Bring biomem diff back into the tree after the nfs_bio.c fix went in.

ok thib beck art

48 files changed, 367 insertions, 403 deletions

diff --git a/sys/arch/alpha/alpha/disksubr.c b/sys/arch/alpha/alpha/disksubr.c
index 3d3ebf48e5f..30fed88279c 100644
--- a/sys/arch/alpha/alpha/disksubr.c
+++ b/sys/arch/alpha/alpha/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.91 2008/06/11 12:35:38 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.92 2008/06/12 06:58:31 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.21 1996/05/03 19:42:03 christos Exp $	*/
 
 /*
@@ -67,7 +67,7 @@ readdisklabel(dev_t dev, void (*strat)(struct buf *),
 
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp)) {
 		msg = "disk label read error";
@@ -121,7 +121,7 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
@@ -135,7 +135,7 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 		csum += *p++;
 	*p = csum;
 
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/amd64/amd64/disksubr.c b/sys/arch/amd64/amd64/disksubr.c
index 17320717cd2..4315574787f 100644
--- a/sys/arch/amd64/amd64/disksubr.c
+++ b/sys/arch/amd64/amd64/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.55 2008/06/11 12:35:40 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.56 2008/06/12 06:58:33 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.21 1996/05/03 19:42:03 christos Exp $	*/
 
 /*
@@ -141,14 +141,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = partoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/amd64/amd64/dkcsum.c b/sys/arch/amd64/amd64/dkcsum.c
index 9edd48c4745..7eef0cc9d7a 100644
--- a/sys/arch/amd64/amd64/dkcsum.c
+++ b/sys/arch/amd64/amd64/dkcsum.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: dkcsum.c,v 1.13 2008/06/11 12:35:40 deraadt Exp $	*/
+/*	$OpenBSD: dkcsum.c,v 1.14 2008/06/12 06:58:33 deraadt Exp $	*/
 
 /*-
  * Copyright (c) 1997 Niklas Hallqvist.  All rights reserved.
@@ -106,7 +106,7 @@ dkcsumattach(void)
 		/* Read blocks to cksum.  XXX maybe a d_read should be used. */
 		bp->b_blkno = 0;
 		bp->b_bcount = bios_cksumlen * DEV_BSIZE;
-		bp->b_flags = B_BUSY | B_READ;
+		bp->b_flags = B_BUSY | B_READ | B_RAW;
 		bp->b_cylinder = 0;
 		(*bdsw->d_strategy)(bp);
 		if ((error = biowait(bp))) {
diff --git a/sys/arch/arm/arm/disksubr.c b/sys/arch/arm/arm/disksubr.c
index ae50df5e58b..a8750d9646a 100644
--- a/sys/arch/arm/arm/disksubr.c
+++ b/sys/arch/arm/arm/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.50 2008/06/11 12:35:40 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.51 2008/06/12 06:58:33 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.21 1996/05/03 19:42:03 christos Exp $	*/
 
 /*
@@ -114,14 +114,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = partoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/aviion/aviion/disksubr.c b/sys/arch/aviion/aviion/disksubr.c
index dbe54a4f5fb..575ae057f1c 100644
--- a/sys/arch/aviion/aviion/disksubr.c
+++ b/sys/arch/aviion/aviion/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.42 2008/06/11 12:35:40 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.43 2008/06/12 06:58:33 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.21 1996/05/03 19:42:03 christos Exp $	*/
 
 /*
@@ -121,7 +121,7 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/hp300/dev/ct.c b/sys/arch/hp300/dev/ct.c
index 9d333952869..ea1bc02d2f9 100644
--- a/sys/arch/hp300/dev/ct.c
+++ b/sys/arch/hp300/dev/ct.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ct.c,v 1.20 2008/06/11 12:35:40 deraadt Exp $	*/
+/*	$OpenBSD: ct.c,v 1.21 2008/06/12 06:58:33 deraadt Exp $	*/
 /*	$NetBSD: ct.c,v 1.21 1997/04/02 22:37:23 scottr Exp $	*/
 
 /*
@@ -431,7 +431,7 @@ ctcommand(dev, cmd, cnt)
 	}
 
 	while (cnt-- > 0) {
-		bp->b_flags = B_BUSY;
+		bp->b_flags = B_BUSY | B_RAW;
 		if (cmd == MTBSF) {
 			sc->sc_blkno = sc->sc_eofs[sc->sc_eofp];
 			sc->sc_eofp--;
diff --git a/sys/arch/hp300/dev/mt.c b/sys/arch/hp300/dev/mt.c
index 4e7de80a3a8..a0554a5d597 100644
--- a/sys/arch/hp300/dev/mt.c
+++ b/sys/arch/hp300/dev/mt.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: mt.c,v 1.20 2008/06/11 12:35:41 deraadt Exp $	*/
+/*	$OpenBSD: mt.c,v 1.21 2008/06/12 06:58:33 deraadt Exp $	*/
 /*	$NetBSD: mt.c,v 1.8 1997/03/31 07:37:29 scottr Exp $	*/
 
 /*
@@ -407,7 +407,7 @@ mtcommand(dev, cmd, cnt)
 	bp->b_cmd = cmd;
 	bp->b_dev = dev;
 	do {
-		bp->b_flags = B_BUSY | B_CMD;
+		bp->b_flags = B_BUSY | B_CMD | B_RAW;
 		mtstrategy(bp);
 		biowait(bp);
 		if (bp->b_flags & B_ERROR) {
diff --git a/sys/arch/hp300/hp300/disksubr.c b/sys/arch/hp300/hp300/disksubr.c
index 1c5a1480879..8b1df78cbeb 100644
--- a/sys/arch/hp300/hp300/disksubr.c
+++ b/sys/arch/hp300/hp300/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.43 2008/06/11 12:35:41 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.44 2008/06/12 06:58:33 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.9 1997/04/01 03:12:13 scottr Exp $	*/
 
 /*
@@ -68,7 +68,7 @@ readdisklabel(dev_t dev, void (*strat)(struct buf *),
 
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp)) {
 		msg = "disk label I/O error";
@@ -115,7 +115,7 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	bp->b_dev = dev;
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
@@ -123,7 +123,7 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Write it in the regular place. */
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/hppa/hppa/disksubr.c b/sys/arch/hppa/hppa/disksubr.c
index cac0bc04b5f..8c46091d841 100644
--- a/sys/arch/hppa/hppa/disksubr.c
+++ b/sys/arch/hppa/hppa/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.71 2008/06/11 12:35:41 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.72 2008/06/12 06:58:34 deraadt Exp $	*/
 
 /*
  * Copyright (c) 1999 Michael Shalayeff
@@ -107,7 +107,7 @@ readliflabel(struct buf *bp, void (*strat)(struct buf *),
 	/* read LIF volume header */
 	bp->b_blkno = btodb(LIF_VOLSTART);
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp))
 		return "LIF volume header I/O error";
@@ -122,7 +122,7 @@ readliflabel(struct buf *bp, void (*strat)(struct buf *),
 	/* read LIF directory */
 	dbp->b_blkno = lifstodb(lvp->vol_addr);
 	dbp->b_bcount = lp->d_secsize;
-	dbp->b_flags = B_BUSY | B_READ;
+	dbp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(dbp);
 
 	if (biowait(dbp)) {
@@ -154,7 +154,7 @@ readliflabel(struct buf *bp, void (*strat)(struct buf *),
 		/* read LIF directory */
 		dbp->b_blkno = lifstodb(p->dir_addr);
 		dbp->b_bcount = lp->d_secsize;
-		dbp->b_flags = B_BUSY | B_READ;
+		dbp->b_flags = B_BUSY | B_READ | B_RAW;
 		(*strat)(dbp);
 
 		if (biowait(dbp)) {
@@ -218,7 +218,7 @@ finished:
 
 	bp->b_blkno = fsoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 
 	/* if successful, locate disk label within block and validate */
@@ -258,14 +258,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = partoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/hppa64/hppa64/disksubr.c b/sys/arch/hppa64/hppa64/disksubr.c
index 8b72b8b47e2..31580f0c40f 100644
--- a/sys/arch/hppa64/hppa64/disksubr.c
+++ b/sys/arch/hppa64/hppa64/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.55 2008/06/11 12:35:41 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.56 2008/06/12 06:58:34 deraadt Exp $	*/
 
 /*
  * Copyright (c) 1999 Michael Shalayeff
@@ -107,7 +107,7 @@ readliflabel(struct buf *bp, void (*strat)(struct buf *),
 	/* read LIF volume header */
 	bp->b_blkno = btodb(LIF_VOLSTART);
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp))
 		return "LIF volume header I/O error";
@@ -122,7 +122,7 @@ readliflabel(struct buf *bp, void (*strat)(struct buf *),
 	/* read LIF directory */
 	dbp->b_blkno = lifstodb(lvp->vol_addr);
 	dbp->b_bcount = lp->d_secsize;
-	dbp->b_flags = B_BUSY | B_READ;
+	dbp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(dbp);
 
 	if (biowait(dbp)) {
@@ -154,7 +154,7 @@ readliflabel(struct buf *bp, void (*strat)(struct buf *),
 		/* read LIF directory */
 		dbp->b_blkno = lifstodb(p->dir_addr);
 		dbp->b_bcount = lp->d_secsize;
-		dbp->b_flags = B_BUSY | B_READ;
+		dbp->b_flags = B_BUSY | B_READ | B_RAW;
 		(*strat)(dbp);
 
 		if (biowait(dbp)) {
@@ -218,7 +218,7 @@ finished:
 
 	bp->b_blkno = fsoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 
 	/* if successful, locate disk label within block and validate */
@@ -258,14 +258,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = partoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/i386/i386/disksubr.c b/sys/arch/i386/i386/disksubr.c
index 056b495327b..35ec17d595a 100644
--- a/sys/arch/i386/i386/disksubr.c
+++ b/sys/arch/i386/i386/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.96 2008/06/11 12:35:41 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.97 2008/06/12 06:58:34 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.21 1996/05/03 19:42:03 christos Exp $	*/
 
