ffs_alloc.c

早期freebsd实现

	else
		ipref = pip->i_number;
	if (ipref >= fs->fs_ncg * fs->fs_ipg)
		ipref = 0;
	cg = ino_to_cg(fs, ipref);
	ino = (ino_t)ffs_hashalloc(pip, cg, (long)ipref, mode, ffs_nodealloccg);
	if (ino == 0)
		goto noinodes;
	error = VFS_VGET(pvp->v_mount, ino, ap->a_vpp);
	if (error) {
		VOP_VFREE(pvp, ino, mode);
		return (error);
	}
	ip = VTOI(*ap->a_vpp);
	if (ip->i_mode) {
		printf("mode = 0%o, inum = %d, fs = %s\n",
		    ip->i_mode, ip->i_number, fs->fs_fsmnt);
		panic("ffs_valloc: 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;
	}
	ip->i_flags = 0;
	/*
	 * Set up a new generation number for this inode.
	 */
	if (++nextgennumber < (u_long)time.tv_sec)
		nextgennumber = time.tv_sec;
	ip->i_gen = nextgennumber;
	return (0);
noinodes:
	ffs_fserr(fs, ap->a_cred->cr_uid, "out of inodes");
	uprintf("\n%s: create/symlink failed, no inodes free\n", fs->fs_fsmnt);
	return (ENOSPC);
}

/*
 * 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.
 */
static ino_t
ffs_dirpref(fs)
	register struct fs *fs;
{
	int cg, minndir, mincg, avgifree;

	avgifree = fs->fs_cstotal.cs_nifree / fs->fs_ncg;
	minndir = fs->fs_ipg;
	mincg = 0;
	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(fs, cg).cs_ndir;
		}
	return ((ino_t)(fs->fs_ipg * mincg));
}

/*
 * 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.
 */
daddr_t
ffs_blkpref(ip, lbn, indx, bap)
	struct inode *ip;
	daddr_t lbn;
	int indx;
	daddr_t *bap;
{
	register struct fs *fs;
	register int cg;
	int avgbfree, startcg;
	daddr_t nextblk;

	fs = ip->i_fs;
	if (indx % fs->fs_maxbpg == 0 || bap[indx - 1] == 0) {
		if (lbn < NDADDR) {
			cg = ino_to_cg(fs, ip->i_number);
			return (fs->fs_fpg * cg + fs->fs_frag);
		}
		/*
		 * Find a cylinder with greater than average number of
		 * unused data blocks.
		 */
		if (indx == 0 || bap[indx - 1] == 0)
			startcg =
			    ino_to_cg(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);
}

/*
 * 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*/
static u_long
ffs_hashalloc(ip, cg, pref, size, allocator)
	struct inode *ip;
	int cg;
	long pref;
	int size;	/* size for data blocks, mode for inodes */
	u_long (*allocator)();
{
	register struct fs *fs;
	long result;
	int i, icg = cg;

	fs = ip->i_fs;
	/*
	 * 1: preferred cylinder group
	 */
	result = (*allocator)(ip, cg, pref, size);
	if (result)
		return (result);
	/*
	 * 2: quadratic rehash
	 */
	for (i = 1; i < fs->fs_ncg; i *= 2) {
		cg += i;
		if (cg >= fs->fs_ncg)
			cg -= fs->fs_ncg;
		result = (*allocator)(ip, cg, 0, size);
		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 + 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;
	}
	return (NULL);
}

/*
 * Determine whether a fragment can be extended.
 *
 * Check to see if the necessary fragments are available, and 
 * if they are, allocate them.
 */
static daddr_t
ffs_fragextend(ip, cg, bprev, osize, nsize)
	struct inode *ip;
	int cg;
	long bprev;
	int osize, nsize;
{
	register struct fs *fs;
	register struct cg *cgp;
	struct buf *bp;
	long bno;
	int frags, bbase;
	int i, error;

	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);
	}
	error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
		(int)fs->fs_cgsize, NOCRED, &bp);
	if (error) {
		brelse(bp);
		return (NULL);
	}
	cgp = (struct cg *)bp->b_data;
	if (!cg_chkmagic(cgp)) {
		brelse(bp);
		return (NULL);
	}
	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);
		}
	/*
	 * 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--;
	}
	fs->fs_fmod = 1;
	bdwrite(bp);
	return (bprev);
}

