From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ido Schimmel Subject: Re: [PATCH net] net/ipv6: Revert attempt to simplify route replace and append Date: Thu, 5 Jul 2018 00:10:41 +0300 Message-ID: <20180704211041.GB24525@splinter> References: <20180702220312.14590-1-dsahern@kernel.org> <20180703134302.GB15120@splinter.mtl.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: dsahern@kernel.org, netdev@vger.kernel.org, sharpd@cumulusnetworks.com, Thomas.Winter@alliedtelesis.co.nz, petrm@mellanox.com To: David Ahern Return-path: Received: from mail-eopbgr50048.outbound.protection.outlook.com ([40.107.5.48]:65093 "EHLO EUR03-VE1-obe.outbound.protection.outlook.com" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1752653AbeGDVKw (ORCPT ); Wed, 4 Jul 2018 17:10:52 -0400 Content-Disposition: inline In-Reply-To: Sender: netdev-owner@vger.kernel.org List-ID: On Tue, Jul 03, 2018 at 02:02:06PM -0600, David Ahern wrote: > It is unfortunate that mlxsw has to replicate the node lookup code. The kernel can store multiple routes with the same prefix/length, but only one is used for forwarding. Thus when a route is deleted it should be potentially overwritten by a different route in the device's tables. This is why mlxsw stores these nodes. We can have the IPv4/IPv6 code only generate a REPLACE / DELETE notification for routes that are actually used for forwarding and relieve listeners from the need to implement this logic themselves. I think this should work. ?