 /*
@@ -141,14 +141,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = partoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/i386/i386/dkcsum.c b/sys/arch/i386/i386/dkcsum.c
index 0305da7e7eb..16c06fb8b76 100644
--- a/sys/arch/i386/i386/dkcsum.c
+++ b/sys/arch/i386/i386/dkcsum.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: dkcsum.c,v 1.24 2008/06/11 12:35:41 deraadt Exp $	*/
+/*	$OpenBSD: dkcsum.c,v 1.25 2008/06/12 06:58:34 deraadt Exp $	*/
 
 /*-
  * Copyright (c) 1997 Niklas Hallqvist.  All rights reserved.
@@ -106,7 +106,7 @@ dkcsumattach(void)
 		/* Read blocks to cksum.  XXX maybe a d_read should be used. */
 		bp->b_blkno = 0;
 		bp->b_bcount = bios_cksumlen * DEV_BSIZE;
-		bp->b_flags = B_BUSY | B_READ;
+		bp->b_flags = B_BUSY | B_READ | B_RAW;
 		bp->b_cylinder = 0;
 		(*bdsw->d_strategy)(bp);
 		if ((error = biowait(bp))) {
diff --git a/sys/arch/landisk/landisk/disksubr.c b/sys/arch/landisk/landisk/disksubr.c
index 4e5a626b540..f32086e2306 100644
--- a/sys/arch/landisk/landisk/disksubr.c
+++ b/sys/arch/landisk/landisk/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.38 2008/06/11 12:35:42 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.39 2008/06/12 06:58:35 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.21 1996/05/03 19:42:03 christos Exp $	*/
 
 /*
@@ -114,14 +114,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = partoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/luna88k/luna88k/disksubr.c b/sys/arch/luna88k/luna88k/disksubr.c
index 8f513a3263c..f95dd05180f 100644
--- a/sys/arch/luna88k/luna88k/disksubr.c
+++ b/sys/arch/luna88k/luna88k/disksubr.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: disksubr.c,v 1.35 2008/06/11 12:35:42 deraadt Exp $ */
+/* $OpenBSD: disksubr.c,v 1.36 2008/06/12 06:58:35 deraadt Exp $ */
 /* $NetBSD: disksubr.c,v 1.12 2002/02/19 17:09:44 wiz Exp $ */
 
 /*
@@ -124,7 +124,7 @@ readdisklabel(dev_t dev, void (*strat)(struct buf *),
 	bp->b_blkno = LABELSECTOR;
 	bp->b_cylinder = 0;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp)) {
 		msg = "disk label read error";
@@ -179,7 +179,7 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	bp->b_blkno = LABELSECTOR;
 	bp->b_cylinder = 0;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 
 	(*strat)(bp);
 	error = biowait(bp);
@@ -191,7 +191,7 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	if (error)
 		goto done;
 
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/mac68k/mac68k/disksubr.c b/sys/arch/mac68k/mac68k/disksubr.c
index 60913ad8c13..da0ebc88ffc 100644
--- a/sys/arch/mac68k/mac68k/disksubr.c
+++ b/sys/arch/mac68k/mac68k/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.54 2008/06/11 12:35:42 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.55 2008/06/12 06:58:35 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.22 1997/11/26 04:18:20 briggs Exp $	*/
 
 /*
@@ -349,7 +349,7 @@ readdisklabel(dev_t dev, void (*strat)(struct buf *),
 
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = size;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp)) {
 		msg = "disk label I/O error";
@@ -366,7 +366,7 @@ readdisklabel(dev_t dev, void (*strat)(struct buf *),
 	/* Get a MI label */
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp)) {
 		msg = "disk label I/O error";
@@ -423,7 +423,7 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
@@ -437,7 +437,7 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/macppc/macppc/disksubr.c b/sys/arch/macppc/macppc/disksubr.c
index 3d0cddf00da..a43f121f2b2 100644
--- a/sys/arch/macppc/macppc/disksubr.c
+++ b/sys/arch/macppc/macppc/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.59 2008/06/11 12:35:42 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.60 2008/06/12 06:58:36 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.21 1996/05/03 19:42:03 christos Exp $	*/
 
 /*
@@ -111,7 +111,7 @@ readdpmelabel(struct buf *bp, void (*strat)(struct buf *),
 	/* First check for a DPME (HFS) disklabel */
 	bp->b_blkno = 1;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp))
 		return ("DPME partition I/O error");
@@ -129,7 +129,7 @@ readdpmelabel(struct buf *bp, void (*strat)(struct buf *),
 
 		bp->b_blkno = 1+i;
 		bp->b_bcount = lp->d_secsize;
-		bp->b_flags = B_BUSY | B_READ;
+		bp->b_flags = B_BUSY | B_READ | B_RAW;
 		(*strat)(bp);
 		if (biowait(bp))
 			return ("DPME partition I/O error");
@@ -165,7 +165,7 @@ readdpmelabel(struct buf *bp, void (*strat)(struct buf *),
 	/* next, dig out disk label */
 	bp->b_blkno = hfspartoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp))
 		return("disk label I/O error");
@@ -194,14 +194,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = partoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/mvme68k/mvme68k/disksubr.c b/sys/arch/mvme68k/mvme68k/disksubr.c
index bf4dc36761f..52dc3c1f520 100644
--- a/sys/arch/mvme68k/mvme68k/disksubr.c
+++ b/sys/arch/mvme68k/mvme68k/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.62 2008/06/11 12:35:42 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.63 2008/06/12 06:58:36 deraadt Exp $	*/
 /*
  * Copyright (c) 1998 Steve Murphree, Jr.
  * Copyright (c) 1995 Dale Rahn.
@@ -67,7 +67,7 @@ readdisklabel(dev_t dev, void (*strat)(struct buf *),
 
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 	if (error) {
@@ -122,14 +122,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	bsdtocpulabel(lp, (struct mvmedisklabel *)bp->b_data);
 
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/mvme88k/mvme88k/disksubr.c b/sys/arch/mvme88k/mvme88k/disksubr.c
index 0d8477b0a13..07bfbf6ff33 100644
--- a/sys/arch/mvme88k/mvme88k/disksubr.c
+++ b/sys/arch/mvme88k/mvme88k/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.58 2008/06/11 12:35:43 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.59 2008/06/12 06:58:36 deraadt Exp $	*/
 /*
  * Copyright (c) 1998 Steve Murphree, Jr.
  * Copyright (c) 1995 Dale Rahn.
@@ -67,7 +67,7 @@ readdisklabel(dev_t dev, void (*strat)(struct buf *),
 
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 	if (error) {
@@ -123,14 +123,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	bsdtocpulabel(lp, (struct mvmedisklabel *)bp->b_data);
 
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/mvmeppc/mvmeppc/disksubr.c b/sys/arch/mvmeppc/mvmeppc/disksubr.c
index 16cf8c0a2ce..6db60e52d88 100644
--- a/sys/arch/mvmeppc/mvmeppc/disksubr.c
+++ b/sys/arch/mvmeppc/mvmeppc/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.56 2008/06/11 12:35:43 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.57 2008/06/12 06:58:36 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.21 1996/05/03 19:42:03 christos Exp $	*/
 
 /*
@@ -114,14 +114,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = partoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/sgi/sgi/disksubr.c b/sys/arch/sgi/sgi/disksubr.c
index ef7ef682b6f..8d3ad6c0e37 100644
--- a/sys/arch/sgi/sgi/disksubr.c
+++ b/sys/arch/sgi/sgi/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.5 2008/06/11 12:35:43 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.6 2008/06/12 06:58:36 deraadt Exp $	*/
 
 /*
  * Copyright (c) 1999 Michael Shalayeff
@@ -59,7 +59,7 @@ readbsdlabel(struct buf *bp, void (*strat)(struct buf *),
 
 	bp->b_blkno = sec;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 
 	/* if successful, locate disk label within block and validate */
@@ -145,7 +145,7 @@ readsgilabel(struct buf *bp, void (*strat)(struct buf *),
 
 	bp->b_blkno = 0;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 
 	/* if successful, locate disk label within block and validate */
@@ -215,7 +215,7 @@ finished:
 
 	bp->b_blkno = fsoffs + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp)) {
 		msg = "disk label I/O error";
@@ -249,14 +249,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = partoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/socppc/socppc/disksubr.c b/sys/arch/socppc/socppc/disksubr.c
index eee93504473..3a1ee2cdb89 100644
--- a/sys/arch/socppc/socppc/disksubr.c
+++ b/sys/arch/socppc/socppc/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.3 2008/06/11 12:35:43 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.4 2008/06/12 06:58:36 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.21 1996/05/03 19:42:03 christos Exp $	*/
 
 /*
@@ -111,7 +111,7 @@ readdpmelabel(struct buf *bp, void (*strat)(struct buf *),
 	/* First check for a DPME (HFS) disklabel */
 	bp->b_blkno = 1;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp))
 		return ("DPME partition I/O error");
@@ -129,7 +129,7 @@ readdpmelabel(struct buf *bp, void (*strat)(struct buf *),
 
 		bp->b_blkno = 1+i;
 		bp->b_bcount = lp->d_secsize;
-		bp->b_flags = B_BUSY | B_READ;
+		bp->b_flags = B_BUSY | B_READ | B_RAW;
 		(*strat)(bp);
 		if (biowait(bp))
 			return ("DPME partition I/O error");
@@ -165,7 +165,7 @@ readdpmelabel(struct buf *bp, void (*strat)(struct buf *),
 	/* next, dig out disk label */
 	bp->b_blkno = hfspartoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp))
 		return("disk label I/O error");
@@ -194,14 +194,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = partoff + LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/sparc/dev/fd.c b/sys/arch/sparc/dev/fd.c
index ae18fc2933b..5a03717fcb9 100644
--- a/sys/arch/sparc/dev/fd.c
+++ b/sys/arch/sparc/dev/fd.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: fd.c,v 1.60 2008/06/11 12:35:43 deraadt Exp $	*/
+/*	$OpenBSD: fd.c,v 1.61 2008/06/12 06:58:37 deraadt Exp $	*/
 /*	$NetBSD: fd.c,v 1.51 1997/05/24 20:16:19 pk Exp $	*/
 
 /*-
@@ -1911,7 +1911,7 @@ fdformat(dev, finfo, p)
 	if (bp == 0)
 		return (ENOBUFS);
 
