add a new pool for 2k + 2 byte (mcl2k2) clusters.

a certain vendor likes to make chips that specify the rx buffer
sizes in kilobyte increments. unfortunately it places the ethernet
header on the start of the rx buffer, which means if you give it a
mcl2k cluster, the ethernet header will not be ETHER_ALIGNed cos
mcl2k clusters are always allocated on 2k boundarys (cos they pack
into pages well). that in turn means the ip header wont be aligned
correctly.

the current workaround on these chips has been to let non-strict
alignment archs just use the normal 2k cluster, but use whatever
cluster can fit 2k + 2 on strict archs. that turns out to be the
4k cluster, meaning we waste nearly 2k of space on every packet.

properly aligning the ethernet header and ip headers gives a
performance boost, even on non-strict archs.

1 files changed, 12 insertions, 4 deletions

diff --git a/sys/kern/uipc_mbuf.c b/sys/kern/uipc_mbuf.c
index a6e46b5f2f8..e0ba9457212 100644
--- a/sys/kern/uipc_mbuf.c
+++ b/sys/kern/uipc_mbuf.c
@@ -1,4 +1,4 @@
-/*	$OpenBSD: uipc_mbuf.c,v 1.235 2016/10/24 23:58:33 dlg Exp $	*/
+/*	$OpenBSD: uipc_mbuf.c,v 1.236 2016/10/27 03:00:35 dlg Exp $	*/
 /*	$NetBSD: uipc_mbuf.c,v 1.15.4.1 1996/06/13 17:11:44 cgd Exp $	*/
 
 /*
@@ -110,6 +110,7 @@ struct	pool mtagpool;
 /* mbuf cluster pools */
 u_int	mclsizes[MCLPOOLS] = {
 	MCLBYTES,	/* must be at slot 0 */
+	MCLBYTES + 2,	/* ETHER_ALIGNED 2k mbufs */
 	4 * 1024,
 	8 * 1024,
 	9 * 1024,
@@ -145,6 +146,7 @@ void
 mbinit(void)
 {
 	int i;
+	unsigned int lowbits;
 
 #if DIAGNOSTIC
 	if (mclsizes[0] != MCLBYTES)
@@ -161,9 +163,15 @@ mbinit(void)
 	    IPL_NET, 0, "mtagpl", NULL);
 
 	for (i = 0; i < nitems(mclsizes); i++) {
-		snprintf(mclnames[i], sizeof(mclnames[0]), "mcl%dk",
-		    mclsizes[i] >> 10);
-		pool_init(&mclpools[i], mclsizes[i], 0, IPL_NET, 0,
+		lowbits = mclsizes[i] & ((1 << 10) - 1);
+		if (lowbits) {
+			snprintf(mclnames[i], sizeof(mclnames[0]),
+			    "mcl%dk%u", mclsizes[i] >> 10, lowbits);
+		} else {
+			snprintf(mclnames[i], sizeof(mclnames[0]), "mcl%dk",
+			    mclsizes[i] >> 10);
+		}
+		pool_init(&mclpools[i], mclsizes[i], 64, IPL_NET, 0,
 		    mclnames[i], NULL);
 		pool_set_constraints(&mclpools[i], &kp_dma_contig);
 		pool_setlowat(&mclpools[i], mcllowat);