examples/flow_classify: add sample application
authorBernard Iremonger <bernard.iremonger@intel.com>
Tue, 24 Oct 2017 17:28:01 +0000 (18:28 +0100)
committerThomas Monjalon <thomas@monjalon.net>
Tue, 24 Oct 2017 20:26:54 +0000 (22:26 +0200)
The flow_classify sample application exercises the following
librte_flow_classify API's:

rte_flow_classifier_create
rte_flow_classifier_query
rte_flow_classify_table_create
rte_flow_classify_table_entry_add
rte_flow_classify_table_entry_delete

It sets up the IPv4 ACL field definitions.
It creates table_acl and adds and deletes rules using the
librte_table API.

It uses a file of IPv4 five tuple rules for input.

Signed-off-by: Bernard Iremonger <bernard.iremonger@intel.com>
Acked-by: Jasvinder Singh <jasvinder.singh@intel.com>
examples/Makefile
examples/flow_classify/Makefile [new file with mode: 0644]
examples/flow_classify/flow_classify.c [new file with mode: 0644]
examples/flow_classify/ipv4_rules_file.txt [new file with mode: 0644]

index d27eddd..3b3dbbc 100644 (file)
@@ -43,6 +43,7 @@ DIRS-$(CONFIG_RTE_LIBRTE_DISTRIBUTOR) += distributor
 DIRS-y += ethtool
 DIRS-y += exception_path
 DIRS-$(CONFIG_RTE_LIBRTE_EFD) += server_node_efd
+DIRS-$(CONFIG_RTE_LIBRTE_FLOW_CLASSIFY) += flow_classify
 DIRS-y += helloworld
 DIRS-$(CONFIG_RTE_LIBRTE_PIPELINE) += ip_pipeline
 ifeq ($(CONFIG_RTE_LIBRTE_LPM),y)
diff --git a/examples/flow_classify/Makefile b/examples/flow_classify/Makefile
new file mode 100644 (file)
index 0000000..eecdde1
--- /dev/null
@@ -0,0 +1,57 @@
+#   BSD LICENSE
+#
+#   Copyright(c) 2017 Intel Corporation. All rights reserved.
+#   All rights reserved.
+#
+#   Redistribution and use in source and binary forms, with or without
+#   modification, are permitted provided that the following conditions
+#   are met:
+#
+#     * Redistributions of source code must retain the above copyright
+#       notice, this list of conditions and the following disclaimer.
+#     * Redistributions in binary form must reproduce the above copyright
+#       notice, this list of conditions and the following disclaimer in
+#       the documentation and/or other materials provided with the
+#       distribution.
+#     * Neither the name of Intel Corporation nor the names of its
+#       contributors may be used to endorse or promote products derived
+#       from this software without specific prior written permission.
+#
+#   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+#   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+#   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+#   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+#   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+#   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+#   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+#   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+#   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+#   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+#   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+
+ifeq ($(RTE_SDK),)
+$(error "Please define RTE_SDK environment variable")
+endif
+
+# Default target, can be overridden by command line or environment
+RTE_TARGET ?= x86_64-native-linuxapp-gcc
+
+include $(RTE_SDK)/mk/rte.vars.mk
+
+# binary name
+APP = flow_classify
+
+
+# all source are stored in SRCS-y
+SRCS-y := flow_classify.c
+
+CFLAGS += -O3
+CFLAGS += $(WERROR_FLAGS)
+
+# workaround for a gcc bug with noreturn attribute
+# http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12603
+ifeq ($(CONFIG_RTE_TOOLCHAIN_GCC),y)
+CFLAGS_main.o += -Wno-return-type
+endif
+
+include $(RTE_SDK)/mk/rte.extapp.mk
diff --git a/examples/flow_classify/flow_classify.c b/examples/flow_classify/flow_classify.c
new file mode 100644 (file)
index 0000000..e4cabdb
--- /dev/null
@@ -0,0 +1,848 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2017 Intel Corporation. All rights reserved.
+ *   All rights reserved.
+ *
+ *   Redistribution and use in source and binary forms, with or without
+ *   modification, are permitted provided that the following conditions
+ *   are met:
+ *
+ *     * Redistributions of source code must retain the above copyright
+ *       notice, this list of conditions and the following disclaimer.
+ *     * Redistributions in binary form must reproduce the above copyright
+ *       notice, this list of conditions and the following disclaimer in
+ *       the documentation and/or other materials provided with the
+ *       distribution.
+ *     * Neither the name of Intel Corporation nor the names of its
+ *       contributors may be used to endorse or promote products derived
+ *       from this software without specific prior written permission.
