X-Git-Url: https://git.subgeniuskitty.com/unix-history/.git/blobdiff_plain/1b940b136c1f0b68a0a6b5af6072ec374ccb3856..cd66a61f882c077209c4edbf2071544d313d5b7a:/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 b8ad5f8733..d8a029ec69 100644
--- a/usr/src/sys/ufs/lfs/lfs_alloc.c
+++ b/usr/src/sys/ufs/lfs/lfs_alloc.c
@@ -1,6 +1,6 @@
 /* Copyright (c) 1981 Regents of the University of California */
 
-static char vers[] = "@(#)lfs_alloc.c 1.5 %G%";
+static char vers[] = "@(#)lfs_alloc.c 2.1 %G%";
 
 /*	alloc.c	4.8	81/03/08	*/
 
@@ -11,16 +11,40 @@ static char vers[] = "@(#)lfs_alloc.c 1.5 %G%";
 #include "../h/conf.h"
 #include "../h/buf.h"
 #include "../h/inode.h"
-#include "../h/dir.h"
+#include "../h/ndir.h"
 #include "../h/user.h"
 
-long	hashalloc();
-long	alloccg();
-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 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 *
-alloc(dev, ip, bpref, size)
-	dev_t dev;
+alloc(ip, bpref, size)
 	register struct inode *ip;
 	daddr_t bpref;
 	int size;
@@ -30,19 +54,25 @@ alloc(dev, ip, bpref, size)
 	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)
 		panic("alloc: bad size");
-	fs = getfs(dev);
-	if (fs->fs_nbfree == 0 && size == BSIZE)
+	if (size == fs->fs_bsize && fs->fs_cstotal.cs_nbfree == 0)
 		goto nospace;
+	if (u.u_uid != 0 &&
+	    fs->fs_cstotal.cs_nbfree * fs->fs_frag + fs->fs_cstotal.cs_nffree <
+	      fs->fs_dsize * fs->fs_minfree / 100)
+		goto nospace;
+	if (bpref >= fs->fs_size)
+		bpref = 0;
 	if (bpref == 0)
-		cg = itog(ip->i_number, fs);
+		cg = itog(fs, ip->i_number);
 	else
-		cg = dtog(bpref, fs);
-	bno = hashalloc(dev, fs, cg, (long)bpref, size, alloccg);
+		cg = dtog(fs, bpref);
+	bno = (daddr_t)hashalloc(ip, cg, (long)bpref, size, alloccg);
 	if (bno == 0)
 		goto nospace;
-	bp = getblk(dev, bno, size);
+	bp = getblk(ip->i_dev, fsbtodb(fs, bno), size);
 	clrbuf(bp);
 	return (bp);
 nospace:
@@ -52,11 +82,18 @@ nospace:
 	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 *
-realloccg(dev, ip, bprev, osize, nsize)
-	dev_t dev;
+realloccg(ip, bprev, bpref, osize, nsize)
 	register struct inode *ip;
-	daddr_t bprev;
+	daddr_t bprev, bpref;
 	int osize, nsize;
 {
 	daddr_t bno;
@@ -65,37 +102,51 @@ realloccg(dev, ip, bprev, osize, nsize)
 	caddr_t cp;
 	int cg;
 	
