app/testpmd: implement basic support for flow API
authorAdrien Mazarguil <adrien.mazarguil@6wind.com>
Wed, 21 Dec 2016 14:51:22 +0000 (15:51 +0100)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Fri, 23 Dec 2016 09:19:15 +0000 (10:19 +0100)
Add basic management functions for the generic flow API (validate, create,
destroy, flush, query and list). Flow rule objects and properties are
arranged in lists associated with each port.

Signed-off-by: Adrien Mazarguil <adrien.mazarguil@6wind.com>
Acked-by: Olga Shern <olgas@mellanox.com>
14 files changed:
app/test-pmd/cmdline.c
app/test-pmd/config.c
app/test-pmd/csumonly.c
app/test-pmd/flowgen.c
app/test-pmd/icmpecho.c
app/test-pmd/ieee1588fwd.c
app/test-pmd/iofwd.c
app/test-pmd/macfwd.c
app/test-pmd/macswap.c
app/test-pmd/parameters.c
app/test-pmd/rxonly.c
app/test-pmd/testpmd.c
app/test-pmd/testpmd.h
app/test-pmd/txonly.c

index d03a592..5d1c0dd 100644 (file)
@@ -75,6 +75,7 @@
 #include <rte_string_fns.h>
 #include <rte_devargs.h>
 #include <rte_eth_ctrl.h>
+#include <rte_flow.h>
 
 #include <cmdline_rdline.h>
 #include <cmdline_parse.h>
index 8cf537d..9716ce7 100644 (file)
@@ -92,6 +92,8 @@
 #include <rte_ethdev.h>
 #include <rte_string_fns.h>
 #include <rte_cycles.h>
+#include <rte_flow.h>
+#include <rte_errno.h>
 
 #include "testpmd.h"
 
@@ -751,6 +753,502 @@ port_mtu_set(portid_t port_id, uint16_t mtu)
        printf("Set MTU failed. diag=%d\n", diag);
 }
 
+/* Generic flow management functions. */
+
+/** Generate flow_item[] entry. */
+#define MK_FLOW_ITEM(t, s) \
+       [RTE_FLOW_ITEM_TYPE_ ## t] = { \
+               .name = # t, \
+               .size = s, \
+       }
+
+/** Information about known flow pattern items. */
+static const struct {
+       const char *name;
+       size_t size;
+} flow_item[] = {
+       MK_FLOW_ITEM(END, 0),
+       MK_FLOW_ITEM(VOID, 0),
+       MK_FLOW_ITEM(INVERT, 0),
+       MK_FLOW_ITEM(ANY, sizeof(struct rte_flow_item_any)),
+       MK_FLOW_ITEM(PF, 0),
+       MK_FLOW_ITEM(VF, sizeof(struct rte_flow_item_vf)),
+       MK_FLOW_ITEM(PORT, sizeof(struct rte_flow_item_port)),
+       MK_FLOW_ITEM(RAW, sizeof(struct rte_flow_item_raw)), /* +pattern[] */
+       MK_FLOW_ITEM(ETH, sizeof(struct rte_flow_item_eth)),
+       MK_FLOW_ITEM(VLAN, sizeof(struct rte_flow_item_vlan)),
+       MK_FLOW_ITEM(IPV4, sizeof(struct rte_flow_item_ipv4)),
+       MK_FLOW_ITEM(IPV6, sizeof(struct rte_flow_item_ipv6)),
+       MK_FLOW_ITEM(ICMP, sizeof(struct rte_flow_item_icmp)),
+       MK_FLOW_ITEM(UDP, sizeof(struct rte_flow_item_udp)),
+       MK_FLOW_ITEM(TCP, sizeof(struct rte_flow_item_tcp)),
+       MK_FLOW_ITEM(SCTP, sizeof(struct rte_flow_item_sctp)),
+       MK_FLOW_ITEM(VXLAN, sizeof(struct rte_flow_item_vxlan)),
+};
+
+/** Compute storage space needed by item specification. */
+static void
+flow_item_spec_size(const struct rte_flow_item *item,
+                   size_t *size, size_t *pad)
+{
+       if (!item->spec)
+               goto empty;
+       switch (item->type) {
+               union {
+                       const struct rte_flow_item_raw *raw;
+               } spec;
+
+       case RTE_FLOW_ITEM_TYPE_RAW:
+               spec.raw = item->spec;
+               *size = offsetof(struct rte_flow_item_raw, pattern) +
+                       spec.raw->length * sizeof(*spec.raw->pattern);
+               break;
+       default:
+empty:
+               *size = 0;
+               break;
+       }
+       *pad = RTE_ALIGN_CEIL(*size, sizeof(double)) - *size;
+}
+
+/** Generate flow_action[] entry. */
+#define MK_FLOW_ACTION(t, s) \
+       [RTE_FLOW_ACTION_TYPE_ ## t] = { \
+               .name = # t, \
+               .size = s, \
+       }
+
+/** Information about known flow actions. */
+static const struct {
+       const char *name;
+       size_t size;
+} flow_action[] = {
+       MK_FLOW_ACTION(END, 0),
+       MK_FLOW_ACTION(VOID, 0),
+       MK_FLOW_ACTION(PASSTHRU, 0),
+       MK_FLOW_ACTION(MARK, sizeof(struct rte_flow_action_mark)),
+       MK_FLOW_ACTION(FLAG, 0),
+       MK_FLOW_ACTION(QUEUE, sizeof(struct rte_flow_action_queue)),
+       MK_FLOW_ACTION(DROP, 0),
+       MK_FLOW_ACTION(COUNT, 0),
+       MK_FLOW_ACTION(DUP, sizeof(struct rte_flow_action_dup)),
+       MK_FLOW_ACTION(RSS, sizeof(struct rte_flow_action_rss)), /* +queue[] */
+       MK_FLOW_ACTION(PF, 0),
+       MK_FLOW_ACTION(VF, sizeof(struct rte_flow_action_vf)),
+};
+
+/** Compute storage space needed by action configuration. */