+ *
+ *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
+ *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
+ *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
+ *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
+ *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
+ *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
+ *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
+ *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
+ *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <stdint.h>
+#include <inttypes.h>
+#include <getopt.h>
+
+#include <rte_eal.h>
+#include <rte_ethdev.h>
+#include <rte_cycles.h>
+#include <rte_lcore.h>
+#include <rte_mbuf.h>
+#include <rte_flow.h>
+#include <rte_flow_classify.h>
+#include <rte_table_acl.h>
+
+#define RX_RING_SIZE 128
+#define TX_RING_SIZE 512
+
+#define NUM_MBUFS 8191
+#define MBUF_CACHE_SIZE 250
+#define BURST_SIZE 32
+
+#define MAX_NUM_CLASSIFY 30
+#define FLOW_CLASSIFY_MAX_RULE_NUM 91
+#define FLOW_CLASSIFY_MAX_PRIORITY 8
+#define FLOW_CLASSIFIER_NAME_SIZE 64
+
+#define COMMENT_LEAD_CHAR      ('#')
+#define OPTION_RULE_IPV4       "rule_ipv4"
+#define RTE_LOGTYPE_FLOW_CLASSIFY      RTE_LOGTYPE_USER3
+#define flow_classify_log(format, ...) \
+               RTE_LOG(ERR, FLOW_CLASSIFY, format, ##__VA_ARGS__)
+
+#define uint32_t_to_char(ip, a, b, c, d) do {\
+               *a = (unsigned char)(ip >> 24 & 0xff);\
+               *b = (unsigned char)(ip >> 16 & 0xff);\
+               *c = (unsigned char)(ip >> 8 & 0xff);\
+               *d = (unsigned char)(ip & 0xff);\
+       } while (0)
+
+enum {
+       CB_FLD_SRC_ADDR,
+       CB_FLD_DST_ADDR,
+       CB_FLD_SRC_PORT,
+       CB_FLD_SRC_PORT_DLM,
+       CB_FLD_SRC_PORT_MASK,
+       CB_FLD_DST_PORT,
+       CB_FLD_DST_PORT_DLM,
+       CB_FLD_DST_PORT_MASK,
+       CB_FLD_PROTO,
+       CB_FLD_PRIORITY,
+       CB_FLD_NUM,
+};
+
+static struct{
+       const char *rule_ipv4_name;
+} parm_config;
+const char cb_port_delim[] = ":";
+
+static const struct rte_eth_conf port_conf_default = {
+       .rxmode = { .max_rx_pkt_len = ETHER_MAX_LEN }
+};
+
+struct flow_classifier {
+       struct rte_flow_classifier *cls;
+       uint32_t table_id[RTE_FLOW_CLASSIFY_TABLE_MAX];
+};
+
+struct flow_classifier_acl {
+       struct flow_classifier cls;
+} __rte_cache_aligned;
+
+/* ACL field definitions for IPv4 5 tuple rule */
+
+enum {
+       PROTO_FIELD_IPV4,
+       SRC_FIELD_IPV4,
+       DST_FIELD_IPV4,
+       SRCP_FIELD_IPV4,
+       DSTP_FIELD_IPV4,
+       NUM_FIELDS_IPV4
+};
+
+enum {
+       PROTO_INPUT_IPV4,
+       SRC_INPUT_IPV4,
+       DST_INPUT_IPV4,
+       SRCP_DESTP_INPUT_IPV4
+};
+
+static struct rte_acl_field_def ipv4_defs[NUM_FIELDS_IPV4] = {
+       /* first input field - always one byte long. */
+       {
+               .type = RTE_ACL_FIELD_TYPE_BITMASK,
+               .size = sizeof(uint8_t),
+               .field_index = PROTO_FIELD_IPV4,
+               .input_index = PROTO_INPUT_IPV4,
+               .offset = sizeof(struct ether_hdr) +
+                       offsetof(struct ipv4_hdr, next_proto_id),
+       },
+       /* next input field (IPv4 source address) - 4 consecutive bytes. */
+       {
+               /* rte_flow uses a bit mask for IPv4 addresses */
+               .type = RTE_ACL_FIELD_TYPE_BITMASK,
+               .size = sizeof(uint32_t),
+               .field_index = SRC_FIELD_IPV4,
+               .input_index = SRC_INPUT_IPV4,
+               .offset = sizeof(struct ether_hdr) +
+                       offsetof(struct ipv4_hdr, src_addr),
+       },
+       /* next input field (IPv4 destination address) - 4 consecutive bytes. */
+       {
+               /* rte_flow uses a bit mask for IPv4 addresses */
+               .type = RTE_ACL_FIELD_TYPE_BITMASK,
+               .size = sizeof(uint32_t),
+               .field_index = DST_FIELD_IPV4,
+               .input_index = DST_INPUT_IPV4,
+               .offset = sizeof(struct ether_hdr) +
+                       offsetof(struct ipv4_hdr, dst_addr),
+       },
+       /*
+        * Next 2 fields (src & dst ports) form 4 consecutive bytes.
+        * They share the same input index.
