[unix-history] / usr / src / sys / ufs / ffs / ufs_bmap.c

/*
 * Copyright (c) 1989, 1991, 1993
 *	The Regents of the University of California.  All rights reserved.
 * (c) UNIX System Laboratories, Inc.
 * All or some portions of this file are derived from material licensed
 * to the University of California by American Telephone and Telegraph
 * Co. or Unix System Laboratories, Inc. and are reproduced herein with
 * the permission of UNIX System Laboratories, Inc.
 *
 * %sccs.include.redist.c%
 *
 *	@(#)ufs_bmap.c	8.6 (Berkeley) %G%
 */

#include <sys/param.h>
#include <sys/buf.h>
#include <sys/proc.h>
#include <sys/vnode.h>
#include <sys/mount.h>
#include <sys/resourcevar.h>
#include <sys/trace.h>

#include <miscfs/specfs/specdev.h>

#include <ufs/ufs/quota.h>
#include <ufs/ufs/inode.h>
#include <ufs/ufs/ufsmount.h>
#include <ufs/ufs/ufs_extern.h>

/*
 * Bmap converts a the logical block number of a file to its physical block
 * number on the disk. The conversion is done by using the logical block
 * number to index into the array of block pointers described by the dinode.
 */
int
ufs_bmap(ap)
	struct vop_bmap_args /* {
		struct vnode *a_vp;
		daddr_t  a_bn;
		struct vnode **a_vpp;
		daddr_t *a_bnp;
		int *a_runp;
	} */ *ap;
{
	/*
	 * Check for underlying vnode requests and ensure that logical
	 * to physical mapping is requested.
	 */
	if (ap->a_vpp != NULL)
		*ap->a_vpp = VTOI(ap->a_vp)->i_devvp;
	if (ap->a_bnp == NULL)
		return (0);

	return (ufs_bmaparray(ap->a_vp, ap->a_bn, ap->a_bnp, NULL, NULL,
	    ap->a_runp));
}

/*
 * Indirect blocks are now on the vnode for the file.  They are given negative
 * logical block numbers.  Indirect blocks are addressed by the negative
 * address of the first data block to which they point.  Double indirect blocks
 * are addressed by one less than the address of the first indirect block to
 * which they point.  Triple indirect blocks are addressed by one less than
 * the address of the first double indirect block to which they point.
 *
 * ufs_bmaparray does the bmap conversion, and if requested returns the
 * array of logical blocks which must be traversed to get to a block.
 * Each entry contains the offset into that block that gets you to the
 * next block and the disk address of the block (if it is assigned).
 */

int
ufs_bmaparray(vp, bn, bnp, ap, nump, runp)
	struct vnode *vp;
	register daddr_t bn;
	daddr_t *bnp;
	struct indir *ap;
	int *nump;
	int *runp;
{
	register struct inode *ip;
	struct buf *bp;
	struct ufsmount *ump;
	struct mount *mp;
	struct vnode *devvp;
	struct indir a[NIADDR], *xap;
	daddr_t daddr;
	long metalbn;
	int error, maxrun, num;

	ip = VTOI(vp);
	mp = vp->v_mount;
	ump = VFSTOUFS(mp);
#ifdef DIAGNOSTIC
	if (ap != NULL && nump == NULL || ap == NULL && nump != NULL)
		panic("ufs_bmaparray: invalid arguments");
#endif

	if (runp) {
		/*
		 * XXX
		 * If MAXBSIZE is the largest transfer the disks can handle,
		 * we probably want maxrun to be 1 block less so that we
		 * don't create a block larger than the device can handle.
		 */
		*runp = 0;
		maxrun = MAXBSIZE / mp->mnt_stat.f_iosize - 1;
	}

	xap = ap == NULL ? a : ap;
	if (!nump)
		nump = &num;
	if (error = ufs_getlbns(vp, bn, xap, nump))
		return (error);

	num = *nump;
	if (num == 0) {
		*bnp = blkptrtodb(ump, ip->i_db[bn]);
		if (*bnp == 0)
			*bnp = -1;
		else if (runp)
			for (++bn; bn < NDADDR && *runp < maxrun &&
			    is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
			    ++bn, ++*runp);
		return (0);
	}


