From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id F12BA433E8; Tue, 28 Nov 2023 12:29:41 +0100 (CET) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 8849F40E54; Tue, 28 Nov 2023 12:29:41 +0100 (CET) Received: from szxga01-in.huawei.com (szxga01-in.huawei.com [45.249.212.187]) by mails.dpdk.org (Postfix) with ESMTP id E248D402C2 for ; Tue, 28 Nov 2023 12:29:39 +0100 (CET) Received: from kwepemm000004.china.huawei.com (unknown [172.30.72.56]) by szxga01-in.huawei.com (SkyGuard) with ESMTP id 4SfgBH08tMzmXGK; Tue, 28 Nov 2023 19:25:59 +0800 (CST) Received: from [10.67.121.59] (10.67.121.59) by kwepemm000004.china.huawei.com (7.193.23.18) with Microsoft SMTP Server (version=TLS1_2, cipher=TLS_ECDHE_RSA_WITH_AES_128_GCM_SHA256) id 15.1.2507.35; Tue, 28 Nov 2023 19:29:34 +0800 Message-ID: <659676fa-fce3-1f93-b893-e5f0d0f55ec9@huawei.com> Date: Tue, 28 Nov 2023 19:29:33 +0800 MIME-Version: 1.0 User-Agent: Mozilla/5.0 (Windows NT 10.0; Win64; x64; rv:91.0) Gecko/20100101 Thunderbird/91.2.0 Subject: Re: [PATCH v3 1/6] doc: add RSS hash algorithm feature To: Ferruh Yigit , CC: , , Chengwen Feng , Dongdong Liu , Jie Hai References: <20231123135916.33315-1-lihuisong@huawei.com> <20231125014745.61348-1-lihuisong@huawei.com> <20231125014745.61348-2-lihuisong@huawei.com> <89b73632-8a7a-e54c-2957-04c347027c2e@huawei.com> <8968ac41-bf19-40ac-a057-aad7b46b7ca8@amd.com> <93239148-8a4d-7b3a-9e39-6659566c64dc@huawei.com> <7c165606-8064-41c0-82c2-51453f545577@amd.com> From: "lihuisong (C)" In-Reply-To: <7c165606-8064-41c0-82c2-51453f545577@amd.com> Content-Type: text/plain; charset="UTF-8"; format=flowed Content-Transfer-Encoding: 8bit X-Originating-IP: [10.67.121.59] X-ClientProxiedBy: dggems702-chm.china.huawei.com (10.3.19.179) To kwepemm000004.china.huawei.com (7.193.23.18) X-CFilter-Loop: Reflected X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org 在 2023/11/28 18:09, Ferruh Yigit 写道: > On 11/28/2023 1:21 AM, lihuisong (C) wrote: >> 在 2023/11/27 23:43, Ferruh Yigit 写道: >>> On 11/27/2023 1:12 PM, lihuisong (C) wrote: >>>> 在 2023/11/27 20:19, Ferruh Yigit 写道: >>>>> On 11/25/2023 1:47 AM, Huisong Li wrote: >>>>>> Add hash algorithm feature introduced by 23.11 and fix some RSS >>>>>> features >>>>>> description. >>>>>> >>>>>> Fixes: 34ff088cc241 ("ethdev: set and query RSS hash algorithm") >>>>>> >>>>>> Signed-off-by: Huisong Li >>>>>> Acked-by: Chengwen Feng >>>>>> --- >>>>>>    doc/guides/nics/features.rst | 26 ++++++++++++++++++++++---- >>>>>>    1 file changed, 22 insertions(+), 4 deletions(-) >>>>>> >>>>>> diff --git a/doc/guides/nics/features.rst >>>>>> b/doc/guides/nics/features.rst >>>>>> index 1a1dc16c1e..0d38c5c525 100644 >>>>>> --- a/doc/guides/nics/features.rst >>>>>> +++ b/doc/guides/nics/features.rst >>>>>> @@ -277,10 +277,12 @@ RSS hash >>>>>>    Supports RSS hashing on RX. >>>>>>      * **[uses]     user config**: ``dev_conf.rxmode.mq_mode`` = >>>>>> ``RTE_ETH_MQ_RX_RSS_FLAG``. >>>>>> -* **[uses]     user config**: ``dev_conf.rx_adv_conf.rss_conf``. >>>>>> +* **[uses]     user config**: ``rss_conf.rss_hf``. >>>>>> >>>>> Feature title is "RSS hash", it can be two things, >>>>> 1. "Receive Side Scaling" support >>>>> 2. Provide RSS hash to application >>>>> >>>>> When this document first prepared RSS hash value was always provided to >>>>> the application when RSS enabled. >>>>> So intention with this feature was "Receive Side Scaling" support, >>>>> hence >>>>> 'RTE_ETH_MQ_RX_RSS_FLAG' added. >>>>> >>>>> Later providing RSS has to the application separated as optimization, >>>>> 'RTE_ETH_RX_OFFLOAD_RSS_HASH' & 'RTE_MBUF_F_RX_RSS_HASH' added for this >>>>> support. >>>> What should I do for above two comments？ >>>> To tell application how to use it? >>>> >>> Just tried to give some context. >> got it. >>> >>> >>>>> As the intention of this feature is "Receive Side Scaling" support, we >>>>> shouldn't reduce configuration struct to 'rss_conf.rss_hf'. >>>>> >>>>> Instead perhaps can expand to: >>>>> 'rte_eth_conf.rx_adv_conf.rss_conf', 'rte_eth_rss_conf' >>>>   I just pick their common part.