From mboxrd@z Thu Jan 1 00:00:00 1970 From: "David S. Miller" Subject: Re: PMTU issues due to TOS field manipulation (for DSCP) Date: Wed, 10 Dec 2003 14:51:49 -0800 Sender: netdev-bounce@oss.sgi.com Message-ID: <20031210145149.2bd89e9b.davem@redhat.com> References: <3FD79FF1.8000505@us.ibm.com> Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Cc: ja@ssi.bg, ak@suse.de, ruddk@us.ibm.com, kuznet@ms2.inr.ac.ru, netdev@oss.sgi.com, chester.f.johnson@intel.com Return-path: To: Nivedita Singhvi In-Reply-To: <3FD79FF1.8000505@us.ibm.com> Errors-to: netdev-bounce@oss.sgi.com List-Id: netdev.vger.kernel.org Here is my take on this, as far as Linux is concerned. I agree with the three behaviors proposed by Julian. However I have some slight trouble with the ignore-TOS-for- PMTU idea, implementation wise. Walking the routing hash table for each possible TOS value is going to be computationally expensive, and is inviting computational complexity DDoS attacks by bombing the machine with PMTU ICMP messages. That is the most obvious implementation, and I'm not saying there are not others. I just have no alternatives in mind right now :) But once that issue is resolved I'm more than happy to put a patch in which does this stuff. We even have been speaking about this in other threads on netdev wrt. Julian's patches. TOS is truly a value with only network local meaning and hops are going to modify the value on us. I'm actually surprised this is the first time the issue has been seriously hit.