	/* Get disk address out of indirect block array */
	daddr = ip->i_ib[xap->in_off];

	devvp = VFSTOUFS(vp->v_mount)->um_devvp;
	for (bp = NULL, ++xap; --num; ++xap) {
		/* 
		 * Exit the loop if there is no disk address assigned yet and
		 * the indirect block isn't in the cache, or if we were
		 * looking for an indirect block and we've found it.
		 */

		metalbn = xap->in_lbn;
		if (daddr == 0 && !incore(vp, metalbn) || metalbn == bn)
			break;
		/*
		 * If we get here, we've either got the block in the cache
		 * or we have a disk address for it, go fetch it.
		 */
		if (bp)
			brelse(bp);

		xap->in_exists = 1;
		bp = getblk(vp, metalbn, mp->mnt_stat.f_iosize, 0, 0);
		if (bp->b_flags & (B_DONE | B_DELWRI)) {
			trace(TR_BREADHIT, pack(vp, size), metalbn);
		}
#ifdef DIAGNOSTIC
		else if (!daddr)
			panic("ufs_bmaparry: indirect block not in cache");
#endif
		else {
			trace(TR_BREADMISS, pack(vp, size), metalbn);
			bp->b_blkno = blkptrtodb(ump, daddr);
			bp->b_flags |= B_READ;
			VOP_STRATEGY(bp);
			curproc->p_stats->p_ru.ru_inblock++;	/* XXX */
			if (error = biowait(bp)) {
				brelse(bp);
				return (error);
			}
		}

		daddr = ((daddr_t *)bp->b_data)[xap->in_off];
		if (num == 1 && daddr && runp)
			for (bn = xap->in_off + 1;
			    bn < MNINDIR(ump) && *runp < maxrun &&
			    is_sequential(ump, ((daddr_t *)bp->b_data)[bn - 1],
			    ((daddr_t *)bp->b_data)[bn]);
			    ++bn, ++*runp);
	}
	if (bp)
		brelse(bp);

	daddr = blkptrtodb(ump, daddr);
	*bnp = daddr == 0 ? -1 : daddr;
	return (0);
}

/*
 * Create an array of logical block number/offset pairs which represent the
 * path of indirect blocks required to access a data block.  The first "pair"
 * contains the logical block number of the appropriate single, double or
 * triple indirect block and the offset into the inode indirect block array.
 * Note, the logical block number of the inode single/double/triple indirect
 * block appears twice in the array, once with the offset into the i_ib and
 * once with the offset into the page itself.
 */
int
ufs_getlbns(vp, bn, ap, nump)
	struct vnode *vp;
	register daddr_t bn;
	struct indir *ap;
	int *nump;
{
	long metalbn, realbn;
	struct ufsmount *ump;
	int blockcnt, i, numlevels, off;

	ump = VFSTOUFS(vp->v_mount);
	if (nump)
		*nump = 0;
	numlevels = 0;
	realbn = bn;
	if ((long)bn < 0)
		bn = -(long)bn;

	/* The first NDADDR blocks are direct blocks. */
	if (bn < NDADDR)
		return (0);

	/* 
	 * Determine the number of levels of indirection.  After this loop
	 * is done, blockcnt indicates the number of data blocks possible
	 * at the given level of indirection, and NIADDR - i is the number
	 * of levels of indirection needed to locate the requested block.
	 */
	for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
		if (i == 0)
			return (EFBIG);
		blockcnt *= MNINDIR(ump);
		if (bn < blockcnt)
			break;
	}

	/* Calculate the address of the first meta-block. */
	if (realbn >= 0)
		metalbn = -(realbn - bn + NIADDR - i);
	else
		metalbn = -(-realbn - bn + NIADDR - i);