-	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)
 		panic("realloccg: bad size");
-	fs = getfs(dev);
+	if (u.u_uid != 0 &&
+	    fs->fs_cstotal.cs_nbfree * fs->fs_frag + fs->fs_cstotal.cs_nffree <
+	      fs->fs_dsize * fs->fs_minfree / 100)
+		goto nospace;
 	if (bprev == 0)
 		panic("realloccg: bad bprev");
-	else
-		cg = dtog(bprev, fs);
-	bno = fragextend(dev, fs, cg, (long)bprev, osize, nsize);
+	cg = dtog(fs, bprev);
+	bno = fragextend(ip, cg, (long)bprev, osize, nsize);
 	if (bno != 0) {
-		bp = bread(dev, bno, osize);
+		bp = bread(ip->i_dev, fsbtodb(fs, bno), osize);
+		if (bp->b_flags & B_ERROR) {
+			brelse(bp);
+			return (NULL);
+		}
 		bp->b_bcount = nsize;
 		blkclr(bp->b_un.b_addr + osize, nsize - osize);
 		return (bp);
 	}
-	bno = hashalloc(dev, fs, cg, (long)bprev, nsize, alloccg);
+	if (bpref >= fs->fs_size)
+		bpref = 0;
+	bno = (daddr_t)hashalloc(ip, cg, (long)bpref, nsize, alloccg);
 	if (bno != 0) {
 		/*
 		 * make a new copy
 		 */
-		obp = bread(dev, bprev, osize);
-		bp = getblk(dev, bno, nsize);
+		obp = bread(ip->i_dev, fsbtodb(fs, bprev), osize);
+		if (obp->b_flags & B_ERROR) {
+			brelse(obp);
+			return (NULL);
+		}
+		bp = getblk(ip->i_dev, fsbtodb(fs, 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;
 		brelse(obp);
-		fre(dev, bprev, osize);
+		fre(ip, bprev, (off_t)osize);
 		blkclr(bp->b_un.b_addr + osize, nsize - osize);
-		return(bp);
+		return (bp);
 	}
+nospace:
 	/*
 	 * no space available
 	 */
@@ -105,27 +156,44 @@ realloccg(dev, ip, bprev, osize, nsize)
 	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 *
-ialloc(dev, ipref, mode)
-	dev_t dev;
+ialloc(pip, ipref, mode)
+	register struct inode *pip;
 	ino_t ipref;
 	int mode;
 {
-	daddr_t ino;
+	ino_t ino;
 	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;
-	cg = itog(ipref, fs);
-	ino = hashalloc(dev, fs, cg, (long)ipref, mode, ialloccg);
+	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;
-	ip = iget(dev, ino);
+	ip = iget(pip->i_dev, pip->i_fs, ino);
 	if (ip == NULL) {
-		ifree(dev, ino);
+		ifree(ip, ino, 0);
 		return (NULL);
 	}
 	if (ip->i_mode)
@@ -133,46 +201,89 @@ ialloc(dev, ipref, mode)
 	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);
 }
 
-dipref(dev)
-	dev_t dev;
-{
+/*
+ * 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(fs)
 	register struct fs *fs;
-	int cg, minndir, mincg;
+{
+	int cg, minndir, mincg, avgifree;
 
-	fs = getfs(dev);
-	minndir = fs->fs_cs[0].cs_ndir;
+	avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg;
+	minndir = fs->fs_ipg;
 	mincg = 0;
-	for (cg = 1; cg < fs->fs_ncg; cg++)
-		if (fs->fs_cs[cg].cs_ndir < minndir) {
+	for (cg = 0; cg < fs->fs_ncg; cg++)
+		if (fs->fs_cs(fs, cg).cs_ndir < minndir &&
+		    fs->fs_cs(fs, cg).cs_nifree >= avgifree) {
 			mincg = cg;
-			minndir = fs->fs_cs[cg].cs_ndir;
-			if (minndir == 0)
-				break;
+			minndir = fs->fs_cs(fs, cg).cs_ndir;
 		}
 	return (fs->fs_ipg * mincg);
 }
 
-long
-hashalloc(dev, fs, cg, pref, size, allocator)
-	dev_t dev;
+/*
+ * Select a cylinder to place a large block of data.
+ *
+ * The policy implemented by this algorithm is to maintain a
+ * rotor that sweeps the cylinder groups. When a block is 
+ * needed, the rotor is advanced until a cylinder group with
+ * greater than the average number of free blocks is found.
+ */
+daddr_t
+blkpref(fs)
 	register struct fs *fs;
+{
+	int cg, avgbfree;
+
+	avgbfree = fs->fs_cstotal.cs_nbfree / fs->fs_ncg;
+	for (cg = fs->fs_cgrotor + 1; 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 <= fs->fs_cgrotor; cg++)
+		if (fs->fs_cs(fs, cg).cs_nbfree >= avgbfree) {
+			fs->fs_cgrotor = cg;
+			return (fs->fs_fpg * cg + fs->fs_frag);
+		}
+	return (NULL);
+}
+
+/*
+ * 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 */
-	long (*allocator)();
+	u_long (*allocator)();
 {
+	register struct fs *fs;
 	long result;
 	int i, icg = cg;
 
