I have not gotten a chance to build up a network to start testing things and the ip code abstracts so many things that I can not be sure on some things. I am looking to setup a network with an in between hop that is limited by packet/sec more than bits/secs but it has a large MTU. So even though the PMTU from end to end may be smaller than the MTU for the next hop, I still don't want to fragment until after going through the next hop. After that it will be fine to fragment the packet as needed. It seems to me that the code always uses the PMTU to fragment things when it can, but in this case I still want to have the largest possible packets for the next hop even if they will be fragmented later. Looking at ip_forward() I see: mtu = rt->u.dst.pmtu; .... if (skb->len > mtu && (ntohs(iph->frag_off) & IP_DF)) goto frag_needed; this is fine with me. I would want the PMTU reported back as quickly as possible. But looking in ip_fragment I see that pmtu is used again and I would want just the MTU of the next hop used instead. Any insight into the pmtu var would be helpful. -- Andrew May acmay@acmay.homeip.net - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org
