【4.4BSD源码阅读】Local Filestores

为了方便查阅，所有的源码均给出完整内容。

Fast File System (FFS)

下面是FFS中超级块的数据结构定义：

/*
 * Super block for an FFS file system.
 */
struct fs {
	int32_t	 fs_firstfield;		/* historic file system linked list, */
	int32_t	 fs_unused_1;		/*     used for incore super blocks */
	ufs_daddr_t fs_sblkno;		/* addr of super-block in filesys */
	ufs_daddr_t fs_cblkno;		/* offset of cyl-block in filesys */
	ufs_daddr_t fs_iblkno;		/* offset of inode-blocks in filesys */
	ufs_daddr_t fs_dblkno;		/* offset of first data after cg */
	int32_t	 fs_cgoffset;		/* cylinder group offset in cylinder */
	int32_t	 fs_cgmask;		/* used to calc mod fs_ntrak */
	time_t 	 fs_time;		/* last time written */
	int32_t	 fs_size;		/* number of blocks in fs */
	int32_t	 fs_dsize;		/* number of data blocks in fs */
	int32_t	 fs_ncg;		/* number of cylinder groups */
	int32_t	 fs_bsize;		/* size of basic blocks in fs */
	int32_t	 fs_fsize;		/* size of frag blocks in fs */
	int32_t	 fs_frag;		/* number of frags in a block in fs */
/* these are configuration parameters */
	int32_t	 fs_minfree;		/* minimum percentage of free blocks */
	int32_t	 fs_rotdelay;		/* num of ms for optimal next block */
	int32_t	 fs_rps;		/* disk revolutions per second */
/* these fields can be computed from the others */
	int32_t	 fs_bmask;		/* ``blkoff'' calc of blk offsets */
	int32_t	 fs_fmask;		/* ``fragoff'' calc of frag offsets */
	int32_t	 fs_bshift;		/* ``lblkno'' calc of logical blkno */
	int32_t	 fs_fshift;		/* ``numfrags'' calc number of frags */
/* these are configuration parameters */
	int32_t	 fs_maxcontig;		/* max number of contiguous blks */
	int32_t	 fs_maxbpg;		/* max number of blks per cyl group */
/* these fields can be computed from the others */
	int32_t	 fs_fragshift;		/* block to frag shift */
	int32_t	 fs_fsbtodb;		/* fsbtodb and dbtofsb shift constant */
	int32_t	 fs_sbsize;		/* actual size of super block */
	int32_t	 fs_csmask;		/* csum block offset */
	int32_t	 fs_csshift;		/* csum block number */
	int32_t	 fs_nindir;		/* value of NINDIR */
	int32_t	 fs_inopb;		/* value of INOPB */
	int32_t	 fs_nspf;		/* value of NSPF */
/* yet another configuration parameter */
	int32_t	 fs_optim;		/* optimization preference, see below */
/* these fields are derived from the hardware */
	int32_t	 fs_npsect;		/* # sectors/track including spares */
	int32_t	 fs_interleave;		/* hardware sector interleave */
	int32_t	 fs_trackskew;		/* sector 0 skew, per track */
	int32_t	 fs_headswitch;		/* head switch time, usec */
	int32_t	 fs_trkseek;		/* track-to-track seek, usec */
/* sizes determined by number of cylinder groups and their sizes */
	ufs_daddr_t fs_csaddr;		/* blk addr of cyl grp summary area */
	int32_t	 fs_cssize;		/* size of cyl grp summary area */
	int32_t	 fs_cgsize;		/* cylinder group size */
/* these fields are derived from the hardware */
	int32_t	 fs_ntrak;		/* tracks per cylinder */
	int32_t	 fs_nsect;		/* sectors per track */
	int32_t  fs_spc;			/* sectors per cylinder */
/* this comes from the disk driver partitioning */
	int32_t	 fs_ncyl;		/* cylinders in file system */
/* these fields can be computed from the others */
	int32_t	 fs_cpg;			/* cylinders per group */
	int32_t	 fs_ipg;			/* inodes per group */
	int32_t	 fs_fpg;			/* blocks per group * fs_frag */
/* this data must be re-computed after crashes */
	struct	csum fs_cstotal;	/* cylinder summary information */
/* these fields are cleared at mount time */
	int8_t   fs_fmod;		/* super block modified flag */
	int8_t   fs_clean;		/* file system is clean flag */
	int8_t 	 fs_ronly;		/* mounted read-only flag */
	int8_t   fs_flags;		/* currently unused flag */
	u_char	 fs_fsmnt[MAXMNTLEN];	/* name mounted on */
/* these fields retain the current block allocation info */
	int32_t	 fs_cgrotor;		/* last cg searched */
	struct	csum *fs_csp[MAXCSBUFS];/* list of fs_cs info buffers */
	int32_t	 *fs_maxcluster;	/* max cluster in each cyl group */
	int32_t	 fs_cpc;		/* cyl per cycle in postbl */
	int16_t	 fs_opostbl[16][8];	/* old rotation block list head */
	int32_t	 fs_sparecon[50];	/* reserved for future constants */
	int32_t	 fs_contigsumsize;	/* size of cluster summary array */ 
	int32_t	 fs_maxsymlinklen;	/* max length of an internal symlink */
	int32_t	 fs_inodefmt;		/* format of on-disk inodes */
	u_int64_t fs_maxfilesize;	/* maximum representable file size */
	int64_t	 fs_qbmask;		/* ~fs_bmask for use with 64-bit size */
	int64_t	 fs_qfmask;		/* ~fs_fmask for use with 64-bit size */
	int32_t	 fs_state;		/* validate fs_clean field */
	int32_t	 fs_postblformat;	/* format of positional layout tables */
	int32_t	 fs_nrpos;		/* number of rotational positions */
	int32_t	 fs_postbloff;		/* (u_int16) rotation block list head */
	int32_t	 fs_rotbloff;		/* (u_int8) blocks for each rotation */
	int32_t	 fs_magic;		/* magic number */
	u_int8_t fs_space[1];		/* list of blocks for each rotation */
/* actually longer */
};

