vfs_bio.c

来自「基于组件方式开发操作系统的OSKIT源代码」· C语言 代码 · 共 871 行 · 第 1/2 页

/*	$NetBSD: vfs_bio.c,v 1.43.4.1 1996/06/18 19:58:04 pk Exp $	*/

/*-
 * Copyright (c) 1994 Christopher G. Demetriou
 * Copyright (c) 1982, 1986, 1989, 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.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 * 3. All advertising materials mentioning features or use of this software
 *    must display the following acknowledgement:
 *	This product includes software developed by the University of
 *	California, Berkeley and its contributors.
 * 4. Neither the name of the University nor the names of its contributors
 *    may be used to endorse or promote products derived from this software
 *    without specific prior written permission.
 *
 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 *
 *	@(#)vfs_bio.c	8.6 (Berkeley) 1/11/94
 */

/*
 * Some references:
 *	Bach: The Design of the UNIX Operating System (Prentice Hall, 1986)
 *	Leffler, et al.: The Design and Implementation of the 4.3BSD
 *		UNIX Operating System (Addison Welley, 1989)
 */

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

#include <vm/vm.h>

/* Macros to clear/set/test flags. */
#define	SET(t, f)	(t) |= (f)
#define	CLR(t, f)	(t) &= ~(f)
#define	ISSET(t, f)	((t) & (f))

/*
 * Definitions for the buffer hash lists.
 */
#define	BUFHASH(dvp, lbn)	\
	(&bufhashtbl[((long)(dvp) / sizeof(*(dvp)) + (int)(lbn)) & bufhash])
LIST_HEAD(bufhashhdr, buf) *bufhashtbl, invalhash;
u_long	bufhash;

/*
 * Insq/Remq for the buffer hash lists.
 */
#define	binshash(bp, dp)	LIST_INSERT_HEAD(dp, bp, b_hash)
#define	bremhash(bp)		LIST_REMOVE(bp, b_hash)

/*
 * Definitions for the buffer free lists.
 */
#define	BQUEUES		4		/* number of free buffer queues */

#define	BQ_LOCKED	0		/* super-blocks &c */
#define	BQ_LRU		1		/* lru, useful buffers */
#define	BQ_AGE		2		/* rubbish */
#define	BQ_EMPTY	3		/* buffer headers with no memory */

TAILQ_HEAD(bqueues, buf) bufqueues[BQUEUES];
int needbuffer;

/*
 * Insq/Remq for the buffer free lists.
 */
#define	binsheadfree(bp, dp)	TAILQ_INSERT_HEAD(dp, bp, b_freelist)
#define	binstailfree(bp, dp)	TAILQ_INSERT_TAIL(dp, bp, b_freelist)

static __inline struct buf *bio_doread __P((struct vnode *, daddr_t, int,
					    struct ucred *, int));
int count_lock_queue __P((void));

void
bremfree(bp)
	struct buf *bp;
{
	struct bqueues *dp = NULL;

	/*
	 * We only calculate the head of the freelist when removing
	 * the last element of the list as that is the only time that
	 * it is needed (e.g. to reset the tail pointer).
	 *
	 * NB: This makes an assumption about how tailq's are implemented.
	 */
	if (bp->b_freelist.tqe_next == NULL) {
		for (dp = bufqueues; dp < &bufqueues[BQUEUES]; dp++)
			if (dp->tqh_last == &bp->b_freelist.tqe_next)
				break;
		if (dp == &bufqueues[BQUEUES])
			panic("bremfree: lost tail");
	}
	TAILQ_REMOVE(dp, bp, b_freelist);
}