+        */
+       {
+               /* rte_flow uses a bit mask for protocol ports */
+               .type = RTE_ACL_FIELD_TYPE_BITMASK,
+               .size = sizeof(uint16_t),
+               .field_index = SRCP_FIELD_IPV4,
+               .input_index = SRCP_DESTP_INPUT_IPV4,
+               .offset = sizeof(struct ether_hdr) +
+                       sizeof(struct ipv4_hdr) +
+                       offsetof(struct tcp_hdr, src_port),
+       },
+       {
+               /* rte_flow uses a bit mask for protocol ports */
+               .type = RTE_ACL_FIELD_TYPE_BITMASK,
+               .size = sizeof(uint16_t),
+               .field_index = DSTP_FIELD_IPV4,
+               .input_index = SRCP_DESTP_INPUT_IPV4,
+               .offset = sizeof(struct ether_hdr) +
+                       sizeof(struct ipv4_hdr) +
+                       offsetof(struct tcp_hdr, dst_port),
+       },
+};
+
+/* flow classify data */
+static int num_classify_rules;
+static struct rte_flow_classify_rule *rules[MAX_NUM_CLASSIFY];
+static struct rte_flow_classify_ipv4_5tuple_stats ntuple_stats;
+static struct rte_flow_classify_stats classify_stats = {
+               .stats = (void **)&ntuple_stats
+};
+
+/* parameters for rte_flow_classify_validate and
+ * rte_flow_classify_table_entry_add functions
+ */
+
+static struct rte_flow_item  eth_item = { RTE_FLOW_ITEM_TYPE_ETH,
+       0, 0, 0 };
+static struct rte_flow_item  end_item = { RTE_FLOW_ITEM_TYPE_END,
+       0, 0, 0 };
+
+/* sample actions:
+ * "actions count / end"
+ */
+static struct rte_flow_action count_action = { RTE_FLOW_ACTION_TYPE_COUNT, 0};
+static struct rte_flow_action end_action = { RTE_FLOW_ACTION_TYPE_END, 0};
+static struct rte_flow_action actions[2];
+
+/* sample attributes */
+static struct rte_flow_attr attr;
+
+/* flow_classify.c: * Based on DPDK skeleton forwarding example. */
+
+/*
+ * Initializes a given port using global settings and with the RX buffers
+ * coming from the mbuf_pool passed as a parameter.
+ */
+static inline int
+port_init(uint8_t port, struct rte_mempool *mbuf_pool)
+{
+       struct rte_eth_conf port_conf = port_conf_default;
+       struct ether_addr addr;
+       const uint16_t rx_rings = 1, tx_rings = 1;
+       int retval;
+       uint16_t q;
+
+       if (port >= rte_eth_dev_count())
+               return -1;
+
+       /* Configure the Ethernet device. */
+       retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
+       if (retval != 0)
+               return retval;
+
+       /* Allocate and set up 1 RX queue per Ethernet port. */
+       for (q = 0; q < rx_rings; q++) {
+               retval = rte_eth_rx_queue_setup(port, q, RX_RING_SIZE,
+                               rte_eth_dev_socket_id(port), NULL, mbuf_pool);
+               if (retval < 0)
+                       return retval;
+       }
+
+       /* Allocate and set up 1 TX queue per Ethernet port. */
+       for (q = 0; q < tx_rings; q++) {
+               retval = rte_eth_tx_queue_setup(port, q, TX_RING_SIZE,
+                               rte_eth_dev_socket_id(port), NULL);
+               if (retval < 0)
+                       return retval;
+       }
+
+       /* Start the Ethernet port. */
+       retval = rte_eth_dev_start(port);
+       if (retval < 0)
+               return retval;
+
+       /* Display the port MAC address. */
+       rte_eth_macaddr_get(port, &addr);
+       printf("Port %u MAC: %02" PRIx8 " %02" PRIx8 " %02" PRIx8
+                          " %02" PRIx8 " %02" PRIx8 " %02" PRIx8 "\n",
+                       port,
+                       addr.addr_bytes[0], addr.addr_bytes[1],
+                       addr.addr_bytes[2], addr.addr_bytes[3],
+                       addr.addr_bytes[4], addr.addr_bytes[5]);
+
+       /* Enable RX in promiscuous mode for the Ethernet device. */
+       rte_eth_promiscuous_enable(port);
+
+       return 0;
+}
+
+/*
+ * The lcore main. This is the main thread that does the work, reading from
+ * an input port classifying the packets and writing to an output port.
+ */
+static __attribute__((noreturn)) void
+lcore_main(struct flow_classifier *cls_app)
+{
+       const uint8_t nb_ports = rte_eth_dev_count();
+       uint8_t port;
+       int ret;
+       int i = 0;
+
+       ret = rte_flow_classify_table_entry_delete(cls_app->cls,
+                       cls_app->table_id[0], rules[7]);
+       if (ret)
+               printf("table_entry_delete failed [7] %d\n\n", ret);
+       else
+               printf("table_entry_delete succeeded [7]\n\n");
+
+       /*
+        * Check that the port is on the same NUMA node as the polling thread
+        * for best performance.