先说下一些先决知识，FFS将磁盘分为若干个柱面组（Cylinder Group），所有的块（block） 都可以划分为大小相等的 段（fragment），而每个文件只有最后一个块才能被分段，前面的所有块都是完整的。

在申请空闲块时，只有对齐的位置才能分配，也就是不能将跨越两个块的段组合成一个块。同时，已被分段的块，不能重新合并为块，必须另外分配空闲块，然后将内容复制过去。

负责管理块和段分配的函数为ffs_balloc()，下面为源码：

/*
 * Balloc defines the structure of file system storage
 * by allocating the physical blocks on a device given
 * the inode and the logical block number in a file.
 */
ffs_balloc(ip, lbn, size, cred, bpp, flags)
	register struct inode *ip;  //该文件的inode指针
	register ufs_daddr_t lbn;   //逻辑块号
	int size;                   //申请的数据大小
	struct ucred *cred;         //用户相关信息
	struct buf **bpp;           //内核中的缓冲区，用于执行I/O
	int flags;
{
	register struct fs *fs;
	register ufs_daddr_t nb;
	struct buf *bp, *nbp;
	struct vnode *vp = ITOV(ip);     //根据inode得到vnode
	struct indir indirs[NIADDR + 2];
	ufs_daddr_t newb, *bap, pref;
	int deallocated, osize, nsize, num, i, error;
	ufs_daddr_t *allocib, *blkp, *allocblk, allociblk[NIADDR + 1];

	*bpp = NULL;
	if (lbn < 0)
		return (EFBIG);
	fs = ip->i_fs;