-	bp->b_flags = B_BUSY | B_PHYS | B_FORMAT;
+	bp->b_flags = B_BUSY | B_PHYS | B_FORMAT | B_RAW;
 	bp->b_proc = p;
 	bp->b_dev = dev;
 
diff --git a/sys/arch/sparc/sparc/disksubr.c b/sys/arch/sparc/sparc/disksubr.c
index e5f6d531917..51fa9e50246 100644
--- a/sys/arch/sparc/sparc/disksubr.c
+++ b/sys/arch/sparc/sparc/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.74 2008/06/11 12:35:43 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.75 2008/06/12 06:58:37 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.16 1996/04/28 20:25:59 thorpej Exp $ */
 
 /*
@@ -105,7 +105,7 @@ readdisklabel(dev_t dev, void (*strat)(struct buf *),
 
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp)) {
 		msg = "disk label read error";
@@ -169,7 +169,7 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Write out the updated label. */
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/sparc64/dev/fd.c b/sys/arch/sparc64/dev/fd.c
index 8011833f26e..ad86107fc03 100644
--- a/sys/arch/sparc64/dev/fd.c
+++ b/sys/arch/sparc64/dev/fd.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: fd.c,v 1.23 2008/06/11 12:35:44 deraadt Exp $	*/
+/*	$OpenBSD: fd.c,v 1.24 2008/06/12 06:58:37 deraadt Exp $	*/
 /*	$NetBSD: fd.c,v 1.112 2003/08/07 16:29:35 agc Exp $	*/
 
 /*-
@@ -1965,7 +1965,7 @@ fdformat(dev, finfo, p)
 	if (bp == NULL)
 		return (ENOBUFS);
 
-	bp->b_flags = B_BUSY | B_PHYS | B_FORMAT;
+	bp->b_flags = B_BUSY | B_PHYS | B_FORMAT | B_RAW;
 	bp->b_proc = p;
 	bp->b_dev = dev;
 
diff --git a/sys/arch/sparc64/sparc64/disksubr.c b/sys/arch/sparc64/sparc64/disksubr.c
index 8e45c71930c..8796a0831c3 100644
--- a/sys/arch/sparc64/sparc64/disksubr.c
+++ b/sys/arch/sparc64/sparc64/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.50 2008/06/11 12:35:44 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.51 2008/06/12 06:58:37 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.13 2000/12/17 22:39:18 pk Exp $ */
 
 /*
@@ -98,7 +98,7 @@ readdisklabel(dev_t dev, void (*strat)(struct buf *),
 
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp)) {
 		msg = "disk label read error";
@@ -162,7 +162,7 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Write out the updated label. */
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/arch/vax/uba/ts.c b/sys/arch/vax/uba/ts.c
index 8b6160e9281..9d443e6659d 100644
--- a/sys/arch/vax/uba/ts.c
+++ b/sys/arch/vax/uba/ts.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ts.c,v 1.19 2008/06/11 12:35:44 deraadt Exp $ */
+/*	$OpenBSD: ts.c,v 1.20 2008/06/12 06:58:37 deraadt Exp $ */
 /*	$NetBSD: ts.c,v 1.11 1997/01/11 11:34:43 ragge Exp $ */
 
 /*-
@@ -379,7 +379,7 @@ tscommand (dev, cmd, count)
 		tsleep ((caddr_t)bp, PRIBIO, "tscommand", 0);
 		/* check MOT-flag !!! */
 	}
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 
 	splx(s);
 
diff --git a/sys/arch/vax/vax/disksubr.c b/sys/arch/vax/vax/disksubr.c
index 8b9c7f63693..9f30ae69a6b 100644
--- a/sys/arch/vax/vax/disksubr.c
+++ b/sys/arch/vax/vax/disksubr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: disksubr.c,v 1.56 2008/06/11 12:35:44 deraadt Exp $	*/
+/*	$OpenBSD: disksubr.c,v 1.57 2008/06/12 06:58:38 deraadt Exp $	*/
 /*	$NetBSD: disksubr.c,v 1.21 1999/06/30 18:48:06 ragge Exp $	*/
 
 /*
@@ -76,7 +76,7 @@ readdisklabel(dev_t dev, void (*strat)(struct buf *),
 
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp)) {
 		msg = "I/O error";
@@ -125,14 +125,14 @@ writedisklabel(dev_t dev, void (*strat)(struct buf *), struct disklabel *lp)
 	/* Read it in, slap the new label in, and write it back out */
 	bp->b_blkno = LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if ((error = biowait(bp)) != 0)
 		goto done;
 
 	dlp = (struct disklabel *)(bp->b_data + LABELOFFSET);
 	*dlp = *lp;
-	bp->b_flags = B_BUSY | B_WRITE;
+	bp->b_flags = B_BUSY | B_WRITE | B_RAW;
 	(*strat)(bp);
 	error = biowait(bp);
 
diff --git a/sys/conf/files b/sys/conf/files
index 5e3a667a38f..c841b5267ce 100644
--- a/sys/conf/files
+++ b/sys/conf/files
@@ -1,4 +1,4 @@
-#	$OpenBSD: files,v 1.434 2008/06/11 12:35:44 deraadt Exp $
+#	$OpenBSD: files,v 1.435 2008/06/12 06:58:38 deraadt Exp $
 #	$NetBSD: files,v 1.87 1996/05/19 17:17:50 jonathan Exp $
 
 #	@(#)files.newconf	7.5 (Berkeley) 5/10/93
@@ -707,6 +707,7 @@ file kern/uipc_socket2.c
 file kern/uipc_syscalls.c
 file kern/uipc_usrreq.c
 file kern/vfs_bio.c
+file kern/vfs_biomem.c
 file kern/vfs_cache.c
 file kern/vfs_cluster.c
 file kern/vfs_conf.c
diff --git a/sys/dev/isa/fd.c b/sys/dev/isa/fd.c
index 1c103718d3f..528a0daf223 100644
--- a/sys/dev/isa/fd.c
+++ b/sys/dev/isa/fd.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: fd.c,v 1.73 2008/06/11 12:35:45 deraadt Exp $	*/
+/*	$OpenBSD: fd.c,v 1.74 2008/06/12 06:58:39 deraadt Exp $	*/
 /*	$NetBSD: fd.c,v 1.90 1996/05/12 23:12:03 mycroft Exp $	*/
 
 /*-
@@ -1125,7 +1125,7 @@ fdformat(dev, finfo, p)
         if (bp == NULL)
                 return ENOBUFS;
 
-        bp->b_flags = B_BUSY | B_PHYS | B_FORMAT;
+        bp->b_flags = B_BUSY | B_PHYS | B_FORMAT | B_RAW;
         bp->b_proc = p;
         bp->b_dev = dev;
 
diff --git a/sys/isofs/udf/udf_subr.c b/sys/isofs/udf/udf_subr.c
index 87ef5367e74..3ffa57e28eb 100644
--- a/sys/isofs/udf/udf_subr.c
+++ b/sys/isofs/udf/udf_subr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: udf_subr.c,v 1.16 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: udf_subr.c,v 1.17 2008/06/12 06:58:39 deraadt Exp $	*/
 
 /*
  * Copyright (c) 2006, Miodrag Vallat
@@ -110,7 +110,7 @@ udf_disklabelspoof(dev_t dev, void (*strat)(struct buf *),
 	 */
 	bp->b_blkno = sector * btodb(bsize);
 	bp->b_bcount = bsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags |= B_READ;
 	bp->b_resid = bp->b_blkno / lp->d_secpercyl;
 
 	(*strat)(bp);
@@ -130,7 +130,7 @@ udf_disklabelspoof(dev_t dev, void (*strat)(struct buf *),
 	for (sector = mvds_start; sector < mvds_end; sector++) {
 		bp->b_blkno = sector * btodb(bsize);
 		bp->b_bcount = bsize;
-		bp->b_flags = B_BUSY | B_READ;
+		bp->b_flags |= B_READ;
 		bp->b_resid = bp->b_blkno / lp->d_secpercyl;
 
 		(*strat)(bp);
diff --git a/sys/kern/kern_sched.c b/sys/kern/kern_sched.c
index 1aa25f464a9..51b4c59f8ad 100644
--- a/sys/kern/kern_sched.c
+++ b/sys/kern/kern_sched.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: kern_sched.c,v 1.5 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: kern_sched.c,v 1.6 2008/06/12 06:58:39 deraadt Exp $	*/
 /*
  * Copyright (c) 2007 Artur Grabowski <art@openbsd.org>
  *
@@ -113,6 +113,8 @@ sched_idle(void *v)
 			}
 		}
 
+		splassert(IPL_NONE);
+
 		cpu_idle_enter();
 		while (sched_is_idle())
 			cpu_idle_cycle();
diff --git a/sys/kern/spec_vnops.c b/sys/kern/spec_vnops.c
index 15c2bd86377..b9394598065 100644
--- a/sys/kern/spec_vnops.c
+++ b/sys/kern/spec_vnops.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: spec_vnops.c,v 1.51 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: spec_vnops.c,v 1.52 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: spec_vnops.c,v 1.29 1996/04/22 01:42:38 christos Exp $	*/
 
 /*
@@ -447,7 +447,7 @@ loop:
 		if ((bp->b_flags & B_DELWRI) == 0)
 			panic("spec_fsync: not dirty");
 		bremfree(bp);
-		bp->b_flags |= B_BUSY;
+		buf_acquire(bp);
 		splx(s);
 		bawrite(bp);
 		goto loop;
diff --git a/sys/kern/subr_disk.c b/sys/kern/subr_disk.c
index d4db8e2e299..e3d80ae69ff 100644
--- a/sys/kern/subr_disk.c
+++ b/sys/kern/subr_disk.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: subr_disk.c,v 1.77 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: subr_disk.c,v 1.78 2008/06/12 06:58:39 deraadt Exp $	*/
 /*	$NetBSD: subr_disk.c,v 1.17 1996/03/16 23:17:08 christos Exp $	*/
 
