crypto/ipsec_mb: add chacha_poly PMD
[dpdk.git] / doc / guides / sample_app_ug / flow_filtering.rst
index 5e5a6cd..11da914 100644 (file)
@@ -13,28 +13,7 @@ It is intended as a demonstration of the basic components RTE flow rules.
 Compiling the Application
 -------------------------
 
-To compile the application export the path to the DPDK source tree and go to
-the example directory:
-
-.. code-block:: console
-
-    export RTE_SDK=/path/to/rte_sdk
-
-    cd ${RTE_SDK}/examples/flow_filtering
-
-Set the target, for example:
-
-.. code-block:: console
-
-    export RTE_TARGET=x86_64-native-linux-gcc
-
-See the *DPDK Getting Started* Guide for possible ``RTE_TARGET`` values.
-
-Build the application as follows:
-
-.. code-block:: console
-
-    make
+To compile the sample application see :doc:`compiling`.
 
 
 Running the Application
@@ -44,7 +23,7 @@ To run the example in a ``linux`` environment:
 
 .. code-block:: console
 
-    ./build/flow -l 1 -n 1
+    ./<build_dir>/examples/dpdk-flow_filtering -l 1 -n 1
 
 Refer to *DPDK Getting Started Guide* for general information on running
 applications and the Environment Abstraction Layer (EAL) options.
@@ -74,21 +53,21 @@ The first task is to initialize the Environment Abstraction Layer (EAL).  The
 ``argc`` and ``argv`` arguments are provided to the ``rte_eal_init()``
 function. The value returned is the number of parsed arguments:
 
-.. code-block:: c
-
-    int ret = rte_eal_init(argc, argv);
-    if (ret < 0)
-        rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Initialize EAL. 8<
+    :end-before: >8 End of Initialization of EAL.
+    :dedent: 1
 
 
 The ``main()`` also allocates a mempool to hold the mbufs (Message Buffers)
 used by the application:
 
-.. code-block:: c
-
-   mbuf_pool = rte_pktmbuf_pool_create("mbuf_pool", 4096, 128, 0,
-                                            RTE_MBUF_DEFAULT_BUF_SIZE,
-                                            rte_socket_id());
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Allocates a mempool to hold the mbufs. 8<
+    :end-before: >8 End of allocating a mempool to hold the mbufs.
+    :dedent: 1
 
 Mbufs are the packet buffer structure used by DPDK. They are explained in
 detail in the "Mbuf Library" section of the *DPDK Programmer's Guide*.
@@ -96,33 +75,30 @@ detail in the "Mbuf Library" section of the *DPDK Programmer's Guide*.
 The ``main()`` function also initializes all the ports using the user defined
 ``init_port()`` function which is explained in the next section:
 
-.. code-block:: c
-
-   init_port();
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Initializes all the ports using the user defined init_port(). 8<
+    :end-before: >8 End of Initializing the ports using user defined init_port().
+    :dedent: 1
 
 Once the initialization is complete, we set the flow rule using the
 following code:
 
-.. code-block:: c
-
-   /* create flow for send packet with */
-   flow = generate_ipv4_flow(port_id, selected_queue,
-                                SRC_IP, EMPTY_MASK,
-                                DEST_IP, FULL_MASK, &error);
-   if (!flow) {
-          printf("Flow can't be created %d message: %s\n",
-                       error.type,
-                       error.message ? error.message : "(no stated reason)");
-          rte_exit(EXIT_FAILURE, "error in creating flow");
-   }
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Create flow for send packet with. 8<
+    :end-before: >8 End of creating flow for send packet with.
+    :dedent: 1
 
 In the last part the application is ready to launch the
 ``main_loop()`` function. Which is explained below.
 