/*
 * Initialize buffers and hash links for buffers.
 */
void
bufinit()
{
	register struct buf *bp;
	struct bqueues *dp;
	register int i;
	int base, residual;

	for (dp = bufqueues; dp < &bufqueues[BQUEUES]; dp++)
		TAILQ_INIT(dp);
	bufhashtbl = hashinit(nbuf, M_CACHE, &bufhash);
	base = bufpages / nbuf;
	residual = bufpages % nbuf;
	for (i = 0; i < nbuf; i++) {
		bp = &buf[i];
		bzero((char *)bp, sizeof *bp);
		bp->b_dev = NODEV;
		bp->b_rcred = NOCRED;
		bp->b_wcred = NOCRED;
		bp->b_vnbufs.le_next = NOLIST;
		bp->b_data = buffers + i * MAXBSIZE;
		if (i < residual)
			bp->b_bufsize = (base + 1) * CLBYTES;
		else
			bp->b_bufsize = base * CLBYTES;
		bp->b_flags = B_INVAL;
		dp = bp->b_bufsize ? &bufqueues[BQ_AGE] : &bufqueues[BQ_EMPTY];
		binsheadfree(bp, dp);
		binshash(bp, &invalhash);
	}
}

static __inline struct buf *
bio_doread(vp, blkno, size, cred, async)
	struct vnode *vp;
	daddr_t blkno;
	int size;
	struct ucred *cred;
	int async;
{
	register struct buf *bp;

	bp = getblk(vp, blkno, size, 0, 0);

	/*
	 * If buffer does not have data valid, start a read.
	 * Note that if buffer is B_INVAL, getblk() won't return it.
	 * Therefore, it's valid if it's I/O has completed or been delayed.
	 */
	if (!ISSET(bp->b_flags, (B_DONE | B_DELWRI))) {
		/* Start I/O for the buffer (keeping credentials). */
		SET(bp->b_flags, B_READ | async);
		if (cred != NOCRED && bp->b_rcred == NOCRED) {
			crhold(cred);
			bp->b_rcred = cred;
		}
		VOP_STRATEGY(bp);

		/* Pay for the read. */
		curproc->p_stats->p_ru.ru_inblock++;		/* XXX */
	} else if (async) {
		brelse(bp);
	}

	return (bp);
}

/*
 * Read a disk block.
 * This algorithm described in Bach (p.54).
 */
int
bread(vp, blkno, size, cred, bpp)
	struct vnode *vp;
	daddr_t blkno;
	int size;
	struct ucred *cred;
	struct buf **bpp;
{
	register struct buf *bp;

	/* Get buffer for block. */
	bp = *bpp = bio_doread(vp, blkno, size, cred, 0);

	/* Wait for the read to complete, and return result. */
	return (biowait(bp));
}

/*
 * Read-ahead multiple disk blocks. The first is sync, the rest async.
 * Trivial modification to the breada algorithm presented in Bach (p.55).
 */
int
breadn(vp, blkno, size, rablks, rasizes, nrablks, cred, bpp)
	struct vnode *vp;
	daddr_t blkno; int size;
	daddr_t rablks[]; int rasizes[];
	int nrablks;
	struct ucred *cred;
	struct buf **bpp;
{
	register struct buf *bp;
	int i;

	bp = *bpp = bio_doread(vp, blkno, size, cred, 0);

	/*
	 * For each of the read-ahead blocks, start a read, if necessary.
	 */
	for (i = 0; i < nrablks; i++) {
		/* If it's in the cache, just go on to next one. */
		if (incore(vp, rablks[i]))
			continue;

		/* Get a buffer for the read-ahead block */
		(void) bio_doread(vp, rablks[i], rasizes[i], cred, B_ASYNC);
	}

	/* Otherwise, we had to start a read for it; wait until it's valid. */
	return (biowait(bp));
}

/*
 * Read with single-block read-ahead.  Defined in Bach (p.55), but
 * implemented as a call to breadn().
 * XXX for compatibility with old file systems.
 */
int
breada(vp, blkno, size, rablkno, rabsize, cred, bpp)
	struct vnode *vp;
	daddr_t blkno; int size;
	daddr_t rablkno; int rabsize;
	struct ucred *cred;
	struct buf **bpp;
{

	return (breadn(vp, blkno, size, &rablkno, &rabsize, 1, cred, bpp));	
}

/*
 * Block write.  Described in Bach (p.56)
 */
int
bwrite(bp)
	struct buf *bp;
{
	int rv, sync, wasdelayed, s;

