdwrite(bp);
        return (cg * fs->fs_ipg + ipref);
 }
 
        }
        bdwrite(bp);
        return (cg * fs->fs_ipg + ipref);
 }
 

-fre(dev, bno, size)
-       dev_t dev;
+/*
+ * Free a block or fragment.
+ *
+ * The specified block or fragment is placed back in the
+ * free map. If a fragment is deallocated, a possible 
+ * block reassembly is checked.
+ */
+blkfree(ip, bno, size)
+       register struct inode *ip;

        daddr_t bno;
        daddr_t bno;

-       int size;
+       off_t size;

 {
        register struct fs *fs;
        register struct cg *cgp;
 {
        register struct fs *fs;
        register struct cg *cgp;


@@ -490,105 +860,147 @@ fre(dev, bno, size)
        int cg, blk, frags, bbase;
        register int i;
 
        int cg, blk, frags, bbase;
        register int i;
 

-       if ((unsigned)size > BSIZE || size % FSIZE != 0)
-               panic("free: bad size");
-       fs = getfs(dev);
-       cg = dtog(bno, fs);
-       if (badblock(fs, bno))
-               return;
-       bp = bread(dev, cgtod(cg, fs), BSIZE);
-       if (bp->b_flags & B_ERROR)
+       fs = ip->i_fs;
+       if ((unsigned)size > fs->fs_bsize || fragoff(fs, size) != 0) {
+               printf("dev = 0x%x, bsize = %d, size = %d, fs = %s\n",
+                   ip->i_dev, fs->fs_bsize, size, fs->fs_fsmnt);
+               panic("blkfree: bad size");
+       }
+       cg = dtog(fs, bno);
+       if (badblock(fs, bno)) {
+               printf("bad block %d, ino %d\n", bno, ip->i_number);

                return;
                return;

+       }
+#ifdef SECSIZE
+       bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize,
+           fs->fs_dbsize);
+#else SECSIZE
+       bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize);
+#endif SECSIZE

        cgp = bp->b_un.b_cg;
        cgp = bp->b_un.b_cg;

-       bno %= fs->fs_fpg;
-       if (size == BSIZE) {
-               if (isblock(cgp->cg_free, bno/FRAG))
-                       panic("free: freeing free block");
-               setblock(cgp->cg_free, bno/FRAG);
-               cgp->cg_nbfree++;
-               fs->fs_nbfree++;
-               fs->fs_cs(cg).cs_nbfree++;
-               i = bno * NSPF;
-               cgp->cg_b[i/fs->fs_spc][i%fs->fs_nsect*NRPOS/fs->fs_nsect]++;
+       if (bp->b_flags & B_ERROR || !cg_chkmagic(cgp)) {
+               brelse(bp);
+               return;
+       }
+       cgp->cg_time = time.tv_sec;
+       bno = dtogd(fs, bno);
+       if (size == fs->fs_bsize) {
+               if (isblock(fs, cg_blksfree(cgp), fragstoblks(fs, bno))) {
+                       printf("dev = 0x%x, block = %d, fs = %s\n",
+                           ip->i_dev, bno, fs->fs_fsmnt);
+                       panic("blkfree: freeing free block");
+               }
+               setblock(fs, cg_blksfree(cgp), fragstoblks(fs, bno));
+               cgp->cg_cs.cs_nbfree++;
+               fs->fs_cstotal.cs_nbfree++;
+               fs->fs_cs(fs, cg).cs_nbfree++;
+               i = cbtocylno(fs, bno);
+               cg_blks(fs, cgp, i)[cbtorpos(fs, bno)]++;
+               cg_blktot(cgp)[i]++;

        } else {
        } else {

-               bbase = bno - (bno % FRAG);
+               bbase = bno - fragnum(fs, bno);

                /*
                 * decrement the counts associated with the old frags
                 */
                /*
                 * decrement the counts associated with the old frags
                 */