	/* 
	 * At each iteration, off is the offset into the bap array which is
	 * an array of disk addresses at the current level of indirection.
	 * The logical block number and the offset in that block are stored
	 * into the argument array.
	 */
	ap->in_lbn = metalbn;
	ap->in_off = off = NIADDR - i;
	ap->in_exists = 0;
	ap++;
	for (++numlevels; i <= NIADDR; i++) {
		/* If searching for a meta-data block, quit when found. */
		if (metalbn == realbn)
			break;

		blockcnt /= MNINDIR(ump);
		off = (bn / blockcnt) % MNINDIR(ump);

		++numlevels;
		ap->in_lbn = metalbn;
		ap->in_off = off;
		ap->in_exists = 0;
		++ap;

		metalbn -= -1 + off * blockcnt;
	}
	if (nump)
		*nump = numlevels;
	return (0);
}
Commit	Line	Data
d886ada0	1	/*
ad0f93d2 KB	2	* Copyright (c) 1989, 1991, 1993
ad0f93d2 KB	3	* The Regents of the University of California. All rights reserved.
adb35f79 KB	4	* (c) UNIX System Laboratories, Inc.
	5	* All or some portions of this file are derived from material licensed
	6	* to the University of California by American Telephone and Telegraph
	7	* Co. or Unix System Laboratories, Inc. and are reproduced herein with
	8	* the permission of UNIX System Laboratories, Inc.
d886ada0 MS	9	*
	10	* %sccs.include.redist.c%
	11	*
adb35f79	12	* @(#)ufs_bmap.c 8.6 (Berkeley) %G%
d886ada0 MS	13	*/
	14
	15	#include <sys/param.h>
	16	#include <sys/buf.h>
	17	#include <sys/proc.h>
	18	#include <sys/vnode.h>
	19	#include <sys/mount.h>
	20	#include <sys/resourcevar.h>
	21	#include <sys/trace.h>
	22
	23	#include <miscfs/specfs/specdev.h>
	24
	25	#include <ufs/ufs/quota.h>
	26	#include <ufs/ufs/inode.h>
	27	#include <ufs/ufs/ufsmount.h>
	28	#include <ufs/ufs/ufs_extern.h>
	29
	30	/*
	31	* Bmap converts a the logical block number of a file to its physical block
	32	* number on the disk. The conversion is done by using the logical block
	33	* number to index into the array of block pointers described by the dinode.
	34	*/
	35	int
	36	ufs_bmap(ap)
	37	struct vop_bmap_args /* {
	38	struct vnode *a_vp;
	39	daddr_t a_bn;
	40	struct vnode **a_vpp;
	41	daddr_t *a_bnp;
	42	int *a_runp;
	43	} / ap;
	44	{
	45	/*
	46	* Check for underlying vnode requests and ensure that logical
	47	* to physical mapping is requested.
	48	*/
	49	if (ap->a_vpp != NULL)
	50	*ap->a_vpp = VTOI(ap->a_vp)->i_devvp;
	51	if (ap->a_bnp == NULL)
	52	return (0);
	53
	54	return (ufs_bmaparray(ap->a_vp, ap->a_bn, ap->a_bnp, NULL, NULL,
	55	ap->a_runp));
	56	}
	57
	58	/*
	59	* Indirect blocks are now on the vnode for the file. They are given negative
	60	* logical block numbers. Indirect blocks are addressed by the negative
	61	* address of the first data block to which they point. Double indirect blocks
	62	* are addressed by one less than the address of the first indirect block to
	63	* which they point. Triple indirect blocks are addressed by one less than
	64	* the address of the first double indirect block to which they point.
	65	*
	66	* ufs_bmaparray does the bmap conversion, and if requested returns the
	67	* array of logical blocks which must be traversed to get to a block.