+	fs = ip->i_fs;
 	/*
 	 * 1: preferred cylinder group
 	 */
-	result = (*allocator)(dev, fs, cg, pref, size);
+	result = (*allocator)(ip, cg, pref, size);
 	if (result)
 		return (result);
 	/*
@@ -182,7 +293,7 @@ hashalloc(dev, fs, cg, pref, size, allocator)
 		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);
 	}
@@ -191,97 +302,109 @@ hashalloc(dev, fs, cg, pref, size, allocator)
 	 */
 	cg = icg;
 	for (i = 0; i < fs->fs_ncg; i++) {
-		result = (*allocator)(dev, fs, cg, 0, size);
+		result = (*allocator)(ip, cg, 0, size);
 		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
-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;
 {
+	register struct fs *fs;
 	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) {
+	fs = ip->i_fs;
+	frags = numfrags(fs, nsize);
+	bbase = fragoff(fs, bprev);
+	if (bbase > (bprev + frags - 1) % fs->fs_frag) {
 		/* cannot extend across a block boundry */
-		return (0);
+		return (NULL);
+	}
+	bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), fs->fs_bsize);
+	if (bp->b_flags & B_ERROR) {
+		brelse(bp);
+		return (NULL);
 	}
-	bp = bread(dev, cgtod(cg, fs), BSIZE);
-	if (bp->b_flags & B_ERROR)
-		return (0);
 	cgp = bp->b_un.b_cg;
-	bno = bprev % fs->fs_fpg;
-	for (i = osize / FSIZE; i < frags; i++) {
+	bno = dtogd(fs, bprev);
+	for (i = numfrags(fs, osize); i < frags; i++)
+		if (isclr(cgp->cg_free, bno + i)) {
+			brelse(bp);
+			return (NULL);
+		}
+	/*
+	 * 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(cgp->cg_free, 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(cgp->cg_free, bno + i);
+		cgp->cg_cs.cs_nffree--;
+		fs->fs_cstotal.cs_nffree--;
+		fs->fs_cs(fs, cg).cs_nffree--;
 	}
-	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--;
-		}
-		fs->fs_fmod++;
-		bdwrite(bp);
-		return (bprev);
-	}
-	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
-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;
 {
+	register struct fs *fs;
 	register struct buf *bp;
 	register struct cg *cgp;
 	int bno, frags;
 	int allocsiz;
-	int start, len, loc;
-	int blk, field, subfield, pos;
 	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);
+	bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), fs->fs_bsize);
+	if (bp->b_flags & B_ERROR) {
+		brelse(bp);
+		return (NULL);
+	}
 	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 (size == fs->fs_bsize) {
+		bno = alloccgblk(fs, cgp, bpref);
 		bdwrite(bp);
 		return (bno);
 	}
@@ -290,74 +413,39 @@ alloccg(dev, fs, cg, bpref, size)
 	 * 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 (allocsiz == FRAG) {
+	if (allocsiz == fs->fs_frag) {
 		/*
 		 * 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);
-			return (0);
+			return (NULL);
 		}
-		bno = alloccgblk(dev, fs, cgp, bpref);
-		bpref = bno % fs->fs_fpg;
-		for (i = frags; i < FRAG; i++)
+		bno = alloccgblk(fs, cgp, bpref);
+		bpref = dtogd(fs, bno);
+		for (i = frags; i < fs->fs_frag; i++)
 			setbit(cgp->cg_free, bpref + i);
-		i = FRAG - frags;
-		cgp->cg_nffree += i;
-		fs->fs_nffree += 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;
 		cgp->cg_frsum[i]++;
 		bdwrite(bp);
 		return (bno);
 	}
-	/*
-	 * find the fragment by searching through the free block
-	 * map for an appropriate bit pattern
-	 */
-	if (bpref)
-		start = bpref % fs->fs_fpg / 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));
-	if (loc == 0) {
-		len = start - 1;
-		start = (cgdmin(cg, fs) - cgbase(cg, fs)) / NBBY;
-		loc = scanc(len, &cgp->cg_free[start], fragtbl,
-			1 << (allocsiz - 1));
-		if (loc == 0)
-			panic("alloccg: can't find frag");
-	}
-	bno = (start + len - loc) * NBBY;
-	cgp->cg_frotor = bno;
-	/*
-	 * 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);
-		blk <<= 1;
-		field = around[allocsiz];
-		subfield = inside[allocsiz];
-		for (pos = 0; pos <= FRAG - allocsiz; pos++) {
-			if ((blk & field) == subfield) {
-				bno += i + pos;
-				goto gotit;
-			}
-			field <<= 1;
-			subfield <<= 1;
-		}
-	}
-	panic("alloccg: frag not in block");
-gotit:
+	bno = mapsearch(fs, cgp, bpref, allocsiz);
+	if (bno == 0)
+		return (NULL);
 	for (i = 0; i < frags; i++)
 		clrbit(cgp->cg_free, bno + i);
-	cgp->cg_nffree -= frags;
-	fs->fs_nffree -= frags;
+	cgp->cg_cs.cs_nffree -= frags;
+	fs->fs_cstotal.cs_nffree -= frags;
+	fs->fs_cs(fs, cg).cs_nffree -= frags;
 	cgp->cg_frsum[allocsiz]--;
 	if (frags != allocsiz)
 		cgp->cg_frsum[allocsiz - frags]++;
@@ -365,64 +453,157 @@ gotit:
 	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
-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;
+	int cylno, pos, delta;
+	short *cylbp;
 	register int i;
 