-               blk = ((cgp->cg_free[bbase / NBBY] >> (bbase % NBBY)) &
-                      (0xff >> (NBBY - FRAG)));
-               fragacct(blk, cgp->cg_frsum, -1);
+               blk = blkmap(fs, cg_blksfree(cgp), bbase);
+               fragacct(fs, blk, cgp->cg_frsum, -1);

                /*
                 * deallocate the fragment
                 */
                /*
                 * deallocate the fragment
                 */

-               frags = size / FSIZE;
+               frags = numfrags(fs, size);

                for (i = 0; i < frags; i++) {
                for (i = 0; i < frags; i++) {

-                       if (isset(cgp->cg_free, bno + i))
-                               panic("free: freeing free frag");
-                       setbit(cgp->cg_free, bno + i);
-                       cgp->cg_nffree++;
-                       fs->fs_nffree++;
+                       if (isset(cg_blksfree(cgp), bno + i)) {
+                               printf("dev = 0x%x, block = %d, fs = %s\n",
+                                   ip->i_dev, bno + i, fs->fs_fsmnt);
+                               panic("blkfree: freeing free frag");
+                       }
+                       setbit(cg_blksfree(cgp), bno + i);

                }
                }

+               cgp->cg_cs.cs_nffree += i;
+               fs->fs_cstotal.cs_nffree += i;
+               fs->fs_cs(fs, cg).cs_nffree += i;

                /*
                 * add back in counts associated with the new frags
                 */
                /*
                 * add back in counts associated with the new frags
                 */

-               blk = ((cgp->cg_free[bbase / NBBY] >> (bbase % NBBY)) &
-                      (0xff >> (NBBY - FRAG)));
-               fragacct(blk, cgp->cg_frsum, 1);
+               blk = blkmap(fs, cg_blksfree(cgp), bbase);
+               fragacct(fs, blk, cgp->cg_frsum, 1);

                /*
                 * if a complete block has been reassembled, account for it
                 */
                /*
                 * if a complete block has been reassembled, account for it
                 */

-               if (isblock(cgp->cg_free, bbase / FRAG)) {
-                       cgp->cg_nffree -= FRAG;
-                       fs->fs_nffree -= FRAG;
-                       cgp->cg_nbfree++;
-                       fs->fs_nbfree++;
-                       fs->fs_cs(cg).cs_nbfree++;
-                       i = bbase * NSPF;
-                       cgp->cg_b[i / fs->fs_spc]
-                                [i % fs->fs_nsect * NRPOS / fs->fs_nsect]++;
+               if (isblock(fs, cg_blksfree(cgp), fragstoblks(fs, bbase))) {
+                       cgp->cg_cs.cs_nffree -= fs->fs_frag;
+                       fs->fs_cstotal.cs_nffree -= fs->fs_frag;
+                       fs->fs_cs(fs, cg).cs_nffree -= fs->fs_frag;
+                       cgp->cg_cs.cs_nbfree++;
+                       fs->fs_cstotal.cs_nbfree++;
+                       fs->fs_cs(fs, cg).cs_nbfree++;
+                       i = cbtocylno(fs, bbase);
+                       cg_blks(fs, cgp, i)[cbtorpos(fs, bbase)]++;
+                       cg_blktot(cgp)[i]++;

                }
        }
        fs->fs_fmod++;
        bdwrite(bp);
 }
 
                }
        }
        fs->fs_fmod++;
        bdwrite(bp);
 }
 

-ifree(dev, ino, mode)
-       dev_t dev;
+/*
+ * Free an inode.
+ *
+ * The specified inode is placed back in the free map.
+ */
+ifree(ip, ino, mode)
+       struct inode *ip;

        ino_t ino;
        int mode;
 {
        register struct fs *fs;
        register struct cg *cgp;
        register struct buf *bp;
        ino_t ino;
        int mode;
 {
        register struct fs *fs;
        register struct cg *cgp;
        register struct buf *bp;

-       int i;

        int cg;
 
        int cg;
 

