From patchwork Wed Jun 27 14:55:25 2018 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: =?utf-8?q?N=C3=A9lio_Laranjeiro?= X-Patchwork-Id: 41673 X-Patchwork-Delegate: ferruh.yigit@amd.com Return-Path: X-Original-To: patchwork@dpdk.org Delivered-To: patchwork@dpdk.org Received: from [92.243.14.124] (localhost [127.0.0.1]) by dpdk.org (Postfix) with ESMTP id D0D861BECD; Wed, 27 Jun 2018 16:55:07 +0200 (CEST) Received: from mail-wm0-f65.google.com (mail-wm0-f65.google.com [74.125.82.65]) by dpdk.org (Postfix) with ESMTP id 1EB5D1BF64 for ; Wed, 27 Jun 2018 16:55:06 +0200 (CEST) Received: by mail-wm0-f65.google.com with SMTP id i139-v6so6147899wmf.4 for ; Wed, 27 Jun 2018 07:55:06 -0700 (PDT) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=6wind-com.20150623.gappssmtp.com; s=20150623; h=from:to:subject:date:message-id:in-reply-to:references; bh=j4EJUMJzt9NDT78V+zsXY39TIFa6hRXHPmYPvV86vEk=; b=B9yLr4opidsc2qTtNtbCGezerpRfYWJsNoHI8dU4PhMGK3C765yubU0p7Kaofxsmcw idki8Q7iJDfxsruIz5ksdWIBJTJiQNaIMZ/zAUw/7WVeryNi1vBFDXddckJiKsrq6/0Q ZlNFLMZms6Ag4f3zOwGGw/g0xylsYgijS3l7BR2UDL4D7L/kURzLSMmune0NTDevEGgN 4ZqnW6Hgj7T2iHNMJo5N4g77B3rtknMhBjZ01aU882IOjKg68YH8t/3pS/Ds+OOMhSMG x8OvVyjTSQ3JPiVy6m5+B2TvServV1g02alkwTcfgOea1Ijyzw8upOQY9HAKySTCFQtm T5tg== X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20161025; h=x-gm-message-state:from:to:subject:date:message-id:in-reply-to :references; bh=j4EJUMJzt9NDT78V+zsXY39TIFa6hRXHPmYPvV86vEk=; b=IrLbYQEEzAkIxKKB3uElJvbuJBxuX8zAlczRQSINFc5KCOtmno5KbHcp8S93TP7d3R HxcclaBMIazoQk+CSmDc6RRcWeLCSLDbgGU/qNAdNuqq4Gsj8MOwt3LLewVTGecqwq1U 81LmUl82/6sbYw5LaNLpbvdpg7qijpxgjxhTDf1EbKRYD5i2tMPUYEVHgzcgiXOSETQE yfQtdwKv0whR1TnRNo8SmCrkX489jq01OCf3SEmGBs/aJQNYW2qDud89zsl6Xi6bhKaM etI/aprCAcO6ORmz+Co5a6BZ7XRad0VVbceWHKGKHkdwr1FzJzvnLa7s2l+W5uv3Ep4n ZDtg== X-Gm-Message-State: APt69E3jdMjEp+QF3tVDABhI07yKwhLj+YKXNID+NQdPRouIVvxocKh1 UP7tDkhlfSEQUkogJNISk//JIz2zFQ== X-Google-Smtp-Source: AAOMgpfAxfiQ5f52Cc8fO5ffs4hSvX5krYo8QPKbnx1MimxDbl2YwOvTMo8lJU8YCbgsfQ2i4vXzJA== X-Received: by 2002:a1c:ae94:: with SMTP id x142-v6mr5063811wme.22.1530111305530; Wed, 27 Jun 2018 07:55:05 -0700 (PDT) Received: from laranjeiro-vm.dev.6wind.com (host.78.145.23.62.rev.coltfrance.com. [62.23.145.78]) by smtp.gmail.com with ESMTPSA id d1-v6sm2194359wrm.9.2018.06.27.07.55.04 (version=TLS1_2 cipher=ECDHE-RSA-AES128-GCM-SHA256 bits=128/128); Wed, 27 Jun 2018 07:55:04 -0700 (PDT) From: Nelio Laranjeiro To: dev@dpdk.org, Adrien Mazarguil Date: Wed, 27 Jun 2018 16:55:25 +0200 Message-Id: <20180627145525.8659-1-nelio.laranjeiro@6wind.com> X-Mailer: git-send-email 2.18.0 In-Reply-To: <15e22f72e9b4c56e79809c413ce3001e4f6067d8.1529565844.git.nelio.laranjeiro@6wind.com> References: <15e22f72e9b4c56e79809c413ce3001e4f6067d8.1529565844.git.nelio.laranjeiro@6wind.com> Subject: [dpdk-dev] [PATCH v4] ethdev: add flow API to expand RSS flows X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.15 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org Sender: "dev" Introduce an helper for PMD to expand easily flows items list with RSS action into multiple flow