From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mga11.intel.com (mga11.intel.com [192.55.52.93]) by dpdk.org (Postfix) with ESMTP id 00BBCDE0 for ; Wed, 10 Jun 2015 17:39:56 +0200 (CEST) Received: from fmsmga003.fm.intel.com ([10.253.24.29]) by fmsmga102.fm.intel.com with ESMTP; 10 Jun 2015 08:39:56 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.13,588,1427785200"; d="scan'208";a="506172934" Received: from irsmsx101.ger.corp.intel.com ([163.33.3.153]) by FMSMGA003.fm.intel.com with ESMTP; 10 Jun 2015 08:39:55 -0700 Received: from irsmsx105.ger.corp.intel.com ([169.254.7.73]) by IRSMSX101.ger.corp.intel.com ([169.254.1.217]) with mapi id 14.03.0224.002; Wed, 10 Jun 2015 16:39:54 +0100 From: "Ananyev, Konstantin" To: Olivier MATZ , "O'Driscoll, Tim" , "Zhang, Helin" , "dev@dpdk.org" Thread-Topic: [dpdk-dev] [PATCH v6 01/18] mbuf: redefine packet_type in rte_mbuf Thread-Index: AQHQo4qVFWxBo10iiEOCjOLJ5VWpcJ2l4AmQ Date: Wed, 10 Jun 2015 15:39:53 +0000 Message-ID: <2601191342CEEE43887BDE71AB97725836A078D6@irsmsx105.ger.corp.intel.com> References: <1432284264-17376-1-git-send-email-helin.zhang@intel.com> <1433144045-30847-1-git-send-email-helin.zhang@intel.com> <1433144045-30847-2-git-send-email-helin.zhang@intel.com> <556C1478.9040005@6wind.com> <26FA93C7ED1EAA44AB77D62FBE1D27BA54D536B3@IRSMSX102.ger.corp.intel.com> <55784A82.6020801@6wind.com> In-Reply-To: <55784A82.6020801@6wind.com> Accept-Language: en-IE, en-US Content-Language: en-US X-MS-Has-Attach: X-MS-TNEF-Correlator: x-originating-ip: [163.33.239.180] Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: quoted-printable MIME-Version: 1.0 Subject: Re: [dpdk-dev] [PATCH v6 01/18] mbuf: redefine packet_type in rte_mbuf X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: patches and discussions about DPDK List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 10 Jun 2015 15:39:57 -0000 Hi Olivier, > -----Original Message----- > From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Olivier MATZ > Sent: Wednesday, June 10, 2015 3:33 PM > To: O'Driscoll, Tim; Zhang, Helin; dev@dpdk.org > Subject: Re: [dpdk-dev] [PATCH v6 01/18] mbuf: redefine packet_type in rt= e_mbuf >=20 > Hi Tim, Helin, >=20 > On 06/02/2015 03:27 PM, O'Driscoll, Tim wrote: > > > >> -----Original Message----- > >> From: dev [mailto:dev-bounces@dpdk.org] On Behalf Of Olivier MATZ > >> Sent: Monday, June 1, 2015 9:15 AM > >> To: Zhang, Helin; dev@dpdk.org > >> Subject: Re: [dpdk-dev] [PATCH v6 01/18] mbuf: redefine packet_type in > >> rte_mbuf > >> > >> Hi Helin, > >> > >> +CC Neil > >> > >> On 06/01/2015 09:33 AM, Helin Zhang wrote: > >>> In order to unify the packet type, the field of 'packet_type' in > >>> 'struct rte_mbuf' needs to be extended from 16 to 32 bits. > >>> Accordingly, some fields in 'struct rte_mbuf' are re-organized to > >>> support this change for Vector PMD. As 'struct rte_kni_mbuf' for > >>> KNI should be right mapped to 'struct rte_mbuf', it should be > >>> modified accordingly. In addition, Vector PMD of ixgbe is disabled > >>> by default, as 'struct rte_mbuf' changed. > >>> To avoid breaking ABI compatibility, all the changes would be > >>> enabled by RTE_UNIFIED_PKT_TYPE, which is disabled by default. > >> > >> What are the plans for this compile-time option in the future? > >> > >> I wonder what are the benefits of having this option in terms > >> of ABI compatibility: when it is disabled, it is ABI-compatible but > >> the packet-type feature is not present, and when it is enabled we > >> have the feature but it breaks the compatibility. > >> > >> In my opinion, the v5 is preferable: for this kind of features, I > >> don't see how the ABI can be preserved, and I think packet-type > >> won't be the only feature that will modify the mbuf structure. I think > >> the process described here should be applied: > >> http://dpdk.org/browse/dpdk/tree/doc/guides/rel_notes/abi.rst > >> > >> (starting from "Some ABI changes may be too significant to reasonably > >> maintain multiple versions of"). > >> > >> > >> Regards, > >> Olivier > >> > > > > This is just like the change that Steve (Cunming) Liang submitted for I= nterrupt Mode. We have the same problem in both cases: we > want to find a way to get the features included, but need to comply with = our ABI policy. So, in both cases, the proposal is to add a > config option to enable the change by default, so we maintain backward co= mpatibility. Users that want these changes, and are willing > to accept the associated ABI change, have to specifically enable them. > > > > We can note in the Deprecation Notices in the Release Notes for 2.1 tha= t these config options will be removed in 2.2. The features > will then be enabled by default. > > > > This seems like a good compromise which allows us to get these changes = into 2.1 but avoids breaking the ABI policy. >=20 > Sorry for the late answer. >=20 > After some thoughts on this topic, I understand that having a > compile-time option is perhaps a good compromise between > keeping compatibility and having new features earlier. >=20 > I'm just afraid about having one #ifdef in the code for > each new feature that cannot keep the ABI compatibility. > What do you think about having one option -- let's call > it "CONFIG_RTE_NEXT_ABI" --, that is disabled by default, > and that would surround any new feature that breaks the > ABI? I am not Tim/Helin, but really like that idea :) Konstantin >=20 > This would have several advantages: > - only 2 cases (on or off), the combinatorial is smaller than > having one option per feature > - all next features breaking the abi can be identified by a grep > - the code inside the #ifdef can be enabled in a simple operation > by Thomas after each release. >=20 > Thomas, any comment? >=20 > Regards, > Olivier >=20