-       fs = getfs(dev);
-       if ((unsigned)ino >= fs->fs_ipg*fs->fs_ncg)
+       fs = ip->i_fs;
+       if ((unsigned)ino >= fs->fs_ipg*fs->fs_ncg) {
+               printf("dev = 0x%x, ino = %d, fs = %s\n",
+                   ip->i_dev, ino, fs->fs_fsmnt);

                panic("ifree: range");
                panic("ifree: range");

-       cg = itog(ino, fs);
-       bp = bread(dev, cgtod(cg, fs), BSIZE);
-       if (bp->b_flags & B_ERROR)
-               return;
+       }
+       cg = itog(fs, ino);
+#ifdef SECSIZE
+       bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize,
+           fs->fs_dbsize);
+#else SECSIZE
+       bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), (int)fs->fs_cgsize);
+#endif SECSIZE

        cgp = bp->b_un.b_cg;
        cgp = bp->b_un.b_cg;

+       if (bp->b_flags & B_ERROR || !cg_chkmagic(cgp)) {
+               brelse(bp);
+               return;
+       }
+       cgp->cg_time = time.tv_sec;

        ino %= fs->fs_ipg;
        ino %= fs->fs_ipg;

-       if (isclr(cgp->cg_iused, ino))
+       if (isclr(cg_inosused(cgp), ino)) {
+               printf("dev = 0x%x, ino = %d, fs = %s\n",
+                   ip->i_dev, ino, fs->fs_fsmnt);

                panic("ifree: freeing free inode");
                panic("ifree: freeing free inode");

-       clrbit(cgp->cg_iused, ino);
-       cgp->cg_nifree++;
-       fs->fs_nifree++;
-       fs->fs_cs(cg).cs_nifree++;
+       }
+       clrbit(cg_inosused(cgp), ino);
+       if (ino < cgp->cg_irotor)
+               cgp->cg_irotor = ino;
+       cgp->cg_cs.cs_nifree++;
+       fs->fs_cstotal.cs_nifree++;
+       fs->fs_cs(fs, cg).cs_nifree++;

        if ((mode & IFMT) == IFDIR) {
        if ((mode & IFMT) == IFDIR) {

-               cgp->cg_ndir--;
-               fs->fs_cs(cg).cs_ndir--;
+               cgp->cg_cs.cs_ndir--;
+               fs->fs_cstotal.cs_ndir--;
+               fs->fs_cs(fs, cg).cs_ndir--;

        }
        fs->fs_fmod++;
        bdwrite(bp);
 }
 
 /*
        }
        fs->fs_fmod++;
        bdwrite(bp);
 }
 
 /*

- * find a block of the specified size in the specified cylinder group
+ * Find a block of the specified size in the specified cylinder group.
+ *

  * It is a panic if a request is made to find a block if none are
  * available.
  */
  * It is a panic if a request is made to find a block if none are
  * available.
  */


@@ -608,20 +1020,24 @@ mapsearch(fs, cgp, bpref, allocsiz)
         * map for an appropriate bit pattern
         */
        if (bpref)
         * map for an appropriate bit pattern
         */
        if (bpref)

-               start = bpref % fs->fs_fpg / NBBY;
+               start = dtogd(fs, bpref) / NBBY;

        else
                start = cgp->cg_frotor / NBBY;
        else
                start = cgp->cg_frotor / NBBY;

