From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by smtp.lore.kernel.org (Postfix) with ESMTP id 933FDC05027 for ; Fri, 20 Jan 2023 08:35:43 +0000 (UTC) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 74D1A40150; Fri, 20 Jan 2023 09:35:42 +0100 (CET) Received: from shelob.oktetlabs.ru (shelob.oktetlabs.ru [91.220.146.113]) by mails.dpdk.org (Postfix) with ESMTP id 7A0DE400D5 for ; Fri, 20 Jan 2023 09:35:41 +0100 (CET) Received: from [192.168.38.17] (aros.oktetlabs.ru [192.168.38.17]) (using TLSv1.3 with cipher TLS_AES_128_GCM_SHA256 (128/128 bits) key-exchange X25519 server-signature RSA-PSS (4096 bits) server-digest SHA256) (No client certificate requested) by shelob.oktetlabs.ru (Postfix) with ESMTPSA id A983B50; Fri, 20 Jan 2023 11:35:40 +0300 (MSK) DKIM-Filter: OpenDKIM Filter v2.11.0 shelob.oktetlabs.ru A983B50 DKIM-Signature: v=1; a=rsa-sha256; c=simple/simple; d=oktetlabs.ru; s=default; t=1674203740; bh=iHO0dduNJnCyxroLdAWZa2iqTbDw9SYy0X5zvZScFPo=; h=Date:Subject:To:Cc:References:From:In-Reply-To:From; b=ttw7yboUFB03nJaeFPI3Z0acwOARNjobNRn49KTbErjGFAkhxrdh0ZX6wZ/VkHPLd gSuPN3CT873nrHwB4eMwLdPyzmVgAt6/kbAHPcBzcM2ncH47Dsb9xDVm/b0NvhXX/d WV/mlT9iWZE4diJGMh4qd24wWyzv/H2Z02wSEdvQ= Message-ID: Date: Fri, 20 Jan 2023 11:35:40 +0300 MIME-Version: 1.0 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:102.0) Gecko/20100101 Thunderbird/102.5.0 Subject: Re: [PATCH v6 1/2] ethdev: add query_update sync and async function calls Content-Language: en-US To: Gregory Etelson , dev@dpdk.org Cc: matan@nvidia.com, rasland@nvidia.com, Ori Kam , Thomas Monjalon , Ferruh Yigit References: <20221221073547.988-1-getelson@nvidia.com> <20230119164713.7164-1-getelson@nvidia.com> From: Andrew Rybchenko Organization: OKTET Labs In-Reply-To: <20230119164713.7164-1-getelson@nvidia.com> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit 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 On 1/19/23 19:47, Gregory Etelson wrote: > Current API allows either query or update indirect flow action. > If indirect action must be conditionally updated according to it's > present state application must first issue action query then > analyze returned data and if needed issue update request. > When the update will be processed, action state can change and > the update can invalidate the action. > > The patch adds `rte_flow_action_handle_query_update` function call, "This patch adds ..." -> "Add ..." > and it's async version `rte_flow_async_action_handle_query_update` > to atomically query and update flow action. Sorry, may be I'm missing something, but after reading previous discussion I have a feeling that update could be queried data dependent. However, I don't understand how it could be possible with API below. Just my misunderstanding? > > Application can control query and update order, if that is supported > by port hardware, by setting `qu_mode` parameter to > RTE_FLOW_QU_QUERY_FIRST or RTE_FLOW_QU_UPDATE_FIRST. > > Signed-off-by: Gregory Etelson > --- > v2: Remove RTE_FLOW_QU_DEFAULT query-update mode. > v3: Update release release notes. > Fix doxygen errors. > v4: Add returned errno codes. > v5: Update the patch description. > Fix typos. > v6: Resolve merge conflict with the main branch. > --- > doc/guides/rel_notes/release_23_03.rst | 7 ++ > lib/ethdev/rte_flow.c | 39 ++++++++++ > lib/ethdev/rte_flow.h | 99 ++++++++++++++++++++++++++ > lib/ethdev/rte_flow_driver.h | 15 ++++ > lib/ethdev/version.map | 5 ++ > 5 files changed, 165 insertions(+) > > diff --git a/doc/guides/rel_notes/release_23_03.rst b/doc/guides/rel_notes/release_23_03.rst > index c15f6fbb9f..6941d20abc 100644 > --- a/doc/guides/rel_notes/release_23_03.rst > +++ b/doc/guides/rel_notes/release_23_03.rst > @@ -69,6 +69,13 @@ New Features > ``rte_event_dev_config::nb_single_link_event_port_queues`` parameter > required for eth_rx, eth_tx, crypto and timer eventdev adapters. > > +* **Added functions to atomically query and update indirect flow action.** > + > + Added synchronous and asynchronous functions to atomically query and update > + indirect flow action: > + > + - ``rte_flow_action_handle_query_update`` > + - ``rte_flow_async_action_handle_query_update`` Please, add one more empty line since two should separate sections. > > Removed Items > ------------- > diff --git a/lib/ethdev/rte_flow.c b/lib/ethdev/rte_flow.c > index 7d0c24366c..8b8aa940be 100644 > --- a/lib/ethdev/rte_flow.c > +++ b/lib/ethdev/rte_flow.c > @@ -1883,3 +1883,42 @@ rte_flow_async_action_handle_query(uint16_t port_id, > action_handle, data, user_data, error); > return flow_err(port_id, ret, error); > } > + > +int > +rte_flow_action_handle_query_update(uint16_t port_id, > + struct rte_flow_action_handle *handle, > + const void *update, void *query, > + enum rte_flow_query_update_mode mode, > + struct rte_flow_error *error) > +{ > + struct rte_eth_dev *dev = &rte_eth_devices[port_id]; I know that nearby code does the same, but I still dislike it. If port_id is invalid, it could be out-of-boundary access. Yes, port_id is checked on ops get, but it would be safer to do the assignment after ops check below. Right now there is no bugs, but the future modifications could break it. > + const struct rte_flow_ops *ops = rte_flow_ops_get(port_id, error); > + int ret; > + > + if (!ops || !ops->action_handle_query_update) > + return -ENOTSUP; > + ret = ops->action_handle_query_update(dev, handle, update, query, > + mode, error); > + return flow_err(port_id, ret, error); > +} > + > +int > +rte_flow_async_action_handle_query_update(uint16_t port_id, uint32_t queue_id, > + const struct rte_flow_op_attr *attr, > + struct rte_flow_action_handle *handle, > + const void *update, void *query, > + enum rte_flow_query_update_mode mode, > + void *user_data, > + struct rte_flow_error *error) > +{ > + struct rte_eth_dev *dev = &rte_eth_devices[port_id]; same here > + const struct rte_flow_ops *ops = rte_flow_ops_get(port_id, error); > + int ret; > + > + if (!ops || !ops->async_action_handle_query_update) > + return -ENOTSUP; > + ret = ops->async_action_handle_query_update(dev, queue_id, attr, > + handle, update, query, mode, > + user_data, error); > + return flow_err(port_id, ret, error); > +} > diff --git a/lib/ethdev/rte_flow.h b/lib/ethdev/rte_flow.h > index b60987db4b..280567a3ae 100644 > --- a/lib/ethdev/rte_flow.h > +++ b/lib/ethdev/rte_flow.h > @@ -5622,6 +5622,105 @@ rte_flow_async_action_handle_query(uint16_t port_id, > void *user_data, > struct rte_flow_error *error); > > +/** > + * @warning > + * @b EXPERIMENTAL: this API may change without prior notice. > + * > + * Query_update operational mode. May be "Query and update operation mode."? > + * > + * RTE_FLOW_QU_QUERY_FIRST > + * Force port to query action before update. > + * RTE_FLOW_QU_UPDATE_FIRST > + * Force port to update action before query. > + * > + * @see rte_flow_action_handle_query_update() > + * @see rte_flow_async_action_handle_query_update() > + */ > +enum rte_flow_query_update_mode { > + RTE_FLOW_QU_QUERY_FIRST, /* query before update */ > + RTE_FLOW_QU_UPDATE_FIRST, /* query after update */ > +}; > + > +/** > + * @warning > + * @b EXPERIMENTAL: this API may change without prior notice. > + * > + * Query and/or update indirect flow action. > + * If update parameter is NULL the function queries indirect action. > + * If query parameter is NULL the function updates indirect action. IMHO it is better to highlight in the parameter description below that parameter may be NULL. See below. > + * If both query and update not NULL, the function atomically > + * queries and updates indirect action. Query and update are carried in order > + * specified in the mode parameter. I'd like to double-check what happens if both ops are NULL. I guess it is just NOP. Also it should not be a problem is QUERY_FIRST is specified, but query is NULL. Same for update. If yes in both cases, no extra documentation is required. If it is an error, it should be documented. > + * > + * @param port_id > + * Port identifier of Ethernet device. > + * @param handle > + * Handle for the indirect action object to be updated. > + * @param update > + * Update profile specification used to modify the action pointed by handle. If not NULL, update ... > + * @param query > + * Pointer to storage for the associated query data type. If not NULL, query action data and store in provided location. > + * @param mode > + * Operational mode. > + * @param error > + * Perform verbose error reporting if not NULL. > + * PMDs initialize this structure in case of error only. > + * > + * @return > + * 0 on success, a negative errno value otherwise and rte_errno is set. > + * - (ENOTSUP) if underlying device does not support this functionality. > + * - (EINVAL) - if *handle* invalid > + */ > +__rte_experimental > +int > +rte_flow_action_handle_query_update(uint16_t port_id, > + struct rte_flow_action_handle *handle, > + const void *update, void *query, > + enum rte_flow_query_update_mode mode, > + struct rte_flow_error *error); > + > +/** > + * @warning > + * @b EXPERIMENTAL: this API may change without prior notice. > + * > + * Enqueue async indirect flow action query and/or update > + * > + * @param port_id > + * Port identifier of Ethernet device. > + * @param queue_id > + * Flow queue which is used to update the rule. > + * @param attr > + * Indirect action update operation attributes. > + * @param handle > + * Handle for the indirect action object to be updated. > + * @param update > + * Update profile specification used to modify the action pointed by handle. It should be documented that NULL is allowed > + * @param query > + * Pointer to storage for the associated query data type. > + * Query result returned on async completion event. It should be documented that NULL is allowed > + * @param mode > + * Operational mode. > + * @param user_data > + * The user data that will be returned on async completion event. > + * @param error > + * Perform verbose error reporting if not NULL. > + * PMDs initialize this structure in case of error only. > + * > + * @return > + * 0 on success, a negative errno value otherwise and rte_errno is set. > + * - (ENOTSUP) if underlying device does not support this functionality. > + * - (EINVAL) - if *handle* invalid > + */ > +__rte_experimental > +int > +rte_flow_async_action_handle_query_update(uint16_t port_id, uint32_t queue_id, > + const struct rte_flow_op_attr *attr, > + struct rte_flow_action_handle *handle, > + const void *update, void *query, > + enum rte_flow_query_update_mode mode, > + void *user_data, > + struct rte_flow_error *error); > + > #ifdef __cplusplus > } > #endif [skip]