 /*
@@ -399,7 +399,7 @@ readdoslabel(struct buf *bp, void (*strat)(struct buf *),
 		/* read boot record */
 		bp->b_blkno = part_blkno;
 		bp->b_bcount = lp->d_secsize;
-		bp->b_flags = B_BUSY | B_READ;
+		bp->b_flags = B_BUSY | B_READ | B_RAW;
 		(*strat)(bp);
 		if (biowait(bp)) {
 /*wrong*/		if (partoffp)
@@ -542,7 +542,7 @@ notfat:
 
 	bp->b_blkno = dospartoff + DOS_LABELSECTOR;
 	bp->b_bcount = lp->d_secsize;
-	bp->b_flags = B_BUSY | B_READ;
+	bp->b_flags = B_BUSY | B_READ | B_RAW;
 	(*strat)(bp);
 	if (biowait(bp))
 		return ("disk label I/O error");
diff --git a/sys/kern/vfs_bio.c b/sys/kern/vfs_bio.c
index 11c450fc982..ee1b43b0221 100644
--- a/sys/kern/vfs_bio.c
+++ b/sys/kern/vfs_bio.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: vfs_bio.c,v 1.105 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: vfs_bio.c,v 1.106 2008/06/12 06:58:39 deraadt Exp $	*/
 /*	$NetBSD: vfs_bio.c,v 1.44 1996/06/11 11:15:36 pk Exp $	*/
 
 /*-
@@ -78,14 +78,12 @@ u_long	bufhash;
 /*
  * Definitions for the buffer free lists.
  */
-#define	BQUEUES		6		/* number of free buffer queues */
+#define	BQUEUES		2		/* number of free buffer queues */
 
 #define	BQ_DIRTY	0		/* LRU queue with dirty buffers */
-
+#define	BQ_CLEAN	1		/* LRU queue with clean buffers */
 
 TAILQ_HEAD(bqueues, buf) bufqueues[BQUEUES];
-int bqpages[BQUEUES];		/* pages allocated, per queue */
-int bqpagelow;
 int needbuffer;
 struct bio_ops bioops;
 
@@ -93,7 +91,6 @@ struct bio_ops bioops;
  * Buffer pool for I/O buffers.
  */
 struct pool bufpool;
-struct vm_map *buf_map;
 struct bufhead bufhead = LIST_HEAD_INITIALIZER(bufhead);
 struct buf *buf_get(size_t);
 struct buf *buf_stub(struct vnode *, daddr64_t);
@@ -107,7 +104,7 @@ void buf_put(struct buf *);
 
 struct buf *bio_doread(struct vnode *, daddr64_t, int, int);
 struct buf *getnewbuf(size_t, int, int, int *);
-void buf_init(struct buf *, int);
+void buf_init(struct buf *);
 void bread_cluster_callback(struct buf *);
 
 /*
@@ -117,54 +114,33 @@ void bread_cluster_callback(struct buf *);
  *  numdirtypages - number of pages on BQ_DIRTY queue.
  *  lodirtypages  - low water mark for buffer cleaning daemon.
  *  hidirtypages  - high water mark for buffer cleaning daemon.
- *  numfreepages  - number of pages on BQ_CLEAN and BQ_DIRTY queues. unused.
  *  numcleanpages - number of pages on BQ_CLEAN queue.
  *		    Used to track the need to speedup the cleaner and 
  *		    as a reserve for special processes like syncer.
  *  maxcleanpages - the highest page count on BQ_CLEAN.
  */
-long numbufpages;
-long numdirtypages;
+
+struct bcachestats bcstats;
 long lodirtypages;
 long hidirtypages;
-long numfreepages;
-long numcleanpages;
 long locleanpages;
 long hicleanpages;
 long maxcleanpages;
 
-struct proc *cleanerproc;
-int bd_req;			/* Sleep point for cleaner daemon. */
+/* XXX - should be defined here. */
+extern int bufcachepercent;
 
-int size2cqueue(int *size);
+vsize_t bufkvm;
 
-int
-size2cqueue(int *size)
-{
-	int i = 0, q;
-	int s = *size;
-	s -= 1;
-	while (s > 0) {
-		s = s >> 1;
-		i++;
-	}
-	if (i < PAGE_SHIFT) {
-		i = PAGE_SHIFT; /* < 4096 -> 4096 */
-        }
-	*size = 1 << i;
-	q = (i + 1 - PAGE_SHIFT); /* XXX 4096 is queue 1 */
-	if (q >= BQUEUES)
-		panic("queue %d > BQUEUES %d", q, BQUEUES);
-	if (q == 0)
-		panic("can't return dirty q");
-	return(q);
-}
+struct proc *cleanerproc;
+int bd_req;			/* Sleep point for cleaner daemon. */
 
 void
 bremfree(struct buf *bp)
 {
 	struct bqueues *dp = NULL;
-	int queue;
+
+	splassert(IPL_BIO);
 
 	/*
 	 * We only calculate the head of the freelist when removing
@@ -180,38 +156,26 @@ bremfree(struct buf *bp)
 		if (dp == &bufqueues[BQUEUES])
 			panic("bremfree: lost tail");
 	}
-	numfreepages -= atop(bp->b_bufsize);
 	if (!ISSET(bp->b_flags, B_DELWRI)) {
-		int qs = bp->b_bufsize;
-		queue = size2cqueue(&qs);
-		numcleanpages -= atop(bp->b_bufsize);
-		bqpages[queue] -= atop(bp->b_bufsize);
-	} else
-		numdirtypages -= atop(bp->b_bufsize);
+		bcstats.numcleanpages -= atop(bp->b_bufsize);
+	} else {
+		bcstats.numdirtypages -= atop(bp->b_bufsize);
+	}
 	TAILQ_REMOVE(dp, bp, b_freelist);
+	bcstats.freebufs--;
 }
 
 void
-buf_init(struct buf *bp, int size)
+buf_init(struct buf *bp)
 {
-	int npages, queue;
-
 	splassert(IPL_BIO);
 
-	npages = atop(size);
 	bzero((char *)bp, sizeof *bp);
 	bp->b_vnbufs.le_next = NOLIST;
 	bp->b_freelist.tqe_next = NOLIST;
 	bp->b_synctime = time_uptime + 300;
 	bp->b_dev = NODEV;
-	queue = size2cqueue(&size);
 	LIST_INIT(&bp->b_dep);
-	numbufpages += npages;
-	numfreepages += npages;
-	numcleanpages += npages;
-	bqpages[queue] += npages;
-	if (maxcleanpages < numcleanpages)
-		maxcleanpages = numcleanpages;
 }
 
 /*
@@ -238,7 +202,7 @@ buf_stub(struct vnode *vp, daddr64_t lblkno)
 	bp->b_dev = NODEV;
 	bp->b_bufsize = 0;
 	bp->b_data = NULL;
-	bp->b_flags = B_BUSY;
+	bp->b_flags = 0;
 	bp->b_dev = NODEV;
 	bp->b_blkno = bp->b_lblkno = lblkno;
 	bp->b_iodone = NULL;
@@ -250,8 +214,11 @@ buf_stub(struct vnode *vp, daddr64_t lblkno)
 
 	LIST_INIT(&bp->b_dep);
 
+	buf_acquire_unmapped(bp);
+
 	s = splbio();
 	LIST_INSERT_HEAD(&bufhead, bp, b_list);
+	bcstats.numbufs++;
 	bgetvp(vp, bp);
 	splx(s);
 
@@ -261,39 +228,31 @@ buf_stub(struct vnode *vp, daddr64_t lblkno)
 struct buf *
 buf_get(size_t size)
 {
-	struct bqueues *dp;
 	struct buf *bp;
 	int npages;
-	int queue, qs;
-	void *data;
 
 	splassert(IPL_BIO);
 
 	KASSERT(size > 0);
 
 	size = round_page(size);
-	qs = size;
-	queue = size2cqueue(&qs);
-	npages = atop(qs);
+	npages = atop(size);
 
-	if (numbufpages + npages > bufpages)
+	if (bcstats.numbufpages + npages > bufpages)
 		return (NULL);
 
 	bp = pool_get(&bufpool, PR_WAITOK);
 
-	data = (void *)uvm_km_alloc(buf_map, qs);
-	if (data == NULL) {
-		pool_put(&bufpool, bp);
-		return (NULL);
-	}
-	buf_init(bp, qs);
+	buf_init(bp);
 	bp->b_flags = B_INVAL;
-	bp->b_bufsize = qs;
-	bp->b_data = data;
-	dp = &bufqueues[queue];
-	binsheadfree(bp, dp);
+	buf_alloc_pages(bp, size);
+	bp->b_data = NULL;
+	binsheadfree(bp, &bufqueues[BQ_CLEAN]);
 	binshash(bp, &invalhash);
 	LIST_INSERT_HEAD(&bufhead, bp, b_list);
+	bcstats.numbufs++;
+	bcstats.freebufs++;
+	bcstats.numcleanpages += atop(bp->b_bufsize);
 
 	return (bp);
 }
@@ -303,7 +262,7 @@ buf_put(struct buf *bp)
 {
 	splassert(IPL_BIO);
 #ifdef DIAGNOSTIC
-	if (bp->b_data != NULL)
+	if (bp->b_pobj != NULL)
 		KASSERT(bp->b_bufsize > 0);
 #endif
 #ifdef DIAGNOSTIC
@@ -320,13 +279,10 @@ buf_put(struct buf *bp)
 		panic("buf_put: b_dep is not empty");
 #endif
 	LIST_REMOVE(bp, b_list);
+	bcstats.numbufs--;
 
-	if (bp->b_data != NULL) {
-		bremhash(bp);
-		numbufpages -= atop(bp->b_bufsize);
-		uvm_km_free(buf_map, (vaddr_t)bp->b_data, bp->b_bufsize);
-	}
-
+	if (buf_dealloc_mem(bp) != 0)
+		return;
 	pool_put(&bufpool, bp);
 }
 
@@ -336,39 +292,56 @@ buf_put(struct buf *bp)
 void
 bufinit(void)
 {
-	vaddr_t minaddr, maxaddr;
 	struct bqueues *dp;
 