-	if (bpref) {
-		bpref &= ~(FRAG - 1);
-		bpref %= fs->fs_fpg;
-		if (isblock(cgp->cg_free, bpref/FRAG))
-			goto gotit;
-	} else
+	if (bpref == 0) {
 		bpref = cgp->cg_rotor;
-	for (i = 0; i < cgp->cg_ndblk; i += FRAG) {
-		bpref += FRAG;
-		if (bpref >= cgp->cg_ndblk)
-			bpref = 0;
-		if (isblock(cgp->cg_free, bpref/FRAG)) {
-			cgp->cg_rotor = bpref;
-			goto gotit;
+		goto norot;
+	}
+	bpref &= ~(fs->fs_frag - 1);
+	bpref = dtogd(fs, bpref);
+	/*
+	 * if the requested block is available, use it
+	 */
+	if (isblock(fs, cgp->cg_free, bpref/fs->fs_frag)) {
+		bno = bpref;
+		goto gotit;
+	}
+	/*
+	 * check for a block available on the same cylinder
+	 */
+	cylno = cbtocylno(fs, bpref);
+	if (cgp->cg_btot[cylno] == 0)
+		goto norot;
+	if (fs->fs_cpc == 0) {
+		/*
+		 * block layout info is not available, so just have
+		 * to take any block in this cylinder.
+		 */
+		bpref = howmany(fs->fs_spc * cylno, NSPF(fs));
+		goto norot;
+	}
+	/*
+	 * find a block that is rotationally optimal
+	 */
+	cylbp = cgp->cg_b[cylno];
+	if (fs->fs_rotdelay == 0) {
+		pos = cbtorpos(fs, bpref);
+	} else {
+		/*
+		 * 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 rotational position
+		 */
+		bpref += fs->fs_rotdelay * fs->fs_rps * fs->fs_nsect /
+		    (NSPF(fs) * 1000);
+		pos = cbtorpos(fs, bpref);
+		pos = (pos + 1) % NRPOS;
+	}
+	/*
+	 * check the summary information to see if a block is 
+	 * available in the requested cylinder starting at the
+	 * optimal rotational position and proceeding around.
+	 */
+	for (i = pos; i < NRPOS; i++)
+		if (cylbp[i] > 0)
+			break;
+	if (i == NRPOS)
+		for (i = 0; i < pos; i++)
+			if (cylbp[i] > 0)
+				break;
+	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->fs_postbl[pos][i] == -1)
+			panic("alloccgblk: cyl groups corrupted");
+		for (i = fs->fs_postbl[pos][i];; ) {
+			if (isblock(fs, cgp->cg_free, bno + i)) {
+				bno = (bno + i) * fs->fs_frag;
+				goto gotit;
+			}
+			delta = fs->fs_rotbl[i];
+			if (delta <= 0 || delta > MAXBPC - i)
+				break;
+			i += delta;
 		}
+		panic("alloccgblk: can't find blk in cyl");
 	}
-	panic("alloccgblk: can't find a blk");
-	return (0);
+norot:
+	/*
+	 * no blocks in the requested cylinder, so take next
+	 * available one in this cylinder group.
+	 */
+	bno = mapsearch(fs, cgp, bpref, fs->fs_frag);
+	if (bno == 0)
+		return (NULL);
+	cgp->cg_rotor = bno;
 gotit:
-	clrblock(cgp->cg_free, bpref/FRAG);
-	cgp->cg_nbfree--;
-	fs->fs_nbfree--;
-	fs->fs_cs[cgp->cg_cgx].cs_nbfree--;
-	i = bpref * NSPF;
-	cgp->cg_b[i/fs->fs_spc][i%fs->fs_nsect*NRPOS/fs->fs_nsect]--;
+	clrblock(fs, cgp->cg_free, bno/fs->fs_frag);
+	cgp->cg_cs.cs_nbfree--;
+	fs->fs_cstotal.cs_nbfree--;
+	fs->fs_cs(fs, cgp->cg_cgx).cs_nbfree--;
+	cylno = cbtocylno(fs, bno);
+	cgp->cg_b[cylno][cbtorpos(fs, bno)]--;
+	cgp->cg_btot[cylno]--;
 	fs->fs_fmod++;
-	return (cgp->cg_cgx * fs->fs_fpg + bpref);
+	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;
 {
+	register struct fs *fs;
 	register struct buf *bp;
 	register struct cg *cgp;
 	int i;
 