+        */
+       for (port = 0; port < nb_ports; port++)
+               if (rte_eth_dev_socket_id(port) > 0 &&
+                       rte_eth_dev_socket_id(port) != (int)rte_socket_id()) {
+                       printf("\n\n");
+                       printf("WARNING: port %u is on remote NUMA node\n",
+                              port);
+                       printf("to polling thread.\n");
+                       printf("Performance will not be optimal.\n");
+
+                       printf("\nCore %u forwarding packets. ",
+                              rte_lcore_id());
+                       printf("[Ctrl+C to quit]\n");
+               }
+       /* Run until the application is quit or killed. */
+       for (;;) {
+               /*
+                * Receive packets on a port, classify them and forward them
+                * on the paired port.
+                * The mapping is 0 -> 1, 1 -> 0, 2 -> 3, 3 -> 2, etc.
+                */
+               for (port = 0; port < nb_ports; port++) {
+                       /* Get burst of RX packets, from first port of pair. */
+                       struct rte_mbuf *bufs[BURST_SIZE];
+                       const uint16_t nb_rx = rte_eth_rx_burst(port, 0,
+                                       bufs, BURST_SIZE);
+
+                       if (unlikely(nb_rx == 0))
+                               continue;
+
+                       for (i = 0; i < MAX_NUM_CLASSIFY; i++) {
+                               if (rules[i]) {
+                                       ret = rte_flow_classifier_query(
+                                               cls_app->cls,
+                                               cls_app->table_id[0],
+                                               bufs, nb_rx, rules[i],
+                                               &classify_stats);
+                                       if (ret)
+                                               printf(
+                                                       "rule [%d] query failed ret [%d]\n\n",
+                                                       i, ret);
+                                       else {
+                                               printf(
+                                               "rule[%d] count=%"PRIu64"\n",
+                                               i, ntuple_stats.counter1);
+
+                                               printf("proto = %d\n",
+                                               ntuple_stats.ipv4_5tuple.proto);
+                                       }
+                               }
+                       }
+
+                       /* Send burst of TX packets, to second port of pair. */
+                       const uint16_t nb_tx = rte_eth_tx_burst(port ^ 1, 0,
+                                       bufs, nb_rx);
+
+                       /* Free any unsent packets. */
+                       if (unlikely(nb_tx < nb_rx)) {
+                               uint16_t buf;
+
+                               for (buf = nb_tx; buf < nb_rx; buf++)
+                                       rte_pktmbuf_free(bufs[buf]);
+                       }
+               }
+       }
+}
+
+/*
+ * Parse IPv4 5 tuple rules file, ipv4_rules_file.txt.
+ * Expected format:
+ * <src_ipv4_addr>'/'<masklen> <space> \
+ * <dst_ipv4_addr>'/'<masklen> <space> \
+ * <src_port> <space> ":" <src_port_mask> <space> \
+ * <dst_port> <space> ":" <dst_port_mask> <space> \
+ * <proto>'/'<proto_mask> <space> \
+ * <priority>
+ */
+
+static int
+get_cb_field(char **in, uint32_t *fd, int base, unsigned long lim,
+               char dlm)
+{
+       unsigned long val;
+       char *end;
+
+       errno = 0;
+       val = strtoul(*in, &end, base);
+       if (errno != 0 || end[0] != dlm || val > lim)
+               return -EINVAL;
+       *fd = (uint32_t)val;
+       *in = end + 1;
+       return 0;
+}
+
+static int
+parse_ipv4_net(char *in, uint32_t *addr, uint32_t *mask_len)
+{
+       uint32_t a, b, c, d, m;
+
+       if (get_cb_field(&in, &a, 0, UINT8_MAX, '.'))
+               return -EINVAL;
+       if (get_cb_field(&in, &b, 0, UINT8_MAX, '.'))
+               return -EINVAL;
+       if (get_cb_field(&in, &c, 0, UINT8_MAX, '.'))