 
-.. code-block:: c
-
-   main_loop();
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Launching main_loop(). 8<
+    :end-before: >8 End of launching main_loop().
+    :dedent: 1
 
 The Port Initialization  Function
 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
@@ -130,176 +106,54 @@ The Port Initialization  Function
 The main functional part of the port initialization used in the flow filtering
 application is shown below:
 
-.. code-block:: c
-
-   init_port(void)
-   {
-           int ret;
-           uint16_t i;
-           struct rte_eth_conf port_conf = {
-                   .rxmode = {
-                           .split_hdr_size = 0,
-                           },
-                   .txmode = {
-                           .offloads =
-                                   DEV_TX_OFFLOAD_VLAN_INSERT |
-                                   DEV_TX_OFFLOAD_IPV4_CKSUM  |
-                                   DEV_TX_OFFLOAD_UDP_CKSUM   |
-                                   DEV_TX_OFFLOAD_TCP_CKSUM   |
-                                   DEV_TX_OFFLOAD_SCTP_CKSUM  |
-                                   DEV_TX_OFFLOAD_TCP_TSO,
-                   },
-           };
-           struct rte_eth_txconf txq_conf;
-           struct rte_eth_rxconf rxq_conf;
-           struct rte_eth_dev_info dev_info;
-
-           printf(":: initializing port: %d\n", port_id);
-           ret = rte_eth_dev_configure(port_id,
-                   nr_queues, nr_queues, &port_conf);
-           if (ret < 0) {
-                   rte_exit(EXIT_FAILURE,
-                           ":: cannot configure device: err=%d, port=%u\n",
-                           ret, port_id);
-           }
-
-           rte_eth_dev_info_get(port_id, &dev_info);
-           rxq_conf = dev_info.default_rxconf;
-           rxq_conf.offloads = port_conf.rxmode.offloads;
-           /* only set Rx queues: something we care only so far */
-           for (i = 0; i < nr_queues; i++) {
-                   ret = rte_eth_rx_queue_setup(port_id, i, 512,
-                           rte_eth_dev_socket_id(port_id),
-                           &rxq_conf,
-                           mbuf_pool);
-                   if (ret < 0) {
-                            rte_exit(EXIT_FAILURE,
-                                    ":: Rx queue setup failed: err=%d, port=%u\n",
-                                    ret, port_id);
-                   }
-           }
-
-           txq_conf = dev_info.default_txconf;
-           txq_conf.offloads = port_conf.txmode.offloads;
-
-           for (i = 0; i < nr_queues; i++) {
-                   ret = rte_eth_tx_queue_setup(port_id, i, 512,
-                           rte_eth_dev_socket_id(port_id),
-                           &txq_conf);
-                   if (ret < 0) {
-                           rte_exit(EXIT_FAILURE,
-                                   ":: Tx queue setup failed: err=%d, port=%u\n",
-                                   ret, port_id);
-                   }
-          }
-
-           ret = rte_eth_promiscuous_enable(port_id);
-           if (ret != 0) {
-                   rte_exit(EXIT_FAILURE,
-                           ":: cannot enable promiscuous mode: err=%d, port=%u\n",
-                           ret, port_id);
-           }
-
-           ret = rte_eth_dev_start(port_id);
-           if (ret < 0) {
-                   rte_exit(EXIT_FAILURE,
-                           "rte_eth_dev_start:err=%d, port=%u\n",
-                           ret, port_id);
-           }
-
-           assert_link_status();
-
-           printf(":: initializing port: %d done\n", port_id);
-   }
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Port initialization used in flow filtering. 8<
+    :end-before: >8 End of Port initialization used in flow filtering.
 
 The Ethernet port is configured with default settings using the
 ``rte_eth_dev_configure()`` function and the ``port_conf_default`` struct:
 
-.. code-block:: c
-
-   struct rte_eth_conf port_conf = {
-           .rxmode = {
-                   .split_hdr_size = 0,
-                   },
-           .txmode = {
-                   .offloads =
-                           DEV_TX_OFFLOAD_VLAN_INSERT |
-                           DEV_TX_OFFLOAD_IPV4_CKSUM  |
-                           DEV_TX_OFFLOAD_UDP_CKSUM   |
-                           DEV_TX_OFFLOAD_TCP_CKSUM   |
-                           DEV_TX_OFFLOAD_SCTP_CKSUM  |
-                           DEV_TX_OFFLOAD_TCP_TSO,
-                   },
-           };
-
-   ret = rte_eth_dev_configure(port_id, nr_queues, nr_queues, &port_conf);
-   if (ret < 0) {
-        rte_exit(EXIT_FAILURE,
-                 ":: cannot configure device: err=%d, port=%u\n",
-                 ret, port_id);
-   }
-   rte_eth_dev_info_get(port_id, &dev_info);
-   rxq_conf = dev_info.default_rxconf;
-   rxq_conf.offloads = port_conf.rxmode.offloads;
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Ethernet port configured with default settings. 8<
+    :end-before: >8 End of ethernet port configured with default settings.
+    :dedent: 1
 
 For this example we are configuring number of rx and tx queues that are connected
 to a single port.
 