/*
 * Determine whether a block can be allocated.
 *
 * Check to see if a block of the appropriate size is available,
 * and if it is, allocate it.
 */
static daddr_t
ffs_alloccg(ip, cg, bpref, size)
	struct inode *ip;
	int cg;
	daddr_t bpref;
	int size;
{
	register struct fs *fs;
	register struct cg *cgp;
	struct buf *bp;
	register int i;
	int error, bno, frags, allocsiz;

	fs = ip->i_fs;
	if (fs->fs_cs(fs, cg).cs_nbfree == 0 && size == fs->fs_bsize)
		return (NULL);
	error = bread(ip->i_devvp, fsbtodb(fs, cgtod(fs, cg)),
		(int)fs->fs_cgsize, NOCRED, &bp);
	if (error) {
		brelse(bp);
		return (NULL);
	}
	cgp = (struct cg *)bp->b_data;
	if (!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 = ffs_alloccgblk(fs, cgp, bpref);
		bdwrite(bp);
		return (bno);
	}
	/*
	 * check to see if any fragments are already available
	 * allocsiz is the size which will be allocated, hacking
	 * it down to a smaller size if necessary
	 */
	frags = numfrags(fs, size);
	for (allocsiz = frags; allocsiz < fs->fs_frag; allocsiz++)
		if (cgp->cg_frsum[allocsiz] != 0)
			break;
	if (allocsiz == fs->fs_frag) {
		/*
		 * no fragments were available, so a block will be 
		 * allocated, and hacked up
		 */
		if (cgp->cg_cs.cs_nbfree == 0) {
			brelse(bp);
			return (NULL);
		}
		bno = ffs_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 = 1;
		cgp->cg_frsum[i]++;
		bdwrite(bp);
		return (bno);
	}
	bno = ffs_mapsearch(fs, cgp, bpref, allocsiz);
	if (bno < 0) {
		brelse(bp);
		return (NULL);
	}
	for (i = 0; i < frags; i++)
		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 = 1;
	cgp->cg_frsum[allocsiz]--;
	if (frags != allocsiz)
		cgp->cg_frsum[allocsiz - frags]++;
	bdwrite(bp);
	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.
 */
static daddr_t
ffs_alloccgblk(fs, cgp, bpref)
	register struct fs *fs;
	register struct cg *cgp;
	daddr_t bpref;
{
	daddr_t bno, blkno;
	int cylno, pos, delta;
	short *cylbp;
	register int i;

	if (bpref == 0 || dtog(fs, bpref) != cgp->cg_cgx) {
		bpref = cgp->cg_rotor;
		goto norot;
	}
	bpref = blknum(fs, bpref);
	bpref = dtogd(fs, bpref);
	/*
	 * if the requested block is available, use it
	 */
	if (ffs_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) {
		/*
		 * Block layout information is not available.
		 * Leaving bpref unchanged means we take the
		 * next available free block following the one 
		 * we just allocated. Hopefully this will at
		 * least hit a track cache on drives of unknown
		 * geometry (e.g. SCSI).
		 */
		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) {
		/*
		 * 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("ffs_alloccgblk: cyl groups corrupted");
		}
		for (i = fs_postbl(fs, pos)[i];; ) {
			if (ffs_isblock(fs, cg_blksfree(cgp), bno + i)) {
				bno = blkstofrags(fs, (bno + i));
				goto gotit;
			}
			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("ffs_alloccgblk: can't find blk in cyl");
	}
norot:
	/*
	 * no blocks in the requested cylinder, so take next
	 * available one in this cylinder group.
	 */
	bno = ffs_mapsearch(fs, cgp, bpref, (int)fs->fs_frag);
	if (bno < 0)
		return (NULL);
	cgp->cg_rotor = bno;
gotit:
	blkno = fragstoblks(fs, bno);
	ffs_clrblock(fs, cg_blksfree(cgp), (long)blkno);
	ffs_clusteracct(fs, cgp, blkno, -1);
	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 = 1;
	return (cgp->cg_cgx * fs->fs_fpg + bno);
}

/*
 * Determine whether a cluster can be allocated.
 *
 * We do not currently check for optimal rotational layout if there
 * are multiple choices in the same cylinder group. Instead we just
 * take the first one that we find following bpref.
 */
static daddr_t
ffs_clusteralloc(ip, cg, bpref, len)
	struct inode *ip;