items lists with priority information. For instance a user items list being "eth / end" with rss action types "ipv4-udp ipv6-udp end" needs to be expanded into three items lists: - eth - eth / ipv4 / udp - eth / ipv6 / udp to match the user request. Some drivers are unable to reach such request without this expansion, this API is there to help those. Only PMD should use such API for their internal cooking, the application will still handle a single flow. Signed-off-by: Nelio Laranjeiro --- Changes in v4: - Replace the expanded algorithm with a graph to support also tunnel pattern matching. Changes in v3: - Fix a segmentation fault due to an uninitialized pointer. Changes in v2: - Fix expansion for UDP/TCP layers where L3 may not be in the original items list and thus is missing in the expansion. - Fix size verification for some layers causing a segfault --- lib/librte_ethdev/rte_flow.c | 105 ++++++++++++++++++++++++++++ lib/librte_ethdev/rte_flow_driver.h | 48 +++++++++++++ 2 files changed, 153 insertions(+) diff --git a/lib/librte_ethdev/rte_flow.c b/lib/librte_ethdev/rte_flow.c index 2e87e59f3..96af18e16 100644 --- a/lib/librte_ethdev/rte_flow.c +++ b/lib/librte_ethdev/rte_flow.c @@ -526,3 +526,108 @@ rte_flow_copy(struct rte_flow_desc *desc, size_t len, } return 0; } + +int +rte_flow_expand_rss(struct rte_flow_expand_rss *buf, size_t size, + const struct rte_flow_item *pat, uint64_t types, + const struct rte_flow_expand_node nodes[], + const int node_entry_index) +{ + const int elt_n = 8; + const struct rte_flow_item *item; + const struct rte_flow_expand_node *node = &nodes[node_entry_index]; + const int *next_node; + const int *stack[elt_n]; + int stack_n = 0; + struct rte_flow_item flow_items[elt_n]; + unsigned int i; + size_t lsize; + size_t user_pattern_size = 0; + void *addr = NULL; + + lsize = sizeof(*buf) + + /* Size for the list of patterns. */ + sizeof(*buf->patterns) + + RTE_ALIGN_CEIL(elt_n * sizeof(*item), sizeof(void *)) + + /* Size for priorities. */ + sizeof(*buf->priority) + + RTE_ALIGN_CEIL(elt_n * sizeof(uint32_t), sizeof(void *)); + if (lsize <= size) { + buf->priority = (void *)(buf + 1); + buf->priority[0] = 0; + buf->patterns = (void *)&buf->priority[elt_n]; + buf->patterns[0] = (void *)&buf->patterns[elt_n]; + buf->entries = 0; + addr = buf->patterns[0]; + } + for (item = pat; item->type != RTE_FLOW_ITEM_TYPE_END; item++) { + const struct rte_flow_expand_node *next = NULL; + + for (i = 0; node->next && node->next[i]; ++i) { + next = &nodes[node->next[i]]; + if (next->type == item->type) + break; + } + if (next) + node = next; + user_pattern_size += sizeof(*item); + } + user_pattern_size += sizeof(*item); /**< Handle END item. */ + lsize += user_pattern_size; + /* Copy the user pattern in the first entry of the buffer. */ + if (lsize <= size) { + rte_memcpy(addr, pat, user_pattern_size); + addr = (void *)(((uintptr_t)addr) + user_pattern_size); + buf->priority[buf->entries] = 0; + buf->entries = 1; + } + /* Start expanding. */ + memset(flow_items, 0, sizeof(flow_items)); + user_pattern_size -= sizeof(*item); + next_node = node->next; + stack[stack_n] = next_node; + node = next_node ? &nodes[*next_node] : NULL; + while (node) { + flow_items[stack_n].type = node->type; + if ((node->rss_types & types) == node->rss_types) { + /* + * compute the number of items to copy from the + * expansion and copy it. + * When the stack_n is 0, there are 1 element in it, + * plus the addition END item. + */ + int elt = stack_n + 2; + + flow_items[stack_n + 1].type = RTE_FLOW_ITEM_TYPE_END; + lsize += elt * sizeof(*item) + user_pattern_size; + if (lsize <= size) { + size_t n = elt * sizeof(*item); + + buf->priority[buf->entries] = stack_n + 1; + buf->patterns[buf->entries++] = addr; + rte_memcpy(addr, buf->patterns[0], + user_pattern_size); + addr = (void *)(((uintptr_t)addr) + + user_pattern_size); + rte_memcpy(addr, flow_items, n); + addr = (void *) (((uintptr_t)addr) + n); + } + } + /* Go deeper. */ + if (node->next) { + next_node = node->next; + stack[++stack_n] = next_node; + } else if (*(next_node + 1)) { + /* Follow up with the next possibility. */ + ++next_node; + } else { + /* Move to the next path. */ + if (stack_n) + next_node = stack[--stack_n]; + next_node++; + stack[stack_n] = next_node; + } + node = *next_node ? &nodes[*next_node] : NULL; + }; + return lsize; +} diff --git a/lib/librte_ethdev/rte_flow_driver.h b/lib/librte_ethdev/rte_flow_driver.h index 1c90c600d..538b46e54 100644 --- a/lib/librte_ethdev/rte_flow_driver.h +++ b/lib/librte_ethdev/rte_flow_driver.h @@ -114,6 +114,54 @@ struct rte_flow_ops { const struct rte_flow_ops * rte_flow_ops_get(uint16_t port_id, struct rte_flow_error *error); +/** Macro to create the expansion graph. */ +#define RTE_FLOW_EXPAND_ITEMS(...) \ + (const int []){ \ + __VA_ARGS__, 0, \ + } + +/** structure to generate the expansion graph. */ +struct rte_flow_expand_node { + const int *const next; /**< list of items finalised by 0. */ + const enum rte_flow_item_type type; /**< Item type to add. */ + uint64_t rss_types; /**< RSS bit-field value. */ +}; + +/** + * Expansion structure for RSS flows. + */ +struct rte_flow_expand_rss { + uint32_t entries; /**< Number of entries in the following arrays. */ + struct rte_flow_item **patterns; /**< Expanded pattern array. */ + uint32_t *priority; /**< Priority offset for each expansion. */ +}; + +/** + * Expand RSS flows into several possible flows according to the RSS hash + * fields requested and the driver capabilities. + * + * @param[in,out] buf + * Buffer to store the result expansion. + * @param[in] size + * Size in octets of the buffer. + * @param[in] pat + * User flow pattern. + * @param[in] types + * RSS types expected (see ETH_RSS_*). + * @param[in] nodes. + * Expansion graph of possibilities for the RSS. + * @param[in] node_entry_index + * The index in the \p nodes array as start point. + * + * @return + * The size in octets used to expand. + */ +int +rte_flow_expand_rss(struct rte_flow_expand_rss *buf, size_t size, + const struct rte_flow_item *pat, uint64_t types, + const struct rte_flow_expand_node nodes[], + const int node_entry_index); + #ifdef __cplusplus } #endif