-       len = roundup(fs->fs_fpg - 1, NBBY) / NBBY - start;
-       loc = scanc(len, &cgp->cg_free[start], fragtbl, 1 << (allocsiz - 1));
+       len = howmany(fs->fs_fpg, NBBY) - start;
+       loc = scanc((unsigned)len, (caddr_t)&cg_blksfree(cgp)[start],
+               (caddr_t)fragtbl[fs->fs_frag],
+               (int)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY))));

        if (loc == 0) {
        if (loc == 0) {

-               len = start - 1;
-               start = (cgdmin(cgp->cg_cgx, fs) -
-                        cgbase(cgp->cg_cgx, fs)) / NBBY;
-               loc = scanc(len, &cgp->cg_free[start], fragtbl,
-                       1 << (allocsiz - 1));
+               len = start + 1;
+               start = 0;
+               loc = scanc((unsigned)len, (caddr_t)&cg_blksfree(cgp)[0],
+                       (caddr_t)fragtbl[fs->fs_frag],
+                       (int)(1 << (allocsiz - 1 + (fs->fs_frag % NBBY))));

                if (loc == 0) {
                if (loc == 0) {

+                       printf("start = %d, len = %d, fs = %s\n",
+                           start, len, fs->fs_fsmnt);

                        panic("alloccg: map corrupted");
                        panic("alloccg: map corrupted");

-                       return (0);
+                       /* NOTREACHED */

                }
        }
        bno = (start + len - loc) * NBBY;
                }
        }
        bno = (start + len - loc) * NBBY;


@@ -630,101 +1046,27 @@ mapsearch(fs, cgp, bpref, allocsiz)
         * found the byte in the map
         * sift through the bits to find the selected frag
         */
         * found the byte in the map
         * sift through the bits to find the selected frag
         */

-       for (i = 0; i < NBBY; i += FRAG) {
-               blk = (cgp->cg_free[bno / NBBY] >> i) & (0xff >> NBBY - FRAG);
+       for (i = bno + NBBY; bno < i; bno += fs->fs_frag) {
+               blk = blkmap(fs, cg_blksfree(cgp), bno);

                blk <<= 1;
                field = around[allocsiz];
                subfield = inside[allocsiz];
                blk <<= 1;
                field = around[allocsiz];
                subfield = inside[allocsiz];

-               for (pos = 0; pos <= FRAG - allocsiz; pos++) {
-                       if ((blk & field) == subfield) {
-                               return (bno + i + pos);
-                       }
+               for (pos = 0; pos <= fs->fs_frag - allocsiz; pos++) {
+                       if ((blk & field) == subfield)
+                               return (bno + pos);

                        field <<= 1;
                        subfield <<= 1;
                }
        }
                        field <<= 1;
                        subfield <<= 1;
                }
        }

+       printf("bno = %d, fs = %s\n", bno, fs->fs_fsmnt);

        panic("alloccg: block not in map");
        panic("alloccg: block not in map");

-       return (0);
-}
-
-/*
- * update the frsum fields to reflect addition or deletion 
- * of some frags
- */
-fragacct(fragmap, fraglist, cnt)
-       int fragmap;
-       short fraglist[];
-       int cnt;
-{
-       int inblk;
-       register int field, subfield;
-       register int siz, pos;
-
-       inblk = (int)(fragtbl[fragmap]) << 1;
-       fragmap <<= 1;
-       for (siz = 1; siz < FRAG; siz++) {
-               if (((1 << siz) & inblk) == 0)
-                       continue;
-               field = around[siz];
-               subfield = inside[siz];
-               for (pos = siz; pos <= FRAG; pos++) {
-                       if ((fragmap & field) == subfield) {
-                               fraglist[siz] += cnt;
-                               pos += siz;
-                               field <<= siz;
-                               subfield <<= siz;
-                       }
-                       field <<= 1;
-                       subfield <<= 1;
-               }
-       }
-}
-
-badblock(fs, bn)
-       register struct fs *fs;
-       daddr_t bn;
-{
-
-       if ((unsigned)bn >= fs->fs_size || bn < cgdmin(dtog(bn, fs), fs)) {
-               fserr(fs, "bad block");
-               return (1);
-       }
-       return (0);
-}
-
-/*
- * getfs maps a device number into
- * a pointer to the incore super
- * block.  The algorithm is a linear
- * search through the mount table.
- * A consistency check of the
- * in core free-block and i-node
- * counts is performed.
- *
- * panic: no fs -- the device is not mounted.
- *     this "cannot happen"
- */
-struct fs *
-getfs(dev)
-       dev_t dev;
-{
-       register struct mount *mp;
-       register struct fs *fs;
-
-       for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++)
-               if (mp->m_bufp != NULL && mp->m_dev == dev) {
-                       fs = mp->m_bufp->b_un.b_fs;
-                       if (fs->fs_magic != FS_MAGIC)
-                               panic("getfs: bad magic");
-                       return (fs);
-               }
-       panic("getfs: no fs");
-       return (NULL);
+       return (-1);