+               return -EINVAL;
+       if (get_cb_field(&in, &d, 0, UINT8_MAX, '/'))
+               return -EINVAL;
+       if (get_cb_field(&in, &m, 0, sizeof(uint32_t) * CHAR_BIT, 0))
+               return -EINVAL;
+
+       addr[0] = IPv4(a, b, c, d);
+       mask_len[0] = m;
+       return 0;
+}
+
+static int
+parse_ipv4_5tuple_rule(char *str, struct rte_eth_ntuple_filter *ntuple_filter)
+{
+       int i, ret;
+       char *s, *sp, *in[CB_FLD_NUM];
+       static const char *dlm = " \t\n";
+       int dim = CB_FLD_NUM;
+       uint32_t temp;
+
+       s = str;
+       for (i = 0; i != dim; i++, s = NULL) {
+               in[i] = strtok_r(s, dlm, &sp);
+               if (in[i] == NULL)
+                       return -EINVAL;
+       }
+
+       ret = parse_ipv4_net(in[CB_FLD_SRC_ADDR],
+                       &ntuple_filter->src_ip,
+                       &ntuple_filter->src_ip_mask);
+       if (ret != 0) {
+               flow_classify_log("failed to read source address/mask: %s\n",
+                       in[CB_FLD_SRC_ADDR]);
+               return ret;
+       }
+
+       ret = parse_ipv4_net(in[CB_FLD_DST_ADDR],
+                       &ntuple_filter->dst_ip,
+                       &ntuple_filter->dst_ip_mask);
+       if (ret != 0) {
+               flow_classify_log("failed to read source address/mask: %s\n",
+                       in[CB_FLD_DST_ADDR]);
+               return ret;
+       }
+
+       if (get_cb_field(&in[CB_FLD_SRC_PORT], &temp, 0, UINT16_MAX, 0))
+               return -EINVAL;
+       ntuple_filter->src_port = (uint16_t)temp;
+
+       if (strncmp(in[CB_FLD_SRC_PORT_DLM], cb_port_delim,
+                       sizeof(cb_port_delim)) != 0)
+               return -EINVAL;
+
+       if (get_cb_field(&in[CB_FLD_SRC_PORT_MASK], &temp, 0, UINT16_MAX, 0))
+               return -EINVAL;
+       ntuple_filter->src_port_mask = (uint16_t)temp;
+
+       if (get_cb_field(&in[CB_FLD_DST_PORT], &temp, 0, UINT16_MAX, 0))
+               return -EINVAL;
+       ntuple_filter->dst_port = (uint16_t)temp;
+
+       if (strncmp(in[CB_FLD_DST_PORT_DLM], cb_port_delim,
+                       sizeof(cb_port_delim)) != 0)
+               return -EINVAL;
+
+       if (get_cb_field(&in[CB_FLD_DST_PORT_MASK], &temp, 0, UINT16_MAX, 0))
+               return -EINVAL;
+       ntuple_filter->dst_port_mask = (uint16_t)temp;
+
+       if (get_cb_field(&in[CB_FLD_PROTO], &temp, 0, UINT8_MAX, '/'))
+               return -EINVAL;
+       ntuple_filter->proto = (uint8_t)temp;
+
+       if (get_cb_field(&in[CB_FLD_PROTO], &temp, 0, UINT8_MAX, 0))
+               return -EINVAL;
+       ntuple_filter->proto_mask = (uint8_t)temp;
+
+       if (get_cb_field(&in[CB_FLD_PRIORITY], &temp, 0, UINT16_MAX, 0))
+               return -EINVAL;
+       ntuple_filter->priority = (uint16_t)temp;
+       if (ntuple_filter->priority > FLOW_CLASSIFY_MAX_PRIORITY)
+               ret = -EINVAL;
+
+       return ret;
+}
+
+/* Bypass comment and empty lines */
+static inline int
+is_bypass_line(char *buff)
+{
+       int i = 0;
+
+       /* comment line */
+       if (buff[0] == COMMENT_LEAD_CHAR)
+               return 1;
+       /* empty line */
+       while (buff[i] != '\0') {
+               if (!isspace(buff[i]))
+                       return 0;
+               i++;
+       }
+       return 1;
+}
+
+static uint32_t
+convert_depth_to_bitmask(uint32_t depth_val)
+{
+       uint32_t bitmask = 0;
+       int i, j;
+
+       for (i = depth_val, j = 0; i > 0; i--, j++)
+               bitmask |= (1 << (31 - j));
+       return bitmask;
+}
+
+static int
+add_classify_rule(struct rte_eth_ntuple_filter *ntuple_filter,
+               struct flow_classifier *cls_app)
+{
+       int ret = -1;
+       int key_found;
+       struct rte_flow_error error;
+       struct rte_flow_item_ipv4 ipv4_spec;
+       struct rte_flow_item_ipv4 ipv4_mask;
+       struct rte_flow_item ipv4_udp_item;
+       struct rte_flow_item ipv4_tcp_item;
+       struct rte_flow_item ipv4_sctp_item;
+       struct rte_flow_item_udp udp_spec;