	/*
	 * Remember buffer type, to switch on it later.  If the write was
	 * synchronous, but the file system was mounted with MNT_ASYNC,
	 * convert it to a delayed write.  
	 * XXX note that this relies on delayed tape writes being converted
	 * to async, not sync writes (which is safe, but ugly).
	 */
	sync = !ISSET(bp->b_flags, B_ASYNC);
	if (sync && bp->b_vp && bp->b_vp->v_mount &&
	    ISSET(bp->b_vp->v_mount->mnt_flag, MNT_ASYNC)) {
		bdwrite(bp);
		return (0);
	}
	wasdelayed = ISSET(bp->b_flags, B_DELWRI);
	CLR(bp->b_flags, (B_READ | B_DONE | B_ERROR | B_DELWRI));

	s = splbio();
	if (!sync) {
		/*
		 * If not synchronous, pay for the I/O operation and make
		 * sure the buf is on the correct vnode queue.  We have
		 * to do this now, because if we don't, the vnode may not
		 * be properly notified that its I/O has completed.
		 */
		if (wasdelayed)
			reassignbuf(bp, bp->b_vp);
		else
			curproc->p_stats->p_ru.ru_oublock++;
	}

	/* Initiate disk write.  Make sure the appropriate party is charged. */
	bp->b_vp->v_numoutput++;
	splx(s);
	SET(bp->b_flags, B_WRITEINPROG);
	VOP_STRATEGY(bp);

	if (sync) {
		/*
		 * If I/O was synchronous, wait for it to complete.
		 */
		rv = biowait(bp);

		/*
		 * Pay for the I/O operation, if it's not been paid for, and
		 * make sure it's on the correct vnode queue. (async operatings
		 * were payed for above.)
		 */
		s = splbio();
		if (wasdelayed)
			reassignbuf(bp, bp->b_vp);
		else
			curproc->p_stats->p_ru.ru_oublock++;
		splx(s);

		/* Release the buffer. */
		brelse(bp);

		return (rv);
	} else {
		return (0);
	}
}

int
vn_bwrite(v)
	void *v;
{
	struct vop_bwrite_args *ap = v;

	return (bwrite(ap->a_bp));
}

/*
 * Delayed write.
 *
 * The buffer is marked dirty, but is not queued for I/O.
 * This routine should be used when the buffer is expected
 * to be modified again soon, typically a small write that
 * partially fills a buffer.
 *
 * NB: magnetic tapes cannot be delayed; they must be
 * written in the order that the writes are requested.
 *
 * Described in Leffler, et al. (pp. 208-213).
 */
void
bdwrite(bp)
	struct buf *bp;
{
	int s;

	/*
	 * If the block hasn't been seen before:
	 *	(1) Mark it as having been seen,
	 *	(2) Charge for the write,
	 *	(3) Make sure it's on its vnode's correct block list.
	 */
	if (!ISSET(bp->b_flags, B_DELWRI)) {
		SET(bp->b_flags, B_DELWRI);
		curproc->p_stats->p_ru.ru_oublock++;	/* XXX */
		s = splbio();
		reassignbuf(bp, bp->b_vp);
		splx(s);
	}

	/* If this is a tape block, write the block now. */
	if (bdevsw[major(bp->b_dev)].d_type == D_TAPE) {
		bawrite(bp);
		return;
	}

	/* Otherwise, the "write" is done, so mark and release the buffer. */
	CLR(bp->b_flags, B_NEEDCOMMIT);
	SET(bp->b_flags, B_DONE);
	brelse(bp);
}

/*
 * Asynchronous block write; just an asynchronous bwrite().
 */
void
bawrite(bp)
	struct buf *bp;
{

	SET(bp->b_flags, B_ASYNC);
	VOP_BWRITE(bp);
}

/*
 * Release a buffer on to the free lists.
 * Described in Bach (p. 46).
 */
void
brelse(bp)
	struct buf *bp;
{
	struct bqueues *bufq;
	int s;

	/* Wake up any processes waiting for any buffer to become free. */
	if (needbuffer) {
		needbuffer = 0;
		wakeup(&needbuffer);
	}

	/* Wake up any proceeses waiting for _this_ buffer to become free. */
	if (ISSET(bp->b_flags, B_WANTED)) {
		CLR(bp->b_flags, B_WANTED);
		wakeup(bp);
	}

	/* Block disk interrupts. */
	s = splbio();

	/*
	 * Determine which queue the buffer should be on, then put it there.