+static void
+flow_action_conf_size(const struct rte_flow_action *action,
+                     size_t *size, size_t *pad)
+{
+       if (!action->conf)
+               goto empty;
+       switch (action->type) {
+               union {
+                       const struct rte_flow_action_rss *rss;
+               } conf;
+
+       case RTE_FLOW_ACTION_TYPE_RSS:
+               conf.rss = action->conf;
+               *size = offsetof(struct rte_flow_action_rss, queue) +
+                       conf.rss->num * sizeof(*conf.rss->queue);
+               break;
+       default:
+empty:
+               *size = 0;
+               break;
+       }
+       *pad = RTE_ALIGN_CEIL(*size, sizeof(double)) - *size;
+}
+
+/** Generate a port_flow entry from attributes/pattern/actions. */
+static struct port_flow *
+port_flow_new(const struct rte_flow_attr *attr,
+             const struct rte_flow_item *pattern,
+             const struct rte_flow_action *actions)
+{
+       const struct rte_flow_item *item;
+       const struct rte_flow_action *action;
+       struct port_flow *pf = NULL;
+       size_t tmp;
+       size_t pad;
+       size_t off1 = 0;
+       size_t off2 = 0;
+       int err = ENOTSUP;
+
+store:
+       item = pattern;
+       if (pf)
+               pf->pattern = (void *)&pf->data[off1];
+       do {
+               struct rte_flow_item *dst = NULL;
+
+               if ((unsigned int)item->type > RTE_DIM(flow_item) ||
+                   !flow_item[item->type].name)
+                       goto notsup;
+               if (pf)
+                       dst = memcpy(pf->data + off1, item, sizeof(*item));
+               off1 += sizeof(*item);
+               flow_item_spec_size(item, &tmp, &pad);
+               if (item->spec) {
+                       if (pf)
+                               dst->spec = memcpy(pf->data + off2,
+                                                  item->spec, tmp);
+                       off2 += tmp + pad;
+               }
+               if (item->last) {
+                       if (pf)
+                               dst->last = memcpy(pf->data + off2,
+                                                  item->last, tmp);
+                       off2 += tmp + pad;
+               }
+               if (item->mask) {
+                       if (pf)
+                               dst->mask = memcpy(pf->data + off2,
+                                                  item->mask, tmp);
+                       off2 += tmp + pad;
+               }
+               off2 = RTE_ALIGN_CEIL(off2, sizeof(double));
+       } while ((item++)->type != RTE_FLOW_ITEM_TYPE_END);
+       off1 = RTE_ALIGN_CEIL(off1, sizeof(double));
+       action = actions;
+       if (pf)
+               pf->actions = (void *)&pf->data[off1];
+       do {
+               struct rte_flow_action *dst = NULL;
+
+               if ((unsigned int)action->type > RTE_DIM(flow_action) ||
+                   !flow_action[action->type].name)
+                       goto notsup;
+               if (pf)
+                       dst = memcpy(pf->data + off1, action, sizeof(*action));
+               off1 += sizeof(*action);
+               flow_action_conf_size(action, &tmp, &pad);
+               if (action->conf) {
+                       if (pf)
+                               dst->conf = memcpy(pf->data + off2,
+                                                  action->conf, tmp);
+                       off2 += tmp + pad;
+               }
+               off2 = RTE_ALIGN_CEIL(off2, sizeof(double));
+       } while ((action++)->type != RTE_FLOW_ACTION_TYPE_END);
+       if (pf != NULL)
+               return pf;
+       off1 = RTE_ALIGN_CEIL(off1, sizeof(double));
+       tmp = RTE_ALIGN_CEIL(offsetof(struct port_flow, data), sizeof(double));
+       pf = calloc(1, tmp + off1 + off2);
+       if (pf == NULL)
+               err = errno;
+       else {
+               *pf = (const struct port_flow){
+                       .size = tmp + off1 + off2,
+                       .attr = *attr,
+               };
+               tmp -= offsetof(struct port_flow, data);
+               off2 = tmp + off1;
+               off1 = tmp;
+               goto store;
+       }
+notsup:
+       rte_errno = err;
+       return NULL;
+}
+
+/** Print a message out of a flow error. */
+static int
+port_flow_complain(struct rte_flow_error *error)
+{
+       static const char *const errstrlist[] = {