-	bp = bread(dev, cgtod(cg, fs), BSIZE);
-	if (bp->b_flags & B_ERROR)
-		return (0);
-	cgp = bp->b_un.b_cg;
-	if (cgp->cg_nifree == 0) {
+	fs = ip->i_fs;
+	if (fs->fs_cs(fs, cg).cs_nifree == 0)
+		return (NULL);
+	bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), fs->fs_bsize);
+	if (bp->b_flags & B_ERROR) {
 		brelse(bp);
-		return (0);
+		return (NULL);
 	}
+	cgp = bp->b_un.b_cg;
 	if (ipref) {
 		ipref %= fs->fs_ipg;
 		if (isclr(cgp->cg_iused, ipref))
@@ -439,25 +620,33 @@ ialloccg(dev, fs, cg, ipref, mode)
 		}
 	}
 	brelse(bp);
-	return (0);
+	return (NULL);
 gotit:
 	setbit(cgp->cg_iused, ipref);
-	cgp->cg_nifree--;
-	fs->fs_nifree--;
-	fs->fs_cs[cg].cs_nifree--;
+	cgp->cg_cs.cs_nifree--;
+	fs->fs_cstotal.cs_nifree--;
+	fs->fs_cs(fs, cg).cs_nifree--;
 	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);
 }
 
-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.
+ */
+fre(ip, bno, size)
+	register struct inode *ip;
 	daddr_t bno;
-	int size;
+	off_t size;
 {
 	register struct fs *fs;
 	register struct cg *cgp;
@@ -465,124 +654,196 @@ fre(dev, bno, size)
 	int cg, blk, frags, bbase;
 	register int i;
 