+	/* XXX - for now */
+	bufpages = bufcachepercent = bufkvm = 0;
+
+	/*
+	 * If MD code doesn't say otherwise, use 10% of kvm for mappings and
+	 * 10% physmem for pages.
+	 */
+	if (bufcachepercent == 0)
+		bufcachepercent = 10;
+	if (bufpages == 0)
+		bufpages = physmem * bufcachepercent / 100;
+
+	if (bufkvm == 0)
+		bufkvm = (VM_MAX_KERNEL_ADDRESS - VM_MIN_KERNEL_ADDRESS) / 10;
+
+	/*
+	 * Don't use more than twice the amount of bufpages for mappings.
+	 * It's twice since we map things sparsely.
+	 */
+	if (bufkvm > bufpages * PAGE_SIZE)
+		bufkvm = bufpages * PAGE_SIZE;
+	/*
+	 * Round bufkvm to MAXPHYS because we allocate chunks of va space
+	 * in MAXPHYS chunks.
+	 */
+	bufkvm &= ~(MAXPHYS - 1);
+
 	pool_init(&bufpool, sizeof(struct buf), 0, 0, 0, "bufpl", NULL);
 	pool_setipl(&bufpool, IPL_BIO);
 	for (dp = bufqueues; dp < &bufqueues[BQUEUES]; dp++)
 		TAILQ_INIT(dp);
-	minaddr = vm_map_min(kernel_map);
-	buf_map = uvm_km_suballoc(kernel_map, &minaddr, &maxaddr,
-	    ptoa(bufpages), 0, FALSE, NULL);
 
-	/* 
-	 * XXX don't starve any one queue below 5% of the total number
-	 * of buffer cache pages.
-	 */
-	bqpagelow = bufpages / 20; 
+	/*
+	 * hmm - bufkvm is an argument because it's static, while
+	 * bufpages is global because it can change while running.
+ 	 */
+	buf_mem_init(bufkvm);
 
 	bufhashtbl = hashinit(bufpages / 4, M_CACHE, M_WAITOK, &bufhash);
 	hidirtypages = (bufpages / 4) * 3;
 	lodirtypages = bufpages / 2;
 
 	/*
-	 * Reserve 5% of bufpages for syncer's needs,
-	 * but not more than 25% and if possible
-	 * not less than 2 * MAXBSIZE. locleanpages
-	 * value must be not too small
+	 * When we hit 95% of pages being clean, we bring them down to
+	 * 90% to have some slack.
 	 */
-	hicleanpages = bufpages / 2;
-	locleanpages = hicleanpages / 2;
-	if (locleanpages < atop(2 * MAXBSIZE))
-		locleanpages = atop(2 * MAXBSIZE);
-	if (locleanpages > bufpages / 4)
-		locleanpages = bufpages / 4;
+	hicleanpages = bufpages - (bufpages / 20);
+	locleanpages = bufpages - (bufpages / 10);
 
 	maxcleanpages = locleanpages;
 }
@@ -388,8 +361,9 @@ bio_doread(struct vnode *vp, daddr64_t blkno, int size, int async)
 	 */
 	if (!ISSET(bp->b_flags, (B_DONE | B_DELWRI))) {
 		SET(bp->b_flags, B_READ | async);
+		bcstats.pendingreads++;
+		bcstats.numreads++;
 		VOP_STRATEGY(bp);
-
 		/* Pay for the read. */
 		curproc->p_stats->p_ru.ru_inblock++;		/* XXX */
 	} else if (async) {
@@ -477,7 +451,7 @@ bread_cluster_callback(struct buf *bp)
 	}
 
 	free(xbpp, M_TEMP);
-	bp->b_data = NULL;
+	bp->b_pobj = NULL;
 	buf_put(bp);
 }
 
@@ -485,9 +459,8 @@ int
 bread_cluster(struct vnode *vp, daddr64_t blkno, int size, struct buf **rbpp)
 {
 	struct buf *bp, **xbpp;
-	int howmany, i, maxra, inc;
+	int howmany, maxra, i, inc;
 	daddr64_t sblkno;
-	size_t spill;
 
 	*rbpp = bio_doread(vp, blkno, size, 0);
 
@@ -544,11 +517,16 @@ bread_cluster(struct vnode *vp, daddr64_t blkno, int size, struct buf **rbpp)
 	inc = btodb(size);
 
 	for (i = 0; i < howmany; i++) {
+		bcstats.pendingreads++;
+		bcstats.numreads++;
 		SET(xbpp[i]->b_flags, B_READ | B_ASYNC);
 		binshash(xbpp[i], BUFHASH(vp, xbpp[i]->b_lblkno));
 		xbpp[i]->b_blkno = sblkno + (i * inc);
 		xbpp[i]->b_bufsize = xbpp[i]->b_bcount = size;
-		xbpp[i]->b_data = bp->b_data + (i * size);
+		xbpp[i]->b_data = NULL;
+		xbpp[i]->b_pobj = bp->b_pobj;
+		xbpp[i]->b_poffs = bp->b_poffs + (i * size);
+		buf_acquire_unmapped(xbpp[i]);
 	}
 
 	bp->b_blkno = sblkno;
@@ -557,12 +535,8 @@ bread_cluster(struct vnode *vp, daddr64_t blkno, int size, struct buf **rbpp)
 	bp->b_saveaddr = (void *)xbpp;
 	bp->b_iodone = bread_cluster_callback;
 	bp->b_vp = vp;
-	spill = bp->b_bufsize - bp->b_bcount;
-	if (spill) {
-		uvm_km_free(buf_map, (vaddr_t) bp->b_data + bp->b_bcount,
-		    spill);
-		numbufpages -= atop(spill);
-	}
+	bcstats.pendingreads++;
+	bcstats.numreads++;
 	VOP_STRATEGY(bp);
 	curproc->p_stats->p_ru.ru_inblock++;
 
@@ -609,6 +583,8 @@ bwrite(struct buf *bp)
 		else
 			mp->mnt_stat.f_syncwrites++;
 	}
+	bcstats.pendingwrites++;
+	bcstats.numwrites++;
 
 	wasdelayed = ISSET(bp->b_flags, B_DELWRI);
 	CLR(bp->b_flags, (B_READ | B_DONE | B_ERROR | B_DELWRI));
@@ -723,6 +699,11 @@ buf_dirty(struct buf *bp)
 {
 	splassert(IPL_BIO);
 
+#ifdef DIAGNOSTIC
+	if (!ISSET(bp->b_flags, B_BUSY))
+		panic("Trying to dirty buffer on freelist!");
+#endif
+
 	if (ISSET(bp->b_flags, B_DELWRI) == 0) {
 		SET(bp->b_flags, B_DELWRI);
 		bp->b_synctime = time_uptime + 35;
@@ -738,6 +719,10 @@ buf_undirty(struct buf *bp)
 {
 	splassert(IPL_BIO);
 
+#ifdef DIAGNOSTIC
+	if (!ISSET(bp->b_flags, B_BUSY))
+		panic("Trying to undirty buffer on freelist!");
+#endif
 	if (ISSET(bp->b_flags, B_DELWRI)) {
 		CLR(bp->b_flags, B_DELWRI);
 		reassignbuf(bp);
@@ -769,8 +754,6 @@ brelse(struct buf *bp)
 		SET(bp->b_flags, B_INVAL);
 
 	if (ISSET(bp->b_flags, B_INVAL)) {
-		int queue, qs;
-
 		/*
 		 * If the buffer is invalid, place it in the clean queue, so it
 		 * can be reused.
@@ -789,37 +772,29 @@ brelse(struct buf *bp)
 		 * If the buffer has no associated data, place it back in the
 		 * pool.
 		 */
-		if (bp->b_data == NULL) {
+		if (bp->b_data == NULL && bp->b_pobj == NULL) {
 			buf_put(bp);
 			splx(s);
 			return;
 		}
 
-		qs = bp->b_bufsize;
-		queue = size2cqueue(&qs);
-		numcleanpages += atop(bp->b_bufsize);
-		bqpages[queue] += atop(bp->b_bufsize);
-		if (maxcleanpages < numcleanpages)
-			maxcleanpages = numcleanpages;
-		binsheadfree(bp, &bufqueues[queue]);
+		bcstats.numcleanpages += atop(bp->b_bufsize);
+		if (maxcleanpages < bcstats.numcleanpages)
+			maxcleanpages = bcstats.numcleanpages;
+		binsheadfree(bp, &bufqueues[BQ_CLEAN]);
 	} else {
 		/*
 		 * It has valid data.  Put it on the end of the appropriate
 		 * queue, so that it'll stick around for as long as possible.
 		 */
-		int queue, qs;
-		numfreepages += atop(bp->b_bufsize);
-		qs = bp->b_bufsize;
-		queue = size2cqueue(&qs);
 
 		if (!ISSET(bp->b_flags, B_DELWRI)) {
-			numcleanpages += atop(bp->b_bufsize);
-			bqpages[queue] += atop(bp->b_bufsize);
-			if (maxcleanpages < numcleanpages)
-				maxcleanpages = numcleanpages;
-			bufq = &bufqueues[queue];
+			bcstats.numcleanpages += atop(bp->b_bufsize);
+			if (maxcleanpages < bcstats.numcleanpages)
+				maxcleanpages = bcstats.numcleanpages;
+			bufq = &bufqueues[BQ_CLEAN];
 		} else {
-			numdirtypages += atop(bp->b_bufsize);
+			bcstats.numdirtypages += atop(bp->b_bufsize);
 			bufq = &bufqueues[BQ_DIRTY];
 		}
 		if (ISSET(bp->b_flags, B_AGE)) {
@@ -832,7 +807,9 @@ brelse(struct buf *bp)
 	}
 
 	/* Unlock the buffer. */
-	CLR(bp->b_flags, (B_AGE | B_ASYNC | B_BUSY | B_NOCACHE | B_DEFERRED));
+	bcstats.freebufs++;
+	CLR(bp->b_flags, (B_AGE | B_ASYNC | B_NOCACHE | B_DEFERRED));
+	buf_release(bp);
 