+       struct rte_flow_item_udp udp_mask;
+       struct rte_flow_item udp_item;
+       struct rte_flow_item_tcp tcp_spec;
+       struct rte_flow_item_tcp tcp_mask;
+       struct rte_flow_item tcp_item;
+       struct rte_flow_item_sctp sctp_spec;
+       struct rte_flow_item_sctp sctp_mask;
+       struct rte_flow_item sctp_item;
+       struct rte_flow_item pattern_ipv4_5tuple[4];
+       struct rte_flow_classify_rule *rule;
+       uint8_t ipv4_proto;
+
+       if (num_classify_rules >= MAX_NUM_CLASSIFY) {
+               printf(
+                       "\nINFO:  classify rule capacity %d reached\n",
+                       num_classify_rules);
+               return ret;
+       }
+
+       /* set up parameters for validate and add */
+       memset(&ipv4_spec, 0, sizeof(ipv4_spec));
+       ipv4_spec.hdr.next_proto_id = ntuple_filter->proto;
+       ipv4_spec.hdr.src_addr = ntuple_filter->src_ip;
+       ipv4_spec.hdr.dst_addr = ntuple_filter->dst_ip;
+       ipv4_proto = ipv4_spec.hdr.next_proto_id;
+
+       memset(&ipv4_mask, 0, sizeof(ipv4_mask));
+       ipv4_mask.hdr.next_proto_id = ntuple_filter->proto_mask;
+       ipv4_mask.hdr.src_addr = ntuple_filter->src_ip_mask;
+       ipv4_mask.hdr.src_addr =
+               convert_depth_to_bitmask(ipv4_mask.hdr.src_addr);
+       ipv4_mask.hdr.dst_addr = ntuple_filter->dst_ip_mask;
+       ipv4_mask.hdr.dst_addr =
+               convert_depth_to_bitmask(ipv4_mask.hdr.dst_addr);
+
+       switch (ipv4_proto) {
+       case IPPROTO_UDP:
+               ipv4_udp_item.type = RTE_FLOW_ITEM_TYPE_IPV4;
+               ipv4_udp_item.spec = &ipv4_spec;
+               ipv4_udp_item.mask = &ipv4_mask;
+               ipv4_udp_item.last = NULL;
+
+               udp_spec.hdr.src_port = ntuple_filter->src_port;
+               udp_spec.hdr.dst_port = ntuple_filter->dst_port;
+               udp_spec.hdr.dgram_len = 0;
+               udp_spec.hdr.dgram_cksum = 0;
+
+               udp_mask.hdr.src_port = ntuple_filter->src_port_mask;
+               udp_mask.hdr.dst_port = ntuple_filter->dst_port_mask;
+               udp_mask.hdr.dgram_len = 0;
+               udp_mask.hdr.dgram_cksum = 0;
+
+               udp_item.type = RTE_FLOW_ITEM_TYPE_UDP;
+               udp_item.spec = &udp_spec;
+               udp_item.mask = &udp_mask;
+               udp_item.last = NULL;
+
+               attr.priority = ntuple_filter->priority;
+               pattern_ipv4_5tuple[1] = ipv4_udp_item;
+               pattern_ipv4_5tuple[2] = udp_item;
+               break;
+       case IPPROTO_TCP:
+               ipv4_tcp_item.type = RTE_FLOW_ITEM_TYPE_IPV4;
+               ipv4_tcp_item.spec = &ipv4_spec;
+               ipv4_tcp_item.mask = &ipv4_mask;
+               ipv4_tcp_item.last = NULL;
+
+               memset(&tcp_spec, 0, sizeof(tcp_spec));
+               tcp_spec.hdr.src_port = ntuple_filter->src_port;
+               tcp_spec.hdr.dst_port = ntuple_filter->dst_port;
+
+               memset(&tcp_mask, 0, sizeof(tcp_mask));
+               tcp_mask.hdr.src_port = ntuple_filter->src_port_mask;
+               tcp_mask.hdr.dst_port = ntuple_filter->dst_port_mask;
+
+               tcp_item.type = RTE_FLOW_ITEM_TYPE_TCP;
+               tcp_item.spec = &tcp_spec;
+               tcp_item.mask = &tcp_mask;
+               tcp_item.last = NULL;
+
+               attr.priority = ntuple_filter->priority;
+               pattern_ipv4_5tuple[1] = ipv4_tcp_item;
+               pattern_ipv4_5tuple[2] = tcp_item;
+               break;
+       case IPPROTO_SCTP:
+               ipv4_sctp_item.type = RTE_FLOW_ITEM_TYPE_IPV4;
+               ipv4_sctp_item.spec = &ipv4_spec;
+               ipv4_sctp_item.mask = &ipv4_mask;
+               ipv4_sctp_item.last = NULL;
+
+               sctp_spec.hdr.src_port = ntuple_filter->src_port;
+               sctp_spec.hdr.dst_port = ntuple_filter->dst_port;
+               sctp_spec.hdr.cksum = 0;
+               sctp_spec.hdr.tag = 0;
+
+               sctp_mask.hdr.src_port = ntuple_filter->src_port_mask;
+               sctp_mask.hdr.dst_port = ntuple_filter->dst_port_mask;