-	if ((unsigned)size > BSIZE || size % FSIZE != 0)
+	fs = ip->i_fs;
+	if ((unsigned)size > fs->fs_bsize || fragoff(fs, size) != 0)
 		panic("free: bad size");
-	fs = getfs(dev);
-	cg = dtog(bno, fs);
+	cg = dtog(fs, bno);
 	if (badblock(fs, bno))
 		return;
-	bp = bread(dev, cgtod(cg, fs), BSIZE);
-	if (bp->b_flags & B_ERROR)
+	bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), fs->fs_bsize);
+	if (bp->b_flags & B_ERROR) {
+		brelse(bp);
 		return;
+	}
 	cgp = bp->b_un.b_cg;
-	bno %= fs->fs_fpg;
-	if (size == BSIZE) {
-		if (isblock(cgp->cg_free, bno/FRAG))
+	bno = dtogd(fs, bno);
+	if (size == fs->fs_bsize) {
+		if (isblock(fs, cgp->cg_free, bno/fs->fs_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]++;
+		setblock(fs, cgp->cg_free, bno/fs->fs_frag);
+		cgp->cg_cs.cs_nbfree++;
+		fs->fs_cstotal.cs_nbfree++;
+		fs->fs_cs(fs, cg).cs_nbfree++;
+		i = cbtocylno(fs, bno);
+		cgp->cg_b[i][cbtorpos(fs, bno)]++;
+		cgp->cg_btot[i]++;
 	} else {
-		bbase = bno - (bno % FRAG);
+		bbase = bno - (bno % fs->fs_frag);
 		/*
 		 * 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, cgp->cg_free, bbase);
+		fragacct(fs, blk, cgp->cg_frsum, -1);
 		/*
 		 * deallocate the fragment
 		 */
-		frags = size / FSIZE;
+		frags = numfrags(fs, size);
 		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++;
 		}
+		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
 		 */
-		blk = ((cgp->cg_free[bbase / NBBY] >> (bbase % NBBY)) &
-		       (0xff >> (NBBY - FRAG)));
-		fragacct(blk, cgp->cg_frsum, 1);
+		blk = blkmap(fs, cgp->cg_free, bbase);
+		fragacct(fs, blk, cgp->cg_frsum, 1);
 		/*
 		 * 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, cgp->cg_free, bbase / fs->fs_frag)) {
+			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);
+			cgp->cg_b[i][cbtorpos(fs, bbase)]++;
+			cgp->cg_btot[i]++;
 		}
 	}
 	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;
-	int i;
 	int cg;
 
-	fs = getfs(dev);
+	fs = ip->i_fs;
 	if ((unsigned)ino >= fs->fs_ipg*fs->fs_ncg)
 		panic("ifree: range");
-	cg = itog(ino, fs);
-	bp = bread(dev, cgtod(cg, fs), BSIZE);
-	if (bp->b_flags & B_ERROR)
+	cg = itog(fs, ino);
+	bp = bread(ip->i_dev, fsbtodb(fs, cgtod(fs, cg)), fs->fs_bsize);
+	if (bp->b_flags & B_ERROR) {
+		brelse(bp);
 		return;
+	}
 	cgp = bp->b_un.b_cg;
 	ino %= fs->fs_ipg;
 	if (isclr(cgp->cg_iused, ino))
 		panic("ifree: freeing free inode");
 	clrbit(cgp->cg_iused, ino);
-	cgp->cg_nifree++;
-	fs->fs_nifree++;
-	fs->fs_cs[cg].cs_nifree++;
+	cgp->cg_cs.cs_nifree++;
+	fs->fs_cstotal.cs_nifree++;
+	fs->fs_cs(fs, cg).cs_nifree++;
 	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);
 }
 