 }
 
 /*
 }
 
 /*

- * Fserr prints the name of a file system
- * with an error diagnostic, in the form
+ * Fserr prints the name of a file system with an error diagnostic.
+ * 
+ * The form of the error message is:

  *     fs: error message
  */
 fserr(fs, cp)
  *     fs: error message
  */
 fserr(fs, cp)


@@ -732,92 +1074,5 @@ fserr(fs, cp)
        char *cp;
 {
 
        char *cp;
 {
 

-       printf("%s: %s\n", fs->fs_fsmnt, cp);
-}
-
-/*
- * Getfsx returns the index in the file system
- * table of the specified device.  The swap device
- * is also assigned a pseudo-index.  The index may
- * be used as a compressed indication of the location
- * of a block, recording
- *     <getfsx(dev),blkno>
- * rather than
- *     <dev, blkno>
- * provided the information need remain valid only
- * as long as the file system is mounted.
- */
-getfsx(dev)
-       dev_t dev;
-{
-       register struct mount *mp;
-
-       if (dev == swapdev)
-               return (MSWAPX);
-       for(mp = &mount[0]; mp < &mount[NMOUNT]; mp++)
-               if (mp->m_dev == dev)
-                       return (mp - &mount[0]);
-       return (-1);
-}
-
-/*
- * Update is the internal name of 'sync'.  It goes through the disk
- * queues to initiate sandbagged IO; goes through the inodes to write
- * modified nodes; and it goes through the mount table to initiate modified
- * super blocks.
- */
-update()
-{
-       register struct inode *ip;
-       register struct mount *mp;
-       register struct buf *bp;
-       struct fs *fs;
-       time_t tim;
-       int i, blks;
-
-       if (updlock)
-               return;
-       updlock++;
-       /*
-        * Write back modified superblocks.
-        * Consistency check that the superblock
-        * of each file system is still in the buffer cache.
-        */
-       for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++)
-               if (mp->m_bufp != NULL) {
-                       fs = mp->m_bufp->b_un.b_fs;
-                       if (fs->fs_fmod == 0)
-                               continue;
-                       if (fs->fs_ronly != 0)
-                               panic("update: rofs mod");
-                       bp = getblk(mp->m_dev, SBLOCK, BSIZE);
-                       fs->fs_fmod = 0;
-                       fs->fs_time = TIME;
-                       if (bp->b_un.b_fs != fs)
-                               panic("update: bad b_fs");
-                       bwrite(bp);
-                       blks = howmany(cssize(fs), BSIZE);
-                       for (i = 0; i < blks; i++) {
-                               bp = getblk(mp->m_dev, csaddr(fs) + (i * FRAG),
-                                       BSIZE);
-                               bwrite(bp);
-                       }
-               }
-       /*
-        * Write back each (modified) inode.
-        */
-       for (ip = inode; ip < inodeNINODE; ip++)
-               if((ip->i_flag&ILOCK)==0 && ip->i_count) {
-                       ip->i_flag |= ILOCK;
-                       ip->i_count++;
-                       tim = TIME;
-                       iupdat(ip, &tim, &tim, 0);
-                       iput(ip);
-               }
-       updlock = 0;
-       /*
-        * Force stale buffer cache information to be flushed,
-        * for all devices.
-        */
-       bflush(NODEV);
+       log(LOG_ERR, "%s: %s\n", fs->fs_fsmnt, cp);

 }
 }