On Tue, 2005-27-12 at 10:03 +0100, Thomas Graf wrote: > In theory this patch is absolutely correct and we should go > that way one day. The problem is that iproute sets NLM_F_EXCL > by default when adding rules so this patch would modify the > behaviour of all existing "ip rule add" usages. > > The explicit use of preferences makes rules unique at all > times so this is a minor issue and we should defer it to the > next point allowing for a binary interface breakage.
Using explicit priorities is also "broken". Has been since day one - Alexey was planning it to fix it "some day". i.e if you add a second exact same rule with exactly the same prio, it will be lifo added. OTOH, if you dont specify a priority the kernel gives you one and it becomes unique that way. The patch is wrong given the reasoning as to why it was created in the first place - and is also wrong for not comparing priorities. However, the spirit of the patch this is right and if Gabor can resubmit with appropriate r_preference comparison, i think we should take it in. Its one of those things that has been broken for years and needed someone to take the initiative to fix them. The second step should be to fix iprule to set NLM_F_EXCL only when priority/preference/order using "ip rule .." I dont think there would be any ABI breakages this way. old binaries continue to set NLM_F_EXCL and would not be added if all tuples are exactly the same; and new binaries will set NLM_F_EXCL only if a priority is defined. Thoughts? cheers, jamal - To unsubscribe from this list: send the line "unsubscribe netdev" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