 /*
- * update the frsum fields to reflect addition or deletion 
- * of some frags
+ * 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.
  */
-fragacct(fragmap, fraglist, cnt)
+daddr_t
+mapsearch(fs, cgp, bpref, allocsiz)
+	register struct fs *fs;
+	register struct cg *cgp;
+	daddr_t bpref;
+	int allocsiz;
+{
+	daddr_t bno;
+	int start, len, loc, i;
+	int blk, field, subfield, pos;
+
+	/*
+	 * find the fragment by searching through the free block
+	 * map for an appropriate bit pattern
+	 */
+	if (bpref)
+		start = dtogd(fs, bpref) / NBBY;
+	else
+		start = cgp->cg_frotor / NBBY;
+	len = howmany(fs->fs_fpg, NBBY) - start;
+	loc = scanc(len, &cgp->cg_free[start], fragtbl[fs->fs_frag],
+		1 << (allocsiz - 1 + (fs->fs_frag % NBBY)));
+	if (loc == 0) {
+		loc = fs->fs_dblkno / NBBY;
+		len = start - loc + 1;
+		start = loc;
+		loc = scanc(len, &cgp->cg_free[start], fragtbl[fs->fs_frag],
+			1 << (allocsiz - 1 + (fs->fs_frag % NBBY)));
+		if (loc == 0) {
+			panic("alloccg: map corrupted");
+			return (NULL);
+		}
+	}
+	bno = (start + len - loc) * NBBY;
+	cgp->cg_frotor = bno;
+	/*
+	 * found the byte in the map
+	 * sift through the bits to find the selected frag
+	 */
+	for (i = bno + NBBY; bno < i; bno += fs->fs_frag) {
+		blk = blkmap(fs, cgp->cg_free, bno);
+		blk <<= 1;
+		field = around[allocsiz];
+		subfield = inside[allocsiz];
+		for (pos = 0; pos <= fs->fs_frag - allocsiz; pos++) {
+			if ((blk & field) == subfield)
+				return (bno + pos);
+			field <<= 1;
+			subfield <<= 1;
+		}
+	}
+	panic("alloccg: block not in map");
+	return (NULL);
+}
+
+/*
+ * Update the frsum fields to reflect addition or deletion 
+ * of some frags.
+ */
+fragacct(fs, fragmap, fraglist, cnt)
+	struct fs *fs;
 	int fragmap;
-	short fraglist[];
+	long fraglist[];
 	int cnt;
 {
 	int inblk;
 	register int field, subfield;
 	register int siz, pos;
 
-	inblk = (int)(fragtbl[fragmap]) << 1;
+	inblk = (int)(fragtbl[fs->fs_frag][fragmap]) << 1;
 	fragmap <<= 1;
-	for (siz = 1; siz < FRAG; siz++) {
-		if (((1 << siz) & inblk) == 0)
+	for (siz = 1; siz < fs->fs_frag; siz++) {
+		if ((inblk & (1 << (siz + (fs->fs_frag % NBBY)))) == 0)
 			continue;
 		field = around[siz];
 		subfield = inside[siz];
-		for (pos = siz; pos <= FRAG; pos++) {
+		for (pos = siz; pos <= fs->fs_frag; pos++) {
 			if ((fragmap & field) == subfield) {
 				fraglist[siz] += cnt;
 				pos += siz;
@@ -595,12 +856,15 @@ fragacct(fragmap, fraglist, cnt)
 	}
 }
 
+/*
+ * Check that a specified block number is in range.
+ */
 badblock(fs, bn)
 	register struct fs *fs;
 	daddr_t bn;
 {
 
-	if ((unsigned)bn >= fs->fs_size || bn < cgdmin(dtog(bn, fs), fs)) {
+	if ((unsigned)bn >= fs->fs_size || bn < cgdmin(fs, dtog(fs, bn))) {
 		fserr(fs, "bad block");
 		return (1);
 	}
@@ -608,13 +872,10 @@ badblock(fs, bn)
 }
 
 /*
- * 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.
+ * 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 super block magic number is performed.
  *
  * panic: no fs -- the device is not mounted.
  *	this "cannot happen"
@@ -626,20 +887,22 @@ getfs(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);
-		}
+	for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++) {
+		if (mp->m_bufp == NULL || mp->m_dev != dev)
+			continue;
+		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);
 }
 