	68	* Each entry contains the offset into that block that gets you to the
	69	* next block and the disk address of the block (if it is assigned).
	70	*/
	71
	72	int
	73	ufs_bmaparray(vp, bn, bnp, ap, nump, runp)
	74	struct vnode *vp;
	75	register daddr_t bn;
	76	daddr_t *bnp;
77	struct indir *ap;
78	int *nump;
79	int *runp;
80	{
81	register struct inode *ip;
82	struct buf *bp;
83	struct ufsmount *ump;
84	struct mount *mp;
85	struct vnode *devvp;
86	struct indir a[NIADDR], *xap;
67e8af50	87	daddr_t daddr;
d886ada0	88	long metalbn;
67e8af50	89	int error, maxrun, num;
d886ada0 MS	90
	91	ip = VTOI(vp);
	92	mp = vp->v_mount;
	93	ump = VFSTOUFS(mp);
	94	#ifdef DIAGNOSTIC
	95	if (ap != NULL && nump == NULL \|\| ap == NULL && nump != NULL)
	96	panic("ufs_bmaparray: invalid arguments");
	97	#endif
	98
	99	if (runp) {
	100	/*
5685f766 KB	101	* XXX
	102	* If MAXBSIZE is the largest transfer the disks can handle,
	103	* we probably want maxrun to be 1 block less so that we
	104	* don't create a block larger than the device can handle.
d886ada0 MS	105	*/
d886ada0 MS	106	*runp = 0;
91124c44	107	maxrun = MAXBSIZE / mp->mnt_stat.f_iosize - 1;
d886ada0 MS	108	}
	109
	110	xap = ap == NULL ? a : ap;
	111	if (!nump)
	112	nump = &num;
	113	if (error = ufs_getlbns(vp, bn, xap, nump))
	114	return (error);
	115
	116	num = *nump;
	117	if (num == 0) {
	118	*bnp = blkptrtodb(ump, ip->i_db[bn]);
	119	if (*bnp == 0)
	120	*bnp = -1;
91124c44	121	else if (runp)
d886ada0 MS	122	for (++bn; bn < NDADDR && *runp < maxrun &&
	123	is_sequential(ump, ip->i_db[bn - 1], ip->i_db[bn]);
	124	++bn, ++*runp);
d886ada0 MS	125	return (0);
	126	}
	127
	128
	129	/* Get disk address out of indirect block array */
	130	daddr = ip->i_ib[xap->in_off];
	131
d886ada0	132	devvp = VFSTOUFS(vp->v_mount)->um_devvp;
d886ada0 MS	133	for (bp = NULL, ++xap; --num; ++xap) {
	134	/*
	135	* Exit the loop if there is no disk address assigned yet and
	136	* the indirect block isn't in the cache, or if we were
	137	* looking for an indirect block and we've found it.
	138	*/
	139
	140	metalbn = xap->in_lbn;
	141	if (daddr == 0 && !incore(vp, metalbn) \|\| metalbn == bn)
	142	break;
	143	/*
	144	* If we get here, we've either got the block in the cache
	145	* or we have a disk address for it, go fetch it.
	146	*/
	147	if (bp)
	148	brelse(bp);
	149
	150	xap->in_exists = 1;
103a3ab5	151	bp = getblk(vp, metalbn, mp->mnt_stat.f_iosize, 0, 0);
d886ada0 MS	152	if (bp->b_flags & (B_DONE \| B_DELWRI)) {
	153	trace(TR_BREADHIT, pack(vp, size), metalbn);
	154	}
	155	#ifdef DIAGNOSTIC
	156	else if (!daddr)
	157	panic("ufs_bmaparry: indirect block not in cache");
	158	#endif
	159	else {
	160	trace(TR_BREADMISS, pack(vp, size), metalbn);
	161	bp->b_blkno = blkptrtodb(ump, daddr);
	162	bp->b_flags \|= B_READ;
	163	VOP_STRATEGY(bp);
	164	curproc->p_stats->p_ru.ru_inblock++; /* XXX */
	165	if (error = biowait(bp)) {
	166	brelse(bp);
	167	return (error);
	168	}
	169	}
	170
cb84e0ab	171	daddr = ((daddr_t *)bp->b_data)[xap->in_off];
91124c44	172	if (num == 1 && daddr && runp)
d886ada0 MS	173	for (bn = xap->in_off + 1;