+               [RTE_FLOW_ERROR_TYPE_NONE] = "no error",
+               [RTE_FLOW_ERROR_TYPE_UNSPECIFIED] = "cause unspecified",
+               [RTE_FLOW_ERROR_TYPE_HANDLE] = "flow rule (handle)",
+               [RTE_FLOW_ERROR_TYPE_ATTR_GROUP] = "group field",
+               [RTE_FLOW_ERROR_TYPE_ATTR_PRIORITY] = "priority field",
+               [RTE_FLOW_ERROR_TYPE_ATTR_INGRESS] = "ingress field",
+               [RTE_FLOW_ERROR_TYPE_ATTR_EGRESS] = "egress field",
+               [RTE_FLOW_ERROR_TYPE_ATTR] = "attributes structure",
+               [RTE_FLOW_ERROR_TYPE_ITEM_NUM] = "pattern length",
+               [RTE_FLOW_ERROR_TYPE_ITEM] = "specific pattern item",
+               [RTE_FLOW_ERROR_TYPE_ACTION_NUM] = "number of actions",
+               [RTE_FLOW_ERROR_TYPE_ACTION] = "specific action",
+       };
+       const char *errstr;
+       char buf[32];
+       int err = rte_errno;
+
+       if ((unsigned int)error->type > RTE_DIM(errstrlist) ||
+           !errstrlist[error->type])
+               errstr = "unknown type";
+       else
+               errstr = errstrlist[error->type];
+       printf("Caught error type %d (%s): %s%s\n",
+              error->type, errstr,
+              error->cause ? (snprintf(buf, sizeof(buf), "cause: %p, ",
+                                       error->cause), buf) : "",
+              error->message ? error->message : "(no stated reason)");
+       return -err;
+}
+
+/** Validate flow rule. */
+int
+port_flow_validate(portid_t port_id,
+                  const struct rte_flow_attr *attr,
+                  const struct rte_flow_item *pattern,
+                  const struct rte_flow_action *actions)
+{
+       struct rte_flow_error error;
+
+       /* Poisoning to make sure PMDs update it in case of error. */
+       memset(&error, 0x11, sizeof(error));
+       if (rte_flow_validate(port_id, attr, pattern, actions, &error))
+               return port_flow_complain(&error);
+       printf("Flow rule validated\n");
+       return 0;
+}
+
+/** Create flow rule. */
+int
+port_flow_create(portid_t port_id,
+                const struct rte_flow_attr *attr,
+                const struct rte_flow_item *pattern,
+                const struct rte_flow_action *actions)
+{
+       struct rte_flow *flow;
+       struct rte_port *port;
+       struct port_flow *pf;
+       uint32_t id;
+       struct rte_flow_error error;
+
+       /* Poisoning to make sure PMDs update it in case of error. */
+       memset(&error, 0x22, sizeof(error));
+       flow = rte_flow_create(port_id, attr, pattern, actions, &error);
+       if (!flow)
+               return port_flow_complain(&error);
+       port = &ports[port_id];
+       if (port->flow_list) {
+               if (port->flow_list->id == UINT32_MAX) {
+                       printf("Highest rule ID is already assigned, delete"
+                              " it first");
+                       rte_flow_destroy(port_id, flow, NULL);
+                       return -ENOMEM;
+               }
+               id = port->flow_list->id + 1;
+       } else
+               id = 0;
+       pf = port_flow_new(attr, pattern, actions);
+       if (!pf) {
+               int err = rte_errno;
+
+               printf("Cannot allocate flow: %s\n", rte_strerror(err));
+               rte_flow_destroy(port_id, flow, NULL);
+               return -err;
+       }
+       pf->next = port->flow_list;
+       pf->id = id;
+       pf->flow = flow;
+       port->flow_list = pf;
+       printf("Flow rule #%u created\n", pf->id);
+       return 0;
+}
+
+/** Destroy a number of flow rules. */
+int
+port_flow_destroy(portid_t port_id, uint32_t n, const uint32_t *rule)
+{
+       struct rte_port *port;
+       struct port_flow **tmp;
+       uint32_t c = 0;
+       int ret = 0;
+
+       if (port_id_is_invalid(port_id, ENABLED_WARN) ||
+           port_id == (portid_t)RTE_PORT_ALL)
+               return -EINVAL;
+       port = &ports[port_id];
+       tmp = &port->flow_list;
+       while (*tmp) {
+               uint32_t i;
+
+               for (i = 0; i != n; ++i) {
+                       struct rte_flow_error error;
+                       struct port_flow *pf = *tmp;
+
+                       if (rule[i] != pf->id)
+                               continue;
+                       /*
+                        * Poisoning to make sure PMDs update it in case
+                        * of error.
