If a network interface has many vlan sub-interfaces (~400 or more) with enabled GVRP, PDUs may become larger than the interface PDU. GARP packet generator can send multiple PDUs, but it relies heavily on skb_tailroom() not exceeding maximal packet size.
Patch to fix this issue is proposed, I'm not sure if it's a correct way to handle this situation, but after a short test looks like it works - now I get multiple PDUs with maximum size of a 1514 bytes (1500 MTU) and switch recognizes them now. MRP protocol can have the same problem, but I can't test it. Signed-Off-By: Vitaly V. Bursov <[email protected]> --- a/net/802/garp.c 2014-01-13 12:55:07.000000000 +0200 +++ b/net/802/garp.c 2014-01-13 12:58:30.000000000 +0200 @@ -210,6 +210,7 @@ { struct sk_buff *skb; struct garp_pdu_hdr *gp; + int extra_size; #define LLC_RESERVE sizeof(struct llc_pdu_un) skb = alloc_skb(app->dev->mtu + LL_RESERVED_SPACE(app->dev), @@ -221,6 +222,13 @@ skb->protocol = htons(ETH_P_802_2); skb_reserve(skb, LL_RESERVED_SPACE(app->dev) + LLC_RESERVE); + /* Reserve extra space to avoid PDUs larger than MTU, + * Other code depends on skb_tailroom() to return exact + * usable packet space left */ + extra_size = skb_tailroom(skb) - (app->dev->mtu - LLC_RESERVE); + if (extra_size > 0) + skb_reserve(skb, extra_size); + gp = (struct garp_pdu_hdr *)__skb_put(skb, sizeof(*gp)); put_unaligned(htons(GARP_PROTOCOL_ID), &gp->protocol); -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