+               sctp_mask.hdr.cksum = 0;
+               sctp_mask.hdr.tag = 0;
+
+               sctp_item.type = RTE_FLOW_ITEM_TYPE_SCTP;
+               sctp_item.spec = &sctp_spec;
+               sctp_item.mask = &sctp_mask;
+               sctp_item.last = NULL;
+
+               attr.priority = ntuple_filter->priority;
+               pattern_ipv4_5tuple[1] = ipv4_sctp_item;
+               pattern_ipv4_5tuple[2] = sctp_item;
+               break;
+       default:
+               return ret;
+       }
+
+       attr.ingress = 1;
+       pattern_ipv4_5tuple[0] = eth_item;
+       pattern_ipv4_5tuple[3] = end_item;
+       actions[0] = count_action;
+       actions[1] = end_action;
+
+       rule = rte_flow_classify_table_entry_add(
+                       cls_app->cls, cls_app->table_id[0], &key_found,
+                       &attr, pattern_ipv4_5tuple, actions, &error);
+       if (rule == NULL) {
+               printf("table entry add failed ipv4_proto = %u\n",
+                       ipv4_proto);
+               ret = -1;
+               return ret;
+       }
+
+       rules[num_classify_rules] = rule;
+       num_classify_rules++;
+       return 0;
+}
+
+static int
+add_rules(const char *rule_path, struct flow_classifier *cls_app)
+{
+       FILE *fh;
+       char buff[LINE_MAX];
+       unsigned int i = 0;
+       unsigned int total_num = 0;
+       struct rte_eth_ntuple_filter ntuple_filter;
+
+       fh = fopen(rule_path, "rb");
+       if (fh == NULL)
+               rte_exit(EXIT_FAILURE, "%s: Open %s failed\n", __func__,
+                       rule_path);
+
+       fseek(fh, 0, SEEK_SET);
+
+       i = 0;
+       while (fgets(buff, LINE_MAX, fh) != NULL) {
+               i++;
+
+               if (is_bypass_line(buff))
+                       continue;
+
+               if (total_num >= FLOW_CLASSIFY_MAX_RULE_NUM - 1) {
+                       printf("\nINFO: classify rule capacity %d reached\n",
+                               total_num);
+                       break;
+               }
+
+               if (parse_ipv4_5tuple_rule(buff, &ntuple_filter) != 0)
+                       rte_exit(EXIT_FAILURE,
+                               "%s Line %u: parse rules error\n",
+                               rule_path, i);
+
+               if (add_classify_rule(&ntuple_filter, cls_app) != 0)
+                       rte_exit(EXIT_FAILURE, "add rule error\n");
+
+               total_num++;
+       }
+
+       fclose(fh);
+       return 0;
+}
+
+/* display usage */
+static void
+print_usage(const char *prgname)
+{
+       printf("%s usage:\n", prgname);
+       printf("[EAL options] --  --"OPTION_RULE_IPV4"=FILE: ");
+       printf("specify the ipv4 rules file.\n");
+       printf("Each rule occupies one line in the file.\n");
+}
+
+/* Parse the argument given in the command line of the application */
+static int
+parse_args(int argc, char **argv)
+{
+       int opt, ret;
+       char **argvopt;
+       int option_index;
+       char *prgname = argv[0];
+       static struct option lgopts[] = {
+               {OPTION_RULE_IPV4, 1, 0, 0},
+               {NULL, 0, 0, 0}
+       };
+
+       argvopt = argv;
+
+       while ((opt = getopt_long(argc, argvopt, "",
+                               lgopts, &option_index)) != EOF) {
+
+               switch (opt) {
+               /* long options */
+               case 0:
+                       if (!strncmp(lgopts[option_index].name,
+                                       OPTION_RULE_IPV4,
+                                       sizeof(OPTION_RULE_IPV4)))
+                               parm_config.rule_ipv4_name = optarg;
+                       break;
+               default:
+                       print_usage(prgname);
+                       return -1;
+               }
+       }
+
+       if (optind >= 0)
+               argv[optind-1] = prgname;
+
+       ret = optind-1;
+       optind = 1; /* reset getopt lib */
+       return ret;
+}
+
+/*
+ * The main function, which does initialization and calls the lcore_main
+ * function.