 /*
- * 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)
@@ -678,17 +941,17 @@ getfsx(dev)
 /*
  * 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.
+ * modified nodes; and it goes through the mount table to initiate
+ * the writing of the modified super blocks.
  */
-update()
+update(flag)
+	int flag;
 {
 	register struct inode *ip;
 	register struct mount *mp;
 	register struct buf *bp;
 	struct fs *fs;
-	time_t tim;
-	int i;
+	int i, blks;
 
 	if (updlock)
 		return;
@@ -698,37 +961,39 @@ update()
 	 * 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");
+	for (mp = &mount[0]; mp < &mount[NMOUNT]; mp++) {
+		if (mp->m_bufp == NULL)
+			continue;
+		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, SBSIZE);
+		fs->fs_fmod = 0;
+		fs->fs_time = time;
+		if (bp->b_un.b_fs != fs)
+			panic("update: bad b_fs");
+		bwrite(bp);
+		blks = howmany(fs->fs_cssize, fs->fs_bsize);
+		for (i = 0; i < blks; i++) {
+			bp = getblk(mp->m_dev,
+			    fsbtodb(fs, fs->fs_csaddr + (i * fs->fs_frag)),
+			    fs->fs_bsize);
 			bwrite(bp);
-			for (i = 0; i < cssize(fs); i += BSIZE) {
-				bp = getblk(mp->m_dev, csaddr(fs) + i / FSIZE,
-					BSIZE);
-				bcopy(fs->fs_cs + i, bp->b_un.b_addr, 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);
-		}
+	for (ip = inode; ip < inodeNINODE; ip++) {
+		if ((ip->i_flag & ILOCK) != 0 || ip->i_count == 0)
+			continue;
+		ip->i_flag |= ILOCK;
+		ip->i_count++;
+		iupdat(ip, &time, &time, 0);
+		iput(ip);
+	}
 	updlock = 0;
 	/*
 	 * Force stale buffer cache information to be flushed,
@@ -736,3 +1001,87 @@ update()
 	 */
 	bflush(NODEV);
 }
+
+/*
+ * block operations
+ *
+ * check if a block is available
+ */
+isblock(fs, cp, h)
+	struct fs *fs;
+	unsigned char *cp;
+	int h;
+{
+	unsigned char mask;
+
+	switch (fs->fs_frag) {
+	case 8:
+		return (cp[h] == 0xff);
+	case 4:
+		mask = 0x0f << ((h & 0x1) << 2);
+		return ((cp[h >> 1] & mask) == mask);
+	case 2:
+		mask = 0x03 << ((h & 0x3) << 1);
+		return ((cp[h >> 2] & mask) == mask);
+	case 1:
+		mask = 0x01 << (h & 0x7);
+		return ((cp[h >> 3] & mask) == mask);
+	default:
+		panic("isblock");
+		return (NULL);
+	}
+}
+
+/*
+ * take a block out of the map
+ */
+clrblock(fs, cp, h)
+	struct fs *fs;
+	unsigned char *cp;
+	int h;
+{
+	switch ((fs)->fs_frag) {
+	case 8:
+		cp[h] = 0;
+		return;
+	case 4:
+		cp[h >> 1] &= ~(0x0f << ((h & 0x1) << 2));
+		return;
+	case 2:
+		cp[h >> 2] &= ~(0x03 << ((h & 0x3) << 1));
+		return;
+	case 1:
+		cp[h >> 3] &= ~(0x01 << (h & 0x7));
+		return;
+	default:
+		panic("clrblock");
+		return;
+	}
+}
+
+/*
+ * put a block into the map
+ */
+setblock(fs, cp, h)
+	struct fs *fs;
+	unsigned char *cp;
+	int h;
+{
+	switch (fs->fs_frag) {
+	case 8:
+		cp[h] = 0xff;
+		return;
+	case 4:
+		cp[h >> 1] |= (0x0f << ((h & 0x1) << 2));
+		return;
+	case 2:
+		cp[h >> 2] |= (0x03 << ((h & 0x3) << 1));
+		return;
+	case 1:
+		cp[h >> 3] |= (0x01 << (h & 0x7));
+		return;
+	default:
+		panic("setblock");
+		return;
+	}
+}