	/*
	 * If the next write will extend the file into a new block,
	 * and the file is currently composed of a fragment
	 * this fragment has to be extended to be a full block.
	 */
	nb = lblkno(fs, ip->i_size);  //该文件的完整块数目
	if (nb < NDADDR && nb < lbn) {  //块数小于正在写的块的块号
		osize = blksize(fs, ip, nb);  //不构成完整块的最后部分的大小（要么为块大小，要么为段大小的倍数）
		if (osize < fs->fs_bsize && osize > 0) {  //最后一个块为分段块
			error = ffs_realloccg(ip, nb,
				ffs_blkpref(ip, nb, (int)nb, &ip->i_db[0]),
				osize, (int)fs->fs_bsize, cred, &bp);  //将段重新升级为块
			if (error)
				return (error);
			ip->i_size = (nb + 1) * fs->fs_bsize;
			vnode_pager_setsize(vp, (u_long)ip->i_size);
			ip->i_db[nb] = dbtofsb(fs, bp->b_blkno);
			ip->i_flag |= IN_CHANGE | IN_UPDATE;
			if (flags & B_SYNC)
				bwrite(bp);
			else
				bawrite(bp);
		}
	}
	/*
	 * The first NDADDR blocks are direct blocks
	 */
	if (lbn < NDADDR) {
		nb = ip->i_db[lbn];
		if (nb != 0 && ip->i_size >= (lbn + 1) * fs->fs_bsize) {
			error = bread(vp, lbn, fs->fs_bsize, NOCRED, &bp);
			if (error) {
				brelse(bp);
				return (error);
			}
			*bpp = bp;
			return (0);
		}
		if (nb != 0) {
			/*
			 * Consider need to reallocate a fragment.
			 */
			osize = fragroundup(fs, blkoff(fs, ip->i_size));
			nsize = fragroundup(fs, size);
			if (nsize <= osize) {
				error = bread(vp, lbn, osize, NOCRED, &bp);
				if (error) {
					brelse(bp);
					return (error);
				}
			} else {
				error = ffs_realloccg(ip, lbn,
				    ffs_blkpref(ip, lbn, (int)lbn,
					&ip->i_db[0]), osize, nsize, cred, &bp);
				if (error)
					return (error);
			}
		} else {
			if (ip->i_size < (lbn + 1) * fs->fs_bsize)
				nsize = fragroundup(fs, size);
			else
				nsize = fs->fs_bsize;
			error = ffs_alloc(ip, lbn,
			    ffs_blkpref(ip, lbn, (int)lbn, &ip->i_db[0]),
			    nsize, cred, &newb);
			if (error)
				return (error);
			bp = getblk(vp, lbn, nsize, 0, 0);
			bp->b_blkno = fsbtodb(fs, newb);
			if (flags & B_CLRBUF)
				clrbuf(bp);
		}
		ip->i_db[lbn] = dbtofsb(fs, bp->b_blkno);
		ip->i_flag |= IN_CHANGE | IN_UPDATE;
		*bpp = bp;
		return (0);
	}
	/*
	 * Determine the number of levels of indirection.
	 */
	pref = 0;
	if (error = ufs_getlbns(vp, lbn, indirs, &num))
		return(error);
#ifdef DIAGNOSTIC
	if (num < 1)
		panic ("ffs_balloc: ufs_bmaparray returned indirect block\n");
#endif
	/*
	 * Fetch the first indirect block allocating if necessary.
	 */
	--num;
	nb = ip->i_ib[indirs[0].in_off];
	allocib = NULL;
	allocblk = allociblk;
	if (nb == 0) {
		pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
	        if (error = ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize,
		    cred, &newb))
			return (error);
		nb = newb;
		*allocblk++ = nb;
		bp = getblk(vp, indirs[1].in_lbn, fs->fs_bsize, 0, 0);
		bp->b_blkno = fsbtodb(fs, nb);
		clrbuf(bp);
		/*
		 * Write synchronously so that indirect blocks
		 * never point at garbage.
		 */
		if (error = bwrite(bp))
			goto fail;
		allocib = &ip->i_ib[indirs[0].in_off];
		*allocib = nb;
		ip->i_flag |= IN_CHANGE | IN_UPDATE;
	}
	/*
	 * Fetch through the indirect blocks, allocating as necessary.
	 */
	for (i = 1;;) {
		error = bread(vp,
		    indirs[i].in_lbn, (int)fs->fs_bsize, NOCRED, &bp);
		if (error) {
			brelse(bp);
			goto fail;
		}
		bap = (ufs_daddr_t *)bp->b_data;
		nb = bap[indirs[i].in_off];
		if (i == num)
			break;
		i += 1;
		if (nb != 0) {
			brelse(bp);
			continue;
		}
		if (pref == 0)
			pref = ffs_blkpref(ip, lbn, 0, (ufs_daddr_t *)0);
		if (error =
		    ffs_alloc(ip, lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
			brelse(bp);
			goto fail;
		}
		nb = newb;
		*allocblk++ = nb;
		nbp = getblk(vp, indirs[i].in_lbn, fs->fs_bsize, 0, 0);
		nbp->b_blkno = fsbtodb(fs, nb);
		clrbuf(nbp);
		/*
		 * Write synchronously so that indirect blocks
		 * never point at garbage.
		 */
		if (error = bwrite(nbp)) {
			brelse(bp);
			goto fail;
		}
		bap[indirs[i - 1].in_off] = nb;
		/*
		 * If required, write synchronously, otherwise use
		 * delayed write.
		 */
		if (flags & B_SYNC) {
			bwrite(bp);
		} else {
			bdwrite(bp);
		}
	}
	/*
	 * Get the data block, allocating if necessary.
	 */
	if (nb == 0) {
		pref = ffs_blkpref(ip, lbn, indirs[i].in_off, &bap[0]);
		if (error = ffs_alloc(ip,
		    lbn, pref, (int)fs->fs_bsize, cred, &newb)) {
			brelse(bp);
			goto fail;
		}
		nb = newb;
		*allocblk++ = nb;
		nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
		nbp->b_blkno = fsbtodb(fs, nb);
		if (flags & B_CLRBUF)
			clrbuf(nbp);
		bap[indirs[i].in_off] = nb;
		/*
		 * If required, write synchronously, otherwise use
		 * delayed write.
		 */
		if (flags & B_SYNC) {
			bwrite(bp);
		} else {
			bdwrite(bp);
		}
		*bpp = nbp;
		return (0);
	}
	brelse(bp);
	if (flags & B_CLRBUF) {
		error = bread(vp, lbn, (int)fs->fs_bsize, NOCRED, &nbp);
		if (error) {
			brelse(nbp);
			goto fail;
		}
	} else {
		nbp = getblk(vp, lbn, fs->fs_bsize, 0, 0);
		nbp->b_blkno = fsbtodb(fs, nb);
	}
	*bpp = nbp;
	return (0);
fail:
	/*
	 * If we have failed part way through block allocation, we
	 * have to deallocate any indirect blocks that we have allocated.
	 */
	for (deallocated = 0, blkp = allociblk; blkp < allocblk; blkp++) {
		ffs_blkfree(ip, *blkp, fs->fs_bsize);
		deallocated += fs->fs_bsize;
	}
	if (allocib != NULL)
		*allocib = 0;
	if (deallocated) {
#ifdef QUOTA
		/*
		 * Restore user's disk quota because allocation failed.
		 */
		(void) chkdq(ip, (long)-btodb(deallocated), cred, FORCE);
#endif
		ip->i_blocks -= btodb(deallocated);
		ip->i_flag |= IN_CHANGE | IN_UPDATE;
	}
	return (error);
}