+ */
+int
+main(int argc, char *argv[])
+{
+       struct rte_mempool *mbuf_pool;
+       uint8_t nb_ports;
+       uint8_t portid;
+       int ret;
+       int socket_id;
+       struct rte_table_acl_params table_acl_params;
+       struct rte_flow_classify_table_params cls_table_params;
+       struct flow_classifier *cls_app;
+       struct rte_flow_classifier_params cls_params;
+       uint32_t size;
+
+       /* Initialize the Environment Abstraction Layer (EAL). */
+       ret = rte_eal_init(argc, argv);
+       if (ret < 0)
+               rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
+
+       argc -= ret;
+       argv += ret;
+
+       /* parse application arguments (after the EAL ones) */
+       ret = parse_args(argc, argv);
+       if (ret < 0)
+               rte_exit(EXIT_FAILURE, "Invalid flow_classify parameters\n");
+
+       /* Check that there is an even number of ports to send/receive on. */
+       nb_ports = rte_eth_dev_count();
+       if (nb_ports < 2 || (nb_ports & 1))
+               rte_exit(EXIT_FAILURE, "Error: number of ports must be even\n");
+
+       /* Creates a new mempool in memory to hold the mbufs. */
+       mbuf_pool = rte_pktmbuf_pool_create("MBUF_POOL", NUM_MBUFS * nb_ports,
+               MBUF_CACHE_SIZE, 0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
+
+       if (mbuf_pool == NULL)
+               rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
+
+       /* Initialize all ports. */
+       for (portid = 0; portid < nb_ports; portid++)
+               if (port_init(portid, mbuf_pool) != 0)
+                       rte_exit(EXIT_FAILURE, "Cannot init port %"PRIu8 "\n",
+                                       portid);
+
+       if (rte_lcore_count() > 1)
+               printf("\nWARNING: Too many lcores enabled. Only 1 used.\n");
+
+       socket_id = rte_eth_dev_socket_id(0);
+
+       /* Memory allocation */
+       size = RTE_CACHE_LINE_ROUNDUP(sizeof(struct flow_classifier_acl));
+       cls_app = rte_zmalloc(NULL, size, RTE_CACHE_LINE_SIZE);
+       if (cls_app == NULL)
+               rte_exit(EXIT_FAILURE, "Cannot allocate classifier memory\n");
+
+       cls_params.name = "flow_classifier";
+       cls_params.socket_id = socket_id;
+       cls_params.type = RTE_FLOW_CLASSIFY_TABLE_TYPE_ACL;
+
+       cls_app->cls = rte_flow_classifier_create(&cls_params);
+       if (cls_app->cls == NULL) {
+               rte_free(cls_app);
+               rte_exit(EXIT_FAILURE, "Cannot create classifier\n");
+       }
+
+       /* initialise ACL table params */
+       table_acl_params.name = "table_acl_ipv4_5tuple";
+       table_acl_params.n_rules = FLOW_CLASSIFY_MAX_RULE_NUM;
+       table_acl_params.n_rule_fields = RTE_DIM(ipv4_defs);
+       memcpy(table_acl_params.field_format, ipv4_defs, sizeof(ipv4_defs));
+
+       /* initialise table create params */
+       cls_table_params.ops = &rte_table_acl_ops,
+       cls_table_params.arg_create = &table_acl_params,
+
+       ret = rte_flow_classify_table_create(cls_app->cls, &cls_table_params,
+                       &cls_app->table_id[0]);
+       if (ret) {
+               rte_flow_classifier_free(cls_app->cls);
+               rte_free(cls_app);
+               rte_exit(EXIT_FAILURE, "Failed to create classifier table\n");
+       }
+
+       /* read file of IPv4 5 tuple rules and initialize parameters
+        * for rte_flow_classify_validate and rte_flow_classify_table_entry_add
+        * API's.
+        */
+       if (add_rules(parm_config.rule_ipv4_name, cls_app)) {
+               rte_flow_classifier_free(cls_app->cls);
+               rte_free(cls_app);
+               rte_exit(EXIT_FAILURE, "Failed to add rules\n");
+       }
+
+       /* Call lcore_main on the master core only. */
+       lcore_main(cls_app);
+
+       return 0;
+}
diff --git a/examples/flow_classify/ipv4_rules_file.txt b/examples/flow_classify/ipv4_rules_file.txt
new file mode 100644 (file)
index 0000000..dfa0631
--- /dev/null
@@ -0,0 +1,14 @@
+#file format:
+#src_ip/masklen dst_ip/masklen src_port : mask dst_port : mask proto/mask priority
+#
+2.2.2.3/24 2.2.2.7/24 32 : 0xffff 33 : 0xffff 17/0xff 0
+9.9.9.3/24 9.9.9.7/24 32 : 0xffff 33 : 0xffff 17/0xff 1
+9.9.9.3/24 9.9.9.7/24 32 : 0xffff 33 : 0xffff 6/0xff 2
+9.9.8.3/24 9.9.8.7/24 32 : 0xffff 33 : 0xffff 6/0xff 3
+6.7.8.9/24 2.3.4.5/24 32 : 0x0000 33 : 0x0000 132/0xff 4
+6.7.8.9/32 192.168.0.36/32 10 : 0xffff 11 : 0xffff 6/0xfe 5
+6.7.8.9/24 192.168.0.36/24 10 : 0xffff 11 : 0xffff 6/0xfe 6
+6.7.8.9/16 192.168.0.36/16 10 : 0xffff 11 : 0xffff 6/0xfe 7
+6.7.8.9/8 192.168.0.36/8 10 : 0xffff 11 : 0xffff 6/0xfe 8
+#error rules
+#9.8.7.6/8 192.168.0.36/8 10 : 0xffff 11 : 0xffff 6/0xfe 9
\ No newline at end of file