-.. code-block:: c
-
-   for (i = 0; i < nr_queues; i++) {
-          ret = rte_eth_rx_queue_setup(port_id, i, 512,
-                                       rte_eth_dev_socket_id(port_id),
-                                       &rxq_conf,
-                                       mbuf_pool);
-          if (ret < 0) {
-                  rte_exit(EXIT_FAILURE,
-                          ":: Rx queue setup failed: err=%d, port=%u\n",
-                          ret, port_id);
-          }
-   }
-
-   for (i = 0; i < nr_queues; i++) {
-          ret = rte_eth_tx_queue_setup(port_id, i, 512,
-                                       rte_eth_dev_socket_id(port_id),
-                                       &txq_conf);
-          if (ret < 0) {
-                  rte_exit(EXIT_FAILURE,
-                           ":: Tx queue setup failed: err=%d, port=%u\n",
-                           ret, port_id);
-          }
-   }
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Configuring number of RX and TX queues connected to single port. 8<
+    :end-before: >8 End of Configuring RX and TX queues connected to single port.
+    :dedent: 1
 
 In the next step we create and apply the flow rule. which is to send packets
 with destination ip equals to 192.168.1.1 to queue number 1. The detail
 explanation of the ``generate_ipv4_flow()`` appears later in this document:
 
-.. code-block:: c
-
-   flow = generate_ipv4_flow(port_id, selected_queue,
-                             SRC_IP, EMPTY_MASK,
-                             DEST_IP, FULL_MASK, &error);
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Create flow for send packet with. 8<
+    :end-before: >8 End of create flow and the flow rule.
+    :dedent: 1
 
 We are setting the RX port to promiscuous mode:
 
-.. code-block:: c
-
-   ret = rte_eth_promiscuous_enable(port_id);
-   if (ret != 0) {
-        rte_exit(EXIT_FAILURE,
-                 ":: cannot enable promiscuous mode: err=%d, port=%u\n",
-                 ret, port_id);
-   }
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Setting the RX port to promiscuous mode. 8<
+    :end-before: >8 End of setting the RX port to promiscuous mode.
+    :dedent: 1
 
 The last step is to start the port.
 
-.. code-block:: c
-
-   ret = rte_eth_dev_start(port_id);
-   if (ret < 0)  {
-        rte_exit(EXIT_FAILURE, "rte_eth_dev_start:err%d, port=%u\n",
-                        ret, port_id);
-   }
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Starting the port. 8<
+    :end-before: >8 End of starting the port.
+    :dedent: 1
 
 
 The main_loop function
@@ -309,66 +163,18 @@ As we saw above the ``main()`` function calls an application function to handle
 the main loop. For the flow filtering application the main_loop function
 looks like the following:
 
-.. code-block:: c
-
-   static void
-   main_loop(void)
-   {
-           struct rte_mbuf *mbufs[32];
-           struct rte_ether_hdr *eth_hdr;
-           uint16_t nb_rx;
-           uint16_t i;
-           uint16_t j;
-
-           while (!force_quit) {
-                   for (i = 0; i < nr_queues; i++) {
-                           nb_rx = rte_eth_rx_burst(port_id,
-                                                   i, mbufs, 32);
-                           if (nb_rx) {
-                                   for (j = 0; j < nb_rx; j++) {
-                                           struct rte_mbuf *m = mbufs[j];
-
-                                           eth_hdr = rte_pktmbuf_mtod(m,
-                                                        struct rte_ether_hdr *);
-                                           print_ether_addr("src=",
-                                                        &eth_hdr->s_addr);
-                                           print_ether_addr(" - dst=",
-                                                        &eth_hdr->d_addr);
-                                           printf(" - queue=0x%x",
-                                                           (unsigned int)i);
-                                           printf("\n");
-                                           rte_pktmbuf_free(m);
-                                   }
-                           }
-                   }
-           }
-           /* closing and releasing resources */
-           rte_flow_flush(port_id, &error);
-           rte_eth_dev_stop(port_id);
-           rte_eth_dev_close(port_id);
-   }
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Main_loop for flow filtering. 8<
+    :end-before: >8 End of reading the packets from all queues.
 