+                        */
+                       memset(&error, 0x33, sizeof(error));
+                       if (rte_flow_destroy(port_id, pf->flow, &error)) {
+                               ret = port_flow_complain(&error);
+                               continue;
+                       }
+                       printf("Flow rule #%u destroyed\n", pf->id);
+                       *tmp = pf->next;
+                       free(pf);
+                       break;
+               }
+               if (i == n)
+                       tmp = &(*tmp)->next;
+               ++c;
+       }
+       return ret;
+}
+
+/** Remove all flow rules. */
+int
+port_flow_flush(portid_t port_id)
+{
+       struct rte_flow_error error;
+       struct rte_port *port;
+       int ret = 0;
+
+       /* Poisoning to make sure PMDs update it in case of error. */
+       memset(&error, 0x44, sizeof(error));
+       if (rte_flow_flush(port_id, &error)) {
+               ret = port_flow_complain(&error);
+               if (port_id_is_invalid(port_id, DISABLED_WARN) ||
+                   port_id == (portid_t)RTE_PORT_ALL)
+                       return ret;
+       }
+       port = &ports[port_id];
+       while (port->flow_list) {
+               struct port_flow *pf = port->flow_list->next;
+
+               free(port->flow_list);
+               port->flow_list = pf;
+       }
+       return ret;
+}
+
+/** Query a flow rule. */
+int
+port_flow_query(portid_t port_id, uint32_t rule,
+               enum rte_flow_action_type action)
+{
+       struct rte_flow_error error;
+       struct rte_port *port;
+       struct port_flow *pf;
+       const char *name;
+       union {
+               struct rte_flow_query_count count;
+       } query;
+
+       if (port_id_is_invalid(port_id, ENABLED_WARN) ||
+           port_id == (portid_t)RTE_PORT_ALL)
+               return -EINVAL;
+       port = &ports[port_id];
+       for (pf = port->flow_list; pf; pf = pf->next)
+               if (pf->id == rule)
+                       break;
+       if (!pf) {
+               printf("Flow rule #%u not found\n", rule);
+               return -ENOENT;
+       }
+       if ((unsigned int)action > RTE_DIM(flow_action) ||
+           !flow_action[action].name)
+               name = "unknown";
+       else
+               name = flow_action[action].name;
+       switch (action) {
+       case RTE_FLOW_ACTION_TYPE_COUNT:
+               break;
+       default:
+               printf("Cannot query action type %d (%s)\n", action, name);
+               return -ENOTSUP;
+       }
+       /* Poisoning to make sure PMDs update it in case of error. */
+       memset(&error, 0x55, sizeof(error));
+       memset(&query, 0, sizeof(query));
+       if (rte_flow_query(port_id, pf->flow, action, &query, &error))
+               return port_flow_complain(&error);
+       switch (action) {
+       case RTE_FLOW_ACTION_TYPE_COUNT:
+               printf("%s:\n"
+                      " hits_set: %u\n"
+                      " bytes_set: %u\n"
+                      " hits: %" PRIu64 "\n"
+                      " bytes: %" PRIu64 "\n",
+                      name,
+                      query.count.hits_set,
+                      query.count.bytes_set,
+                      query.count.hits,
+                      query.count.bytes);
+               break;
+       default:
+               printf("Cannot display result for action type %d (%s)\n",
+                      action, name);
+               break;
+       }
+       return 0;
+}
+
+/** List flow rules. */
+void
+port_flow_list(portid_t port_id, uint32_t n, const uint32_t group[n])
+{
+       struct rte_port *port;
+       struct port_flow *pf;
+       struct port_flow *list = NULL;
+       uint32_t i;
+
+       if (port_id_is_invalid(port_id, ENABLED_WARN) ||
+           port_id == (portid_t)RTE_PORT_ALL)
+               return;
+       port = &ports[port_id];
+       if (!port->flow_list)
+               return;
+       /* Sort flows by group, priority and ID. */
+       for (pf = port->flow_list; pf != NULL; pf = pf->next) {
+               struct port_flow **tmp;
+
+               if (n) {
+                       /* Filter out unwanted groups. */
+                       for (i = 0; i != n; ++i)
+                               if (pf->attr.