 	/* Wake up any processes waiting for any buffer to become free. */
 	if (needbuffer) {
@@ -917,8 +894,10 @@ start:
 		}
 
 		if (!ISSET(bp->b_flags, B_INVAL)) {
-			SET(bp->b_flags, (B_BUSY | B_CACHE));
+			bcstats.cachehits++;
 			bremfree(bp);
+			SET(bp->b_flags, B_CACHE);
+			buf_acquire(bp);
 			splx(s);
 			break;
 		}
@@ -946,6 +925,10 @@ start:
 		bgetvp(vp, bp);
 		splx(s);
 	}
+#ifdef DIAGNOSTIC
+	if (!ISSET(bp->b_flags, B_BUSY))
+		panic("getblk buffer not B_BUSY");
+#endif
 	return (bp);
 }
 
@@ -972,7 +955,7 @@ struct buf *
 getnewbuf(size_t size, int slpflag, int slptimeo, int *ep)
 {
 	struct buf *bp;
-	int s, error, queue, qs;
+	int s;
 
 #if 0		/* we would really like this but sblock update kills it */
 	KASSERT(curproc != syncerproc && curproc != cleanerproc);
@@ -982,72 +965,47 @@ getnewbuf(size_t size, int slpflag, int slptimeo, int *ep)
 	/*
 	 * Wake up cleaner if we're getting low on pages.
 	 */
-	if (numdirtypages >= hidirtypages || numcleanpages <= locleanpages)
+	if (bcstats.numdirtypages >= hidirtypages || bcstats.numcleanpages <= locleanpages)
 		wakeup(&bd_req);
 
+	/*
+	 * If we're above the high water mark for clean pages,
+	 * free down to the low water mark.
+	 */
+	if (bcstats.numcleanpages > hicleanpages) {
+		while (bcstats.numcleanpages > locleanpages) {
+			bp = TAILQ_FIRST(&bufqueues[BQ_CLEAN]);
+			bremfree(bp);
+			if (bp->b_vp)
+				brelvp(bp);
+			bremhash(bp);
+			buf_put(bp);
+		}
+	}
+
 	/* we just ask. it can say no.. */
 getsome:
-	qs = size;
-	queue = size2cqueue(&qs);
-	bp = buf_get(qs); /* XXX use qs instead and no need in buf_get? */
+	bp = buf_get(size);
 	if (bp == NULL) {
-		/*
-		 * No free ones, try to reuse a clean one of the same or
-		 * larger size.
-		 */
-		do {
-			bp = TAILQ_FIRST(&bufqueues[queue]);
-			queue++;
-		} while (bp == NULL && queue < BQUEUES);
-	}
-	if (bp == NULL) {
-		/* we couldn't reuse a free one, nothing of the right size */
-		/* XXX free 20 buffers per q - ugly hack  should really
-		 * reuse big ones without truncating. fix later 
-		 */
-		int q, gotsome = 0;
-		int freemax = 20; 
-		for (q = 1; q < BQUEUES; q++) {
-			int i = freemax;
-			while (bqpages[q] > bqpagelow
-			    && (bp = TAILQ_FIRST(&bufqueues[q]))
-			    && i--) {
-				gotsome++;
-				bremfree(bp);
-				if (LIST_FIRST(&bp->b_dep) != NULL)
-					buf_deallocate(bp);
-
-				if (ISSET(bp->b_flags, B_DELWRI)) {
-					CLR(bp->b_flags, B_DELWRI);
-				}
-
-				if (bp->b_vp)
-					brelvp(bp);
-
-				buf_put(bp);
-			}
+		int freemax = 5;
+		int i = freemax;
+		while ((bp = TAILQ_FIRST(&bufqueues[BQ_CLEAN])) && i--) {
+			bremfree(bp);
+			if (bp->b_vp)
+				brelvp(bp);
+			bremhash(bp);
+			buf_put(bp);
 		}
-		if (gotsome)
+		if (freemax != i)
 			goto getsome;
-	}
-	if (bp == NULL) {
-		/* wait for a free buffer of any kind */
-		needbuffer++;
-		error = tsleep(&needbuffer, slpflag | (PRIBIO + 1),
-		    "getnewbuf", slptimeo);
-		if (ep != NULL) {
-			*ep = error;
-			if (error) {
-				splx(s);
-				return (NULL);
-			}
-		}
-		goto getsome;
+		splx(s);
+		return (NULL);
 	}
 
 	bremfree(bp);
 	/* Buffer is no longer on free lists. */
-	SET(bp->b_flags, B_BUSY);
+	bp->b_flags = 0;
+	buf_acquire(bp);
 
 #ifdef DIAGNOSTIC
 	if (ISSET(bp->b_flags, B_DELWRI))
@@ -1067,7 +1025,6 @@ getsome:
 #endif
 
 	/* clear out various other fields */
-	bp->b_flags = B_BUSY;
 	bp->b_dev = NODEV;
 	bp->b_blkno = bp->b_lblkno = 0;
 	bp->b_iodone = NULL;
@@ -1095,8 +1052,7 @@ buf_daemon(struct proc *p)
 
 	s = splbio();
 	for (;;) {
-		if (!numdirtypages ||
-		    (numdirtypages < hidirtypages && !needbuffer))
+		if (bcstats.numdirtypages < hidirtypages)
 			tsleep(&bd_req, PRIBIO - 7, "cleaner", 0);
 
 		getmicrouptime(&starttime);
@@ -1104,11 +1060,11 @@ buf_daemon(struct proc *p)
 		while ((bp = TAILQ_FIRST(&bufqueues[BQ_DIRTY]))) {
 			struct timeval tv;
 
-			if (numdirtypages < lodirtypages && !needbuffer)
+			if (bcstats.numdirtypages < lodirtypages)
 				break;
 
 			bremfree(bp);
-			SET(bp->b_flags, B_BUSY);
+			buf_acquire(bp);
 			splx(s);
 
 			if (ISSET(bp->b_flags, B_INVAL)) {
@@ -1125,10 +1081,10 @@ buf_daemon(struct proc *p)
 			    buf_countdeps(bp, 0, 0)) {
 				SET(bp->b_flags, B_DEFERRED);
 				s = splbio();
-				numfreepages += atop(bp->b_bufsize);
-				numdirtypages += atop(bp->b_bufsize);
+				bcstats.numdirtypages += atop(bp->b_bufsize);
 				binstailfree(bp, &bufqueues[BQ_DIRTY]);
-				CLR(bp->b_flags, B_BUSY);
+				bcstats.freebufs++;
+				buf_release(bp);
 				continue;
 			}
 
@@ -1154,6 +1110,8 @@ biowait(struct buf *bp)
 {
 	int s;
 
+	KASSERT(!(bp->b_flags & B_ASYNC));
+
 	s = splbio();
 	while (!ISSET(bp->b_flags, B_DONE))
 		tsleep(bp, PRIBIO + 1, "biowait", 0);
@@ -1203,8 +1161,11 @@ biodone(struct buf *bp)
 
 	if (!ISSET(bp->b_flags, B_READ)) {
 		CLR(bp->b_flags, B_WRITEINPROG);
+		bcstats.pendingwrites--;
 		vwakeup(bp->b_vp);
-	}
+	} else if (bcstats.numbufs && 
+	           (!(ISSET(bp->b_flags, B_RAW) || ISSET(bp->b_flags, B_PHYS))))
+		bcstats.pendingreads--;
 
 	if (ISSET(bp->b_flags, B_CALL)) {	/* if necessary, call out */
 		CLR(bp->b_flags, B_CALL);	/* but note callout done */
@@ -1218,66 +1179,3 @@ biodone(struct buf *bp)
 		}
 	}
 }
-
-#if 1
-void
-vfs_bufstats(void) {
-	return;
-}
-/* #ifdef DDB */
-#else
-/*
- * Print out statistics on the current allocation of the buffer pool.
- * Can be enabled to print out on every ``sync'' by setting "syncprt"
- * in vfs_syscalls.c using sysctl.
- */
-void
-vfs_bufstats(void)
-{
-	int s, i, j, count;
-	struct buf *bp;
-	struct bqueues *dp;
-	int counts[MAXBSIZE/PAGE_SIZE+1];
-	int totals[BQUEUES];
-	long ptotals[BQUEUES];
-	long pages;
-	static char *bname[BQUEUES] = { "CLEAN", "DIRTY", "EMPTY" };
-
-	s = splbio();
-	for (dp = bufqueues, i = 0; dp < &bufqueues[BQUEUES]; dp++, i++) {
-		count = 0;
-		pages = 0;
-		for (j = 0; j <= MAXBSIZE/PAGE_SIZE; j++)
-			counts[j] = 0;
-		TAILQ_FOREACH(bp, dp, b_freelist) {
-			counts[bp->b_bufsize/PAGE_SIZE]++;
-			count++;
-			pages += atop(bp->b_bufsize);
-		}
-		totals[i] = count;
-		ptotals[i] = pages;
-		printf("%s: total-%d(%d pages)", bname[i], count, pages);
-		for (j = 0; j <= MAXBSIZE/PAGE_SIZE; j++)
-			if (counts[j] != 0)
-				printf(", %d-%d", j * PAGE_SIZE, counts[j]);
-		printf("\n");
-	}
-	if ((ptotals[BQ_CLEAN] + ptotals[BQ_DIRTY]) != numfreepages)
-		printf("numfreepages counter wrong: %ld != %ld\n",
-		    numfreepages, ptotals[BQ_CLEAN] + ptotals[BQ_DIRTY]);
-	if (ptotals[BQ_CLEAN] != numcleanpages)
-		printf("numcleanpages counter wrong: %ld != %ld\n",
-		    numcleanpages, ptotals[<BQ_CLEAN]);
-	else
-		printf("numcleanpages: %ld\n", numcleanpages);
-	if (numdirtypages != ptotals[BQ_DIRTY])
-		printf("numdirtypages counter wrong: %ld != %ld\n",
-		    numdirtypages, ptotals[BQ_DIRTY]);
-	else
-		printf("numdirtypages: %ld\n", numdirtypages);
-
-	printf("syncer eating up to %ld pages from %ld reserved\n",
-	    maxcleanpages - hicleanpages, locleanpages);
-	splx(s);
-}
-#endif /* DEBUG */
diff --git a/sys/kern/vfs_subr.c b/sys/kern/vfs_subr.c
index b55c0242c8b..dabadded731 100644
--- a/sys/kern/vfs_subr.c
+++ b/sys/kern/vfs_subr.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: vfs_subr.c,v 1.169 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: vfs_subr.c,v 1.170 2008/06/12 06:58:39 deraadt Exp $	*/
 /*	$NetBSD: vfs_subr.c,v 1.53 1996/04/22 01:39:13 christos Exp $	*/
 