😁 >>>> >>>> ok, will fix it. >>>> >>>>>>    * **[uses]     rte_eth_rxconf,rte_eth_rxmode**: >>>>>> ``offloads:RTE_ETH_RX_OFFLOAD_RSS_HASH``. >>>>>>    * **[provides] rte_eth_dev_info**: ``flow_type_rss_offloads``. >>>>>>    * **[provides] mbuf**: ``mbuf.ol_flags:RTE_MBUF_F_RX_RSS_HASH``, >>>>>> ``mbuf.rss``. >>>>>> +* **[related]  API**: ``rte_eth_dev_configure``, >>>>>> ``rte_eth_dev_rss_hash_update`` >>>>>> +  ``rte_eth_dev_rss_hash_conf_get()``. >>>>>> >>>>> ack >>>>> >>>>>>      .. _nic_features_inner_rss: >>>>>> @@ -288,7 +290,7 @@ Supports RSS hashing on RX. >>>>>>    Inner RSS >>>>>>    --------- >>>>>>    -Supports RX RSS hashing on Inner headers. >>>>>> +Supports RX RSS hashing on Inner headers by rte_flow API. >>>>>> >>>>> This should be clarified with details below, not sure if it required to >>>>> limit description to rte_flow. >>>> But this block like rte_flow_action_rss is from rte_flow. >>>> And ethdev ops doesn't support inner RSS. >>>> So I think it is ok. >>>> >>> Yes it is supported by rte_flow, and '[uses]' information should already >>> clarify it. >> Should we remove the 'rte_flow API' wrods I added in above description? >> > I think it can be removed. The latest version(V5, only three fix doc patch) has been sent out. Do we I need to send V6 for this?😂 > > >>>>> And I guess similar confusion exist with the providing hash to user. >>>>> Need to check if rte_flow implementation puts hash to mbuf along with >>>>> doing the RSS, or if it checks 'RTE_ETH_RX_OFFLOAD_RSS_HASH' offload, >>>>> and update below items accordingly. >>>> Do we need to tell user how to use it here? >>>> I feel this document is a little simple and main to list interface for >>>> user. >>>> In addition, it is better that the more detail about RSS should be >>>> presented  in rte_flow features. >>>> >>> No, I am not suggesting to add more detail. >>> >>> My concern is 'RTE_ETH_RX_OFFLOAD_RSS_HASH' information may not be >>> correct, ethdev APIs checks offload flags, but does rte_flow >>> implementation check it? >> As far as I know, It is possibly verified in PMD if have or required. >>> My suggestion is double check that piece of information and fix it if >>> required. >>> >>> >>>>>>    * **[uses]    rte_flow_action_rss**: ``level``. >>>>>>    * **[uses]    rte_eth_rxconf,rte_eth_rxmode**: >>>>>> ``offloads:RTE_ETH_RX_OFFLOAD_RSS_HASH``. >>>>>> @@ -303,9 +305,25 @@ RSS key update >>>>>>    Supports configuration of Receive Side Scaling (RSS) hash >>>>>> computation. Updating >>>>>>    Receive Side Scaling (RSS) hash key. >>>>>>    -* **[implements] eth_dev_ops**: ``rss_hash_update``, >>>>>> ``rss_hash_conf_get``. >>>>>> +* **[implements] eth_dev_ops**: ``dev_configure``, >>>>>> ``rss_hash_update``, ``rss_hash_conf_get``. >>>>>> +* **[uses]     user config**: ``rss_conf.rss_key``, >>>>>> ``rss_conf.rss_key_len`` >>>>>>    * **[provides]   rte_eth_dev_info**: ``hash_key_size``. >>>>>> -* **[related]    API**: ``rte_eth_dev_rss_hash_update()``, >>>>>> +* **[related]    API**: ``rte_eth_dev_configure``, >>>>>> ``rte_eth_dev_rss_hash_update()``, >>>>>> +  ``rte_eth_dev_rss_hash_conf_get()``. >>>>>> + >>>>> ack >>>>> >>>>> There is an inconsistency in the documentation but I think it is >>>>> good to >>>>> use '()' when documenting API, like: 'rte_eth_dev_configure()' >>>> +1 will fix it. >>>>>> + >>>>>> +.. _nic_features_rss_hash_algo_update: >>>>>> + >>>>>> +RSS hash algorithm update >>>>>> +------------------------- >>>>>> + >>>>>> +Supports configuration of Receive Side Scaling (RSS) hash algorithm. >>>>>> Updating >>>>>> +RSS hash algorithm. >>>>>> + >>>>>> +* **[implements] eth_dev_ops**: ``dev_configure``, >>>>>> ``rss_hash_update``, ``rss_hash_conf_get``. >>>>>> +* **[uses]     user config**: ``rss_conf.algorithm`` >>>>>> +* **[provides]   rte_eth_dev_info**: ``rss_algo_capa``. >>>>>> +* **[related]    API**: ``rte_eth_dev_configure``, >>>>>> ``rte_eth_dev_rss_hash_update()``, >>>>>>      ``rte_eth_dev_rss_hash_conf_get()``. >>>>>> >>>>> This document describes features listed in the 'default.ini', so we >>>>> shouldn't have above. >>>>> >>>>> And I don't think RSS hash algorithm update is a big enough feature to >>>>> list in the feature list, perhaps it can be embedded in the RSS support >>>>> block, what do you think? >>>> Yes it is not a bit feature. >>>> so put it to RSS hash, right? >>>> >>> Yes please. >>> >>> . > .