d886ada0 MS	174	bn < MNINDIR(ump) && *runp < maxrun &&
cb84e0ab KB	175	is_sequential(ump, ((daddr_t *)bp->b_data)[bn - 1],
cb84e0ab KB	176	((daddr_t *)bp->b_data)[bn]);
d886ada0	177	++bn, ++*runp);
d886ada0 MS	178	}
	179	if (bp)
	180	brelse(bp);
	181
	182	daddr = blkptrtodb(ump, daddr);
	183	*bnp = daddr == 0 ? -1 : daddr;
	184	return (0);
	185	}
	186
	187	/*
	188	* Create an array of logical block number/offset pairs which represent the
	189	* path of indirect blocks required to access a data block. The first "pair"
	190	* contains the logical block number of the appropriate single, double or
	191	* triple indirect block and the offset into the inode indirect block array.
	192	* Note, the logical block number of the inode single/double/triple indirect
	193	* block appears twice in the array, once with the offset into the i_ib and
	194	* once with the offset into the page itself.
	195	*/
	196	int
	197	ufs_getlbns(vp, bn, ap, nump)
	198	struct vnode *vp;
	199	register daddr_t bn;
	200	struct indir *ap;
	201	int *nump;
	202	{
	203	long metalbn, realbn;
	204	struct ufsmount *ump;
5685f766	205	int blockcnt, i, numlevels, off;
d886ada0 MS	206
	207	ump = VFSTOUFS(vp->v_mount);
	208	if (nump)
	209	*nump = 0;
	210	numlevels = 0;
	211	realbn = bn;
	212	if ((long)bn < 0)
	213	bn = -(long)bn;
	214
	215	/* The first NDADDR blocks are direct blocks. */
	216	if (bn < NDADDR)
	217	return (0);
	218
	219	/*
	220	* Determine the number of levels of indirection. After this loop
5685f766 KB	221	* is done, blockcnt indicates the number of data blocks possible
	222	* at the given level of indirection, and NIADDR - i is the number
	223	* of levels of indirection needed to locate the requested block.
d886ada0	224	*/
5685f766 KB	225	for (blockcnt = 1, i = NIADDR, bn -= NDADDR;; i--, bn -= blockcnt) {
	226	if (i == 0)
	227	return (EFBIG);
	228	blockcnt *= MNINDIR(ump);
	229	if (bn < blockcnt)
d886ada0	230	break;
d886ada0	231	}
d886ada0 MS	232
	233	/* Calculate the address of the first meta-block. */
	234	if (realbn >= 0)
5685f766	235	metalbn = -(realbn - bn + NIADDR - i);
d886ada0	236	else
5685f766	237	metalbn = -(-realbn - bn + NIADDR - i);
d886ada0 MS	238
	239	/*
	240	* At each iteration, off is the offset into the bap array which is
	241	* an array of disk addresses at the current level of indirection.
	242	* The logical block number and the offset in that block are stored
	243	* into the argument array.
	244	*/
d886ada0	245	ap->in_lbn = metalbn;
5685f766	246	ap->in_off = off = NIADDR - i;
d886ada0 MS	247	ap->in_exists = 0;
d886ada0 MS	248	ap++;
5685f766	249	for (++numlevels; i <= NIADDR; i++) {
d886ada0 MS	250	/* If searching for a meta-data block, quit when found. */
	251	if (metalbn == realbn)
	252	break;
	253
5685f766 KB	254	blockcnt /= MNINDIR(ump);
5685f766 KB	255	off = (bn / blockcnt) % MNINDIR(ump);
d886ada0 MS	256
	257	++numlevels;
	258	ap->in_lbn = metalbn;
	259	ap->in_off = off;
	260	ap->in_exists = 0;
	261	++ap;
	262
5685f766	263	metalbn -= -1 + off * blockcnt;
d886ada0 MS	264	}
	265	if (nump)
	266	*nump = numlevels;
	267	return (0);
	268	}