 /*
@@ -120,7 +120,7 @@ void
 vntblinit(void)
 {
 	/* buffer cache may need a vnode for each buffer */
-	maxvnodes = desiredvnodes;
+	maxvnodes = bufpages;
 	pool_init(&vnode_pool, sizeof(struct vnode), 0, 0, 0, "vnodes",
 	    &pool_allocator_nointr);
 	TAILQ_INIT(&vnode_hold_list);
@@ -1256,8 +1256,12 @@ vfs_sysctl(int *name, u_int namelen, void *oldp, size_t *oldlenp, void *newp,
 
 		free(tmpvfsp, M_TEMP);
 		return (ret);
-	}
+	case VFS_BCACHESTAT:	/* buffer cache statistics */
+		ret = sysctl_rdstruct(oldp, oldlenp, newp, &bcstats,
+		    sizeof(struct bcachestats));
+		return(ret);
 
+	}
 	return (EOPNOTSUPP);
 }
 
@@ -1664,7 +1668,7 @@ vfs_syncwait(int verbose)
 			if (bp->b_flags & B_DELWRI) {
 				s = splbio();
 				bremfree(bp);
-				bp->b_flags |= B_BUSY;
+				buf_acquire(bp);
 				splx(s);
 				nbusy++;
 				bawrite(bp);
@@ -1835,7 +1839,7 @@ loop:
 				break;
 			}
 			bremfree(bp);
-			bp->b_flags |= B_BUSY;
+			buf_acquire(bp);
 			/*
 			 * XXX Since there are no node locks for NFS, I believe
 			 * there is a slight chance that a delayed write will
@@ -1873,7 +1877,7 @@ loop:
 		if ((bp->b_flags & B_DELWRI) == 0)
 			panic("vflushbuf: not dirty");
 		bremfree(bp);
-		bp->b_flags |= B_BUSY;
+		buf_acquire(bp);
 		splx(s);
 		/*
 		 * Wait for I/O associated with indirect blocks to complete,
diff --git a/sys/miscfs/specfs/spec_vnops.c b/sys/miscfs/specfs/spec_vnops.c
index 15c2bd86377..b9394598065 100644
--- a/sys/miscfs/specfs/spec_vnops.c
+++ b/sys/miscfs/specfs/spec_vnops.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: spec_vnops.c,v 1.51 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: spec_vnops.c,v 1.52 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: spec_vnops.c,v 1.29 1996/04/22 01:42:38 christos Exp $	*/
 
 /*
@@ -447,7 +447,7 @@ loop:
 		if ((bp->b_flags & B_DELWRI) == 0)
 			panic("spec_fsync: not dirty");
 		bremfree(bp);
-		bp->b_flags |= B_BUSY;
+		buf_acquire(bp);
 		splx(s);
 		bawrite(bp);
 		goto loop;
diff --git a/sys/nfs/nfs_syscalls.c b/sys/nfs/nfs_syscalls.c
index c1ba30237c8..907551a7d59 100644
--- a/sys/nfs/nfs_syscalls.c
+++ b/sys/nfs/nfs_syscalls.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: nfs_syscalls.c,v 1.63 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: nfs_syscalls.c,v 1.64 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: nfs_syscalls.c,v 1.19 1996/02/18 11:53:52 fvdl Exp $	*/
 
 /*
@@ -746,7 +746,8 @@ nfssvc_iod(p)
 			    (B_BUSY|B_DELWRI|B_NEEDCOMMIT|B_NOCACHE))!=B_DELWRI)
 			    continue;
 			bremfree(nbp);
-			nbp->b_flags |= (B_BUSY|B_ASYNC);
+			nbp->b_flags |= B_ASYNC;
+			buf_acquire(nbp);
 			break;
 		    }
 		    /*
diff --git a/sys/nfs/nfs_vnops.c b/sys/nfs/nfs_vnops.c
index 9fc77ab8095..8b32df4b9bf 100644
--- a/sys/nfs/nfs_vnops.c
+++ b/sys/nfs/nfs_vnops.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: nfs_vnops.c,v 1.88 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: nfs_vnops.c,v 1.89 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: nfs_vnops.c,v 1.62.4.1 1996/07/08 20:26:52 jtc Exp $	*/
 
 /*
@@ -2736,7 +2736,8 @@ again:
 				!= (B_DELWRI | B_NEEDCOMMIT))
 				continue;
 			bremfree(bp);
-			bp->b_flags |= (B_BUSY | B_WRITEINPROG);
+			bp->b_flags |= B_WRITEINPROG;
+			buf_acquire(bp);
 			/*
 			 * A list of these buffers is kept so that the
 			 * second loop knows which buffers have actually
@@ -2814,10 +2815,12 @@ loop:
 		if ((passone || !commit) && (bp->b_flags & B_NEEDCOMMIT))
 			continue;
 		bremfree(bp);
-		if (passone || !commit)
-		    bp->b_flags |= (B_BUSY|B_ASYNC);
-		else
-		    bp->b_flags |= (B_BUSY|B_ASYNC|B_WRITEINPROG|B_NEEDCOMMIT);
+		if (passone || !commit) {
+			bp->b_flags |= B_ASYNC;
+		} else {
+			bp->b_flags |= (B_ASYNC|B_WRITEINPROG|B_NEEDCOMMIT);
+		}
+		buf_acquire(bp);
 		splx(s);
 		VOP_BWRITE(bp);
 		goto loop;
@@ -3013,6 +3016,8 @@ nfs_writebp(bp, force)
 	if (retv) {
 		if (force)
 			bp->b_flags |= B_WRITEINPROG;
+		bcstats.pendingwrites++;
+		bcstats.numwrites++;
 		VOP_STRATEGY(bp);
 	}
 
diff --git a/sys/sys/buf.h b/sys/sys/buf.h
index 59b696bf135..330cc90d614 100644
--- a/sys/sys/buf.h
+++ b/sys/sys/buf.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: buf.h,v 1.59 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: buf.h,v 1.60 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: buf.h,v 1.25 1997/04/09 21:12:17 mycroft Exp $	*/
 
 /*
@@ -87,6 +87,12 @@ struct buf {
 	dev_t	b_dev;			/* Device associated with buffer. */
 	caddr_t	b_data;			/* associated data */
 	void	*b_saveaddr;		/* Original b_data for physio. */
+
+	TAILQ_ENTRY(buf) b_valist;	/* LRU of va to reuse. */
+
+	struct uvm_object *b_pobj;	/* Object containing the pages */
+	off_t	b_poffs;		/* Offset within object */
+
 	daddr64_t	b_lblkno;	/* Logical block number. */
 	daddr64_t	b_blkno;	/* Underlying physical block number. */
 					/* Function to call upon completion.
@@ -161,6 +167,8 @@ struct buf *bufq_default_get(struct bufq *);
 #define	B_DEFERRED	0x04000000	/* Skipped over for cleaning */
 #define	B_SCANNED	0x08000000	/* Block already pushed during sync */
 #define	B_PDAEMON	0x10000000	/* I/O started by pagedaemon */
+#define B_RELEASED	0x20000000	/* free this buffer after its kvm */
+#define B_NOTMAPPED	0x40000000	/* BUSY, but not necessarily mapped */
 
 #define	B_BITS	"\010\001AGE\002NEEDCOMMIT\003ASYNC\004BAD\005BUSY\006CACHE" \
     "\007CALL\010DELWRI\012DONE\013EINTR\014ERROR" \
@@ -228,6 +236,23 @@ struct buf *getblk(struct vnode *, daddr64_t, int, int, int);
 struct buf *geteblk(int);
 struct buf *incore(struct vnode *, daddr64_t);
 
+/*
+ * buf_kvm_init initializes the kvm handling for buffers.
+ * buf_acquire sets the B_BUSY flag and ensures that the buffer is
+ * mapped in the kvm.
+ * buf_release clears the B_BUSY flag and allows the buffer to become
+ * unmapped.
+ * buf_unmap is for internal use only. Unmaps the buffer from kvm.
+ */
+void	buf_mem_init(vsize_t);
+void	buf_acquire(struct buf *);
+void	buf_acquire_unmapped(struct buf *);
+void	buf_release(struct buf *);
+int	buf_dealloc_mem(struct buf *);
+void	buf_alloc_pages(struct buf *, vsize_t);
+void	buf_free_pages(struct buf *);
+
+
 void	minphys(struct buf *bp);
 int	physio(void (*strategy)(struct buf *), struct buf *bp, dev_t dev,
 	    int flags, void (*minphys)(struct buf *), struct uio *uio);
diff --git a/sys/sys/mount.h b/sys/sys/mount.h
index 0e12cb233e7..0261a842247 100644
--- a/sys/sys/mount.h
+++ b/sys/sys/mount.h
@@ -1,4 +1,4 @@
-/*	$OpenBSD: mount.h,v 1.86 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: mount.h,v 1.87 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: mount.h,v 1.48 1996/02/18 11:55:47 fvdl Exp $	*/
 
 /*
@@ -465,10 +465,13 @@ struct mount {
 #define VFS_MAXTYPENUM	1	/* int: highest defined filesystem type */
 #define VFS_CONF	2	/* struct: vfsconf for filesystem given
 				   as next argument */
+#define VFS_BCACHESTAT	3	/* struct: buffer cache statistics given 
+				   as next argument */
 #define	CTL_VFSGENCTL_NAMES { \
 	{ 0, 0 }, \
 	{ "maxtypenum", CTLTYPE_INT }, \
-	{ "conf", CTLTYPE_NODE } \
+	{ "conf", CTLTYPE_NODE }, \
+	{ "bcachestat", CTLTYPE_STRUCT } \
 }
 
 /*
@@ -485,6 +488,24 @@ struct vfsconf {
 	struct	vfsconf *vfc_next;	/* next in list */
 };
 