 The main work of the application is reading the packets from all
 queues and printing for each packet the destination queue:
 
-.. code-block:: c
-
-    while (!force_quit) {
-        for (i = 0; i < nr_queues; i++) {
-                   nb_rx = rte_eth_rx_burst(port_id, i, mbufs, 32);
-                if (nb_rx) {
-                        for (j = 0; j < nb_rx; j++) {
-                             struct rte_mbuf *m = mbufs[j];
-                             eth_hdr = rte_pktmbuf_mtod(m, struct rte_ether_hdr *);
-                             print_ether_addr("src=", &eth_hdr->s_addr);
-                             print_ether_addr(" - dst=", &eth_hdr->d_addr);
-                             printf(" - queue=0x%x", (unsigned int)i);
-                             printf("\n");
-                             rte_pktmbuf_free(m);
-                        }
-                }
-           }
-    }
+.. literalinclude:: ../../../examples/flow_filtering/main.c
+    :language: c
+    :start-after: Reading the packets from all queues. 8<
+    :end-before: >8 End of main_loop for flow filtering.
 
 
 The forwarding loop can be interrupted and the application closed using
@@ -381,100 +187,34 @@ The generate_ipv4_flow function
 The generate_ipv4_flow function is responsible for creating the flow rule.
 This function is located in the ``flow_blocks.c`` file.
 
-.. code-block:: c
-
-   static struct rte_flow *
-   generate_ipv4_flow(uint8_t port_id, uint16_t rx_q,
-                   uint32_t src_ip, uint32_t src_mask,
-                   uint32_t dest_ip, uint32_t dest_mask,
-                   struct rte_flow_error *error)
-   {
-           struct rte_flow_attr attr;
-           struct rte_flow_item pattern[MAX_PATTERN_NUM];
-           struct rte_flow_action action[MAX_ACTION_NUM];
-           struct rte_flow *flow = NULL;
-           struct rte_flow_action_queue queue = { .index = rx_q };
-           struct rte_flow_item_ipv4 ip_spec;
-           struct rte_flow_item_ipv4 ip_mask;
-
-           memset(pattern, 0, sizeof(pattern));
-           memset(action, 0, sizeof(action));
-
-           /*
-            * set the rule attribute.
-            * in this case only ingress packets will be checked.
-            */
-           memset(&attr, 0, sizeof(struct rte_flow_attr));
-           attr.ingress = 1;
-
-           /*
-            * create the action sequence.
-            * one action only,  move packet to queue
-            */
-           action[0].type = RTE_FLOW_ACTION_TYPE_QUEUE;
-           action[0].conf = &queue;
-           action[1].type = RTE_FLOW_ACTION_TYPE_END;
-
-           /*
-            * set the first level of the pattern (ETH).
-            * since in this example we just want to get the
-            * ipv4 we set this level to allow all.
-            */
-           pattern[0].type = RTE_FLOW_ITEM_TYPE_ETH;
-
-           /*
-            * setting the second level of the pattern (IP).
-            * in this example this is the level we care about
-            * so we set it according to the parameters.
-            */
-           memset(&ip_spec, 0, sizeof(struct rte_flow_item_ipv4));
-           memset(&ip_mask, 0, sizeof(struct rte_flow_item_ipv4));
-           ip_spec.hdr.dst_addr = htonl(dest_ip);
-           ip_mask.hdr.dst_addr = dest_mask;
-           ip_spec.hdr.src_addr = htonl(src_ip);
-           ip_mask.hdr.src_addr = src_mask;
-           pattern[1].type = RTE_FLOW_ITEM_TYPE_IPV4;
-           pattern[1].spec = &ip_spec;
-           pattern[1].mask = &ip_mask;
-
-           /* the final level must be always type end */
-           pattern[2].type = RTE_FLOW_ITEM_TYPE_END;
-
-           int res = rte_flow_validate(port_id, &attr, pattern, action, error);
-           if(!res)
-               flow = rte_flow_create(port_id, &attr, pattern, action, error);
-
-           return flow;
-   }
+.. literalinclude:: ../../../examples/flow_filtering/flow_blocks.c
+    :language: c
+    :start-after: Function responsible for creating the flow rule. 8<
+    :end-before: >8 End of function responsible for creating the flow rule.
 
