tg3: Drop non-VLAN rx pkts larger than the MTU

The hardware's MTU is configured so that it will accept packets whose
VLAN tag is left inline.  This means that the hardware will also accept
other packets larger than the desired MTU size.  This patch checks for
and discards packets that are larger than the requested MTU but are not
VLAN tagged frames.

Signed-off-by: Matt Carlson <mcarlson@broadcom.com>
Signed-off-by: Benjamin Li <benli@broadcom.com>
Signed-off-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
This commit is contained in:
Matt Carlson 2009-02-25 14:21:52 +00:00 committed by David S. Miller
parent a6435f3a53
commit f7b493e021

View File

@ -4339,6 +4339,13 @@ static int tg3_rx(struct tg3 *tp, int budget)
skb->ip_summed = CHECKSUM_NONE;
skb->protocol = eth_type_trans(skb, tp->dev);
if (len > (tp->dev->mtu + ETH_HLEN) &&
skb->protocol != htons(ETH_P_8021Q)) {
dev_kfree_skb(skb);
goto next_pkt;
}
#if TG3_VLAN_TAG_USED
if (tp->vlgrp != NULL &&
desc->type_flags & RXD_FLAG_VLAN) {
@ -6824,7 +6831,8 @@ static int tg3_reset_hw(struct tg3 *tp, int reset_phy)
__tg3_set_mac_addr(tp, 0);
/* MTU + ethernet header + FCS + optional VLAN tag */
tw32(MAC_RX_MTU_SIZE, tp->dev->mtu + ETH_HLEN + 8);
tw32(MAC_RX_MTU_SIZE,
tp->dev->mtu + ETH_HLEN + ETH_FCS_LEN + VLAN_HLEN);
/* The slot time is changed by tg3_setup_phy if we
* run at gigabit with half duplex.