+/* buffer cache statistics */
+struct bcachestats {
+	long numbufs; 		/* number of buffers allocated */
+	long freebufs;		/* number of free buffers */
+	long numbufpages; 	/* number of pages in buffer cache */
+	long numfreepages; 	/* number of free pages */
+	long numdirtypages; 	/* number of dirty free pages */
+	long numcleanpages; 	/* number of clean free pages */
+	long pendingwrites;	/* number of pending writes */
+	long pendingreads;	/* number of pending reads */
+	long numwrites;		/* total writes started */
+	long numreads;		/* total reads started */
+	long cachehits;		/* total reads found in cache */
+};
+#ifdef _KERNEL
+extern struct bcachestats bcstats;
+#endif
+
 /*
  * Operations supported on mounted file system.
  */
diff --git a/sys/ufs/ext2fs/ext2fs_bmap.c b/sys/ufs/ext2fs/ext2fs_bmap.c
index 1506fefffef..2d87ee895f4 100644
--- a/sys/ufs/ext2fs/ext2fs_bmap.c
+++ b/sys/ufs/ext2fs/ext2fs_bmap.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ext2fs_bmap.c,v 1.17 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: ext2fs_bmap.c,v 1.18 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: ext2fs_bmap.c,v 1.5 2000/03/30 12:41:11 augustss Exp $	*/
 
 /*
@@ -190,6 +190,7 @@ ext2fs_bmaparray(struct vnode *vp, int32_t bn, daddr64_t *bnp,
 			bp->b_flags |= B_READ;
 			VOP_STRATEGY(bp);
 			curproc->p_stats->p_ru.ru_inblock++;	/* XXX */
+			bcstats.pendingreads++;
 			if ((error = biowait(bp)) != 0) {
 				brelse(bp);
 				return (error);
diff --git a/sys/ufs/ext2fs/ext2fs_inode.c b/sys/ufs/ext2fs/ext2fs_inode.c
index f8cbb04f19d..62a0f438814 100644
--- a/sys/ufs/ext2fs/ext2fs_inode.c
+++ b/sys/ufs/ext2fs/ext2fs_inode.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ext2fs_inode.c,v 1.41 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: ext2fs_inode.c,v 1.42 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: ext2fs_inode.c,v 1.24 2001/06/19 12:59:18 wiz Exp $	*/
 
 /*
@@ -466,6 +466,8 @@ ext2fs_indirtrunc(struct inode *ip, int32_t lbn, int32_t dbn, int32_t lastbn, in
 	bp = getblk(vp, lbn, (int)fs->e2fs_bsize, 0, 0);
 	if (!(bp->b_flags & (B_DONE | B_DELWRI))) {
 		curproc->p_stats->p_ru.ru_inblock++;	/* pay for read */
+		bcstats.pendingreads++;
+		bcstats.numreads++;
 		bp->b_flags |= B_READ;
 		if (bp->b_bcount > bp->b_bufsize)
 			panic("ext2fs_indirtrunc: bad buffer size");
diff --git a/sys/ufs/ffs/ffs_inode.c b/sys/ufs/ffs/ffs_inode.c
index 99594c39f93..8500cb8d2f0 100644
--- a/sys/ufs/ffs/ffs_inode.c
+++ b/sys/ufs/ffs/ffs_inode.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ffs_inode.c,v 1.54 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: ffs_inode.c,v 1.55 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: ffs_inode.c,v 1.10 1996/05/11 18:27:19 mycroft Exp $	*/
 
 /*
@@ -501,6 +501,8 @@ ffs_indirtrunc(struct inode *ip, daddr64_t lbn, daddr64_t dbn,
 	bp = getblk(vp, lbn, (int)fs->fs_bsize, 0, 0);
 	if (!(bp->b_flags & (B_DONE | B_DELWRI))) {
 		curproc->p_stats->p_ru.ru_inblock++;	/* pay for read */
+		bcstats.pendingreads++;
+		bcstats.numreads++;
 		bp->b_flags |= B_READ;
 		if (bp->b_bcount > bp->b_bufsize)
 			panic("ffs_indirtrunc: bad buffer size");
diff --git a/sys/ufs/ffs/ffs_softdep.c b/sys/ufs/ffs/ffs_softdep.c
index e6fe34b1559..b7ef8b81994 100644
--- a/sys/ufs/ffs/ffs_softdep.c
+++ b/sys/ufs/ffs/ffs_softdep.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ffs_softdep.c,v 1.96 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: ffs_softdep.c,v 1.97 2008/06/12 06:58:40 deraadt Exp $	*/
 
 /*
  * Copyright 1998, 2000 Marshall Kirk McKusick. All Rights Reserved.
@@ -4691,9 +4691,9 @@ softdep_fsync_mountdev(vp, waitfor)
 		/* 
 		 * If it is already scheduled, skip to the next buffer.
 		 */
+		splassert(IPL_BIO);
 		if (bp->b_flags & B_BUSY)
 			continue;
-		bp->b_flags |= B_BUSY;
 
 		if ((bp->b_flags & B_DELWRI) == 0) {
 			FREE_LOCK(&lk);
@@ -4705,10 +4705,10 @@ softdep_fsync_mountdev(vp, waitfor)
 		 */
 		if ((wk = LIST_FIRST(&bp->b_dep)) == NULL ||
 		    wk->wk_type != D_BMSAFEMAP) {
-			bp->b_flags &= ~B_BUSY;
 			continue;
 		}
 		bremfree(bp);
+		buf_acquire(bp);
 		FREE_LOCK(&lk);
 		(void) bawrite(bp);
 		ACQUIRE_LOCK(&lk);
@@ -5616,7 +5616,7 @@ getdirtybuf(bp, waitfor)
 	if ((bp->b_flags & B_DELWRI) == 0)
 		return (0);
 	bremfree(bp);
-	bp->b_flags |= B_BUSY;
+	buf_acquire(bp);
 	return (1);
 }
 
diff --git a/sys/ufs/ffs/ffs_vnops.c b/sys/ufs/ffs/ffs_vnops.c
index e03549518ab..12b1e2c3552 100644
--- a/sys/ufs/ffs/ffs_vnops.c
+++ b/sys/ufs/ffs/ffs_vnops.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ffs_vnops.c,v 1.50 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: ffs_vnops.c,v 1.51 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: ffs_vnops.c,v 1.7 1996/05/11 18:27:24 mycroft Exp $	*/
 
 /*
@@ -449,7 +449,8 @@ loop:
 		}
 
 		bremfree(bp);
-		bp->b_flags |= B_BUSY | B_SCANNED;
+		buf_acquire(bp);
+		bp->b_flags |= B_SCANNED;
 		splx(s);
 		/*
 		 * On our final pass through, do all I/O synchronously
diff --git a/sys/ufs/ufs/ufs_bmap.c b/sys/ufs/ufs/ufs_bmap.c
index 68c8ea8521e..d39fb8db443 100644
--- a/sys/ufs/ufs/ufs_bmap.c
+++ b/sys/ufs/ufs/ufs_bmap.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: ufs_bmap.c,v 1.28 2008/06/11 12:35:46 deraadt Exp $	*/
+/*	$OpenBSD: ufs_bmap.c,v 1.29 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: ufs_bmap.c,v 1.3 1996/02/09 22:36:00 christos Exp $	*/
 
 /*
@@ -173,6 +173,8 @@ ufs_bmaparray(struct vnode *vp, daddr64_t bn, daddr64_t *bnp, struct indir *ap,
 		else {
 			bp->b_blkno = blkptrtodb(ump, daddr);
 			bp->b_flags |= B_READ;
+			bcstats.pendingreads++;
+			bcstats.numreads++;
 			VOP_STRATEGY(bp);
 			curproc->p_stats->p_ru.ru_inblock++;	/* XXX */
 			if ((error = biowait(bp)) != 0) {
diff --git a/sys/ufs/ufs/ufs_dirhash.c b/sys/ufs/ufs/ufs_dirhash.c
index 9184a466372..a9a51ec513c 100644
--- a/sys/ufs/ufs/ufs_dirhash.c
+++ b/sys/ufs/ufs/ufs_dirhash.c
@@ -1,4 +1,4 @@
-/* $OpenBSD: ufs_dirhash.c,v 1.19 2008/06/11 12:35:47 deraadt Exp $	*/
+/* $OpenBSD: ufs_dirhash.c,v 1.20 2008/06/12 06:58:40 deraadt Exp $	*/
 /*
  * Copyright (c) 2001, 2002 Ian Dowse.  All rights reserved.
  *
@@ -211,7 +211,6 @@ ufsdirhash_build(struct inode *ip)
 			if (UFS_BUFATOFF(ip, (off_t)pos, NULL, &bp) != 0)
 				goto fail;
 		}
-
 		/* Add this entry to the hash. */
 		ep = (struct direct *)((char *)bp->b_data + (pos & bmask));
 		if (ep->d_reclen == 0 || ep->d_reclen >
diff --git a/sys/uvm/uvm_swap.c b/sys/uvm/uvm_swap.c
index b1590559352..1388975190c 100644
--- a/sys/uvm/uvm_swap.c
+++ b/sys/uvm/uvm_swap.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: uvm_swap.c,v 1.81 2008/06/11 12:35:47 deraadt Exp $	*/
+/*	$OpenBSD: uvm_swap.c,v 1.82 2008/06/12 06:58:40 deraadt Exp $	*/
 /*	$NetBSD: uvm_swap.c,v 1.40 2000/11/17 11:39:39 mrg Exp $	*/
 
 /*
@@ -1980,7 +1980,7 @@ uvm_swap_io(pps, startslot, npages, flags)
 	 * fill in the bp.   we currently route our i/o through
 	 * /dev/drum's vnode [swapdev_vp].
 	 */
-	bp->b_flags = B_BUSY | B_NOCACHE | (flags & (B_READ|B_ASYNC));
+	bp->b_flags = B_BUSY | B_NOCACHE | B_RAW | (flags & (B_READ|B_ASYNC));
 	bp->b_proc = &proc0;	/* XXX */
 	bp->b_vnbufs.le_next = NOLIST;
 	bp->b_data = (caddr_t)kva;