 The first part of the function is declaring the structures that will be used.
 
-.. code-block:: c
-
-   struct rte_flow_attr attr;
-   struct rte_flow_item pattern[MAX_PATTERN_NUM];
-   struct rte_flow_action action[MAX_ACTION_NUM];
-   struct rte_flow *flow;
-   struct rte_flow_error error;
-   struct rte_flow_action_queue queue = { .index = rx_q };
-   struct rte_flow_item_ipv4 ip_spec;
-   struct rte_flow_item_ipv4 ip_mask;
+.. literalinclude:: ../../../examples/flow_filtering/flow_blocks.c
+    :language: c
+    :start-after: Declaring structs being used. 8<
+    :end-before: >8 End of declaring structs being used.
+    :dedent: 1
 
 The following part create the flow attributes, in our case ingress.
 
-.. code-block:: c
-
-   memset(&attr, 0, sizeof(struct rte_flow_attr));
-   attr.ingress = 1;
+.. literalinclude:: ../../../examples/flow_filtering/flow_blocks.c
+    :language: c
+    :start-after: Set the rule attribute, only ingress packets will be checked. 8<
+    :end-before: >8 End of setting the rule attribute.
+    :dedent: 1
 
 The third part defines the action to be taken when a packet matches
 the rule. In this case send the packet to queue.
 
-.. code-block:: c
-
-   action[0].type = RTE_FLOW_ACTION_TYPE_QUEUE;
-   action[0].conf = &queue;
-   action[1].type = RTE_FLOW_ACTION_TYPE_END;
+.. literalinclude:: ../../../examples/flow_filtering/flow_blocks.c
+    :language: c
+    :start-after: Function responsible for creating the flow rule. 8<
+    :end-before: >8 End of setting the rule attribute.
 
 The fourth part is responsible for creating the pattern and is built from
 number of steps. In each step we build one level of the pattern starting with
@@ -482,34 +222,32 @@ the lowest one.
 
 Setting the first level of the pattern ETH:
 
-.. code-block:: c
-
-   pattern[0].type = RTE_FLOW_ITEM_TYPE_ETH;
+.. literalinclude:: ../../../examples/flow_filtering/flow_blocks.c
+    :language: c
+    :start-after: IPv4 we set this level to allow all. 8<
+    :end-before: >8 End of setting the first level of the pattern.
+    :dedent: 1
 
 Setting the second level of the pattern IP:
 
-.. code-block:: c
-
-   memset(&ip_spec, 0, sizeof(struct rte_flow_item_ipv4));
-   memset(&ip_mask, 0, sizeof(struct rte_flow_item_ipv4));
-   ip_spec.hdr.dst_addr = htonl(dest_ip);
-   ip_mask.hdr.dst_addr = dest_mask;
-   ip_spec.hdr.src_addr = htonl(src_ip);
-   ip_mask.hdr.src_addr = src_mask;
-   pattern[1].type = RTE_FLOW_ITEM_TYPE_IPV4;
-   pattern[1].spec = &ip_spec;
-   pattern[1].mask = &ip_mask;
+.. literalinclude:: ../../../examples/flow_filtering/flow_blocks.c
+    :language: c
+    :start-after: Setting the second level of the pattern. 8<
+    :end-before: >8 End of setting the second level of the pattern.
+    :dedent: 1
 
 Closing the pattern part.
 
-.. code-block:: c
-
-   pattern[2].type = RTE_FLOW_ITEM_TYPE_END;
+.. literalinclude:: ../../../examples/flow_filtering/flow_blocks.c
+    :language: c
+    :start-after: The final level must be always type end. 8<
+    :end-before: >8 End of final level must be always type end.
+    :dedent: 1
 
 The last part of the function is to validate the rule and create it.
 
-.. code-block:: c
-
-   int res = rte_flow_validate(port_id, &attr, pattern, action, &error);
-   if (!res)
-        flow = rte_flow_create(port_id, &attr, pattern, action, &error);
+.. literalinclude:: ../../../examples/flow_filtering/flow_blocks.c
+    :language: c
+    :start-after: Validate the rule and create it. 8<
+    :end-before: >8 End of validation the rule and create it.
+    :dedent: 1