[9fans] A snippet from a NetBSD mailing list

[Lucio De Re <lucio@proxima.alt.za>]

I think congratulations and heartfelt thanks are in order for the
Plan 9 team.  I haven't yet tried to install 4ed, but from all
reports, it seems formidable.

To return to the subject of my mail, I wonder if the message below is
pertinent.  I know I have a 3Com 3C905CX-TXM that neither NetBSD nor
Plan 9 (3ed) knew how to handle.

I know the patch isn't useful outside of NetBSD, but perhaps the
pertinent details can be extracted from it.

Of course, I don't know that this hasn't been fixed in 4ed :-(

++L

----- Forwarded message from enami tsugutomo <enami@sm.sony.co.jp> -----

From: enami tsugutomo <enami@sm.sony.co.jp>
To: tech-net@netbsd.org
Subject: changes to elinkxl.c:ex_set_mc()
Sender: tech-net-owner@netbsd.org
Precedence: list

Hi, all.

In the function ex_set_mc() defined in elinkxl.c, there is two
possible bugs.

One is that it assumes someone set IFF_ALLMULTI for us, but actually
it's not true.  We need to set by our self.

Other is that for newer cards which supports multicast hash filtering,
we need to clear hash bit before programming them.

Appened diff is created by me and Frank.  But since I have only
pre-905B card, I can't test the latter.  So, I beg someone to test or
at least review this.

enami.

Index: elinkxl.c
===================================================================
RCS file: /cvsroot/syssrc/sys/dev/ic/elinkxl.c,v
retrieving revision 1.62
diff -u -r1.62 elinkxl.c
--- elinkxl.c	2002/04/06 19:28:01	1.62
+++ elinkxl.c	2002/04/30 06:28:17
@@ -711,7 +711,9 @@
 	return (error);
 }

-#define	ex_mchash(addr)	(ether_crc32_be((addr), ETHER_ADDR_LEN) & 0xff)
+#define	MCHASHSIZE		256
+#define	ex_mchash(addr)		(ether_crc32_be((addr), ETHER_ADDR_LEN) & \
+				    (MCHASHSIZE - 1))

 /*
  * Set multicast receive filter. Also take care of promiscuous mode
@@ -728,28 +730,44 @@
 	int i;
 	u_int16_t mask = FIL_INDIVIDUAL | FIL_BRDCST;

-	if (ifp->if_flags & IFF_PROMISC)
+	if (ifp->if_flags & IFF_PROMISC) {
 		mask |= FIL_PROMISC;
+		goto allmulti;
+	}

-	if (!(ifp->if_flags & IFF_MULTICAST))
-		goto out;
+	ETHER_FIRST_MULTI(estep, ec, enm);
+	if (enm == NULL)
+		goto nomulti;

-	if (!(sc->ex_conf & EX_CONF_90XB) || ifp->if_flags & IFF_ALLMULTI) {
-		mask |= (ifp->if_flags & IFF_MULTICAST) ? FIL_MULTICAST : 0;
-	} else {
-		ETHER_FIRST_MULTI(estep, ec, enm);
-		while (enm != NULL) {
-			if (memcmp(enm->enm_addrlo, enm->enm_addrhi,
-			    ETHER_ADDR_LEN) != 0)
-				goto out;
-			i = ex_mchash(enm->enm_addrlo);
-			bus_space_write_2(sc->sc_iot, sc->sc_ioh,
-			    ELINK_COMMAND, ELINK_SETHASHFILBIT | i);
-			ETHER_NEXT_MULTI(estep, enm);
-		}
-		mask |= FIL_MULTIHASH;
-	}
- out:
+	if ((sc->ex_conf & EX_CONF_90XB) == 0)
+		/* No multicast hash filtering. */
+		goto allmulti;
+
+	for (i = 0; i < MCHASHSIZE; i++)
+		bus_space_write_2(sc->sc_iot, sc->sc_ioh,
+		    ELINK_COMMAND, ELINK_CLEARHASHFILBIT | i);
+
+	do {
+		if (memcmp(enm->enm_addrlo, enm->enm_addrhi,
+		    ETHER_ADDR_LEN) != 0)
+			goto allmulti;
+
+		i = ex_mchash(enm->enm_addrlo);
+		bus_space_write_2(sc->sc_iot, sc->sc_ioh,
+		    ELINK_COMMAND, ELINK_SETHASHFILBIT | i);
+		ETHER_NEXT_MULTI(estep, enm);
+	} while (enm != NULL);
+	mask |= FIL_MULTIHASH;
+
+nomulti:
+	ifp->if_flags &= ~IFF_ALLMULTI;
+	bus_space_write_2(sc->sc_iot, sc->sc_ioh, ELINK_COMMAND,
+	    SET_RX_FILTER | mask);
+	return;
+
+allmulti:
+	ifp->if_flags |= IFF_ALLMULTI;
+	mask |= FIL_MULTICAST;
 	bus_space_write_2(sc->sc_iot, sc->sc_ioh, ELINK_COMMAND,
 	    SET_RX_FILTER | mask);
 }
Index: elinkxlreg.h
===================================================================
RCS file: /cvsroot/syssrc/sys/dev/ic/elinkxlreg.h,v
retrieving revision 1.10
diff -u -r1.10 elinkxlreg.h
--- elinkxlreg.h	2001/12/28 20:35:47	1.10
+++ elinkxlreg.h	2002/04/30 06:28:17
@@ -164,6 +164,7 @@
 #define ELINK_DNUNSTALL		0x3003
 #define ELINK_TXRECLTHRESH	0xc000
 #define ELINK_TXSTARTTHRESH	0x9800
+#define ELINK_CLEARHASHFILBIT	0xc800
 #define ELINK_SETHASHFILBIT	0xcc00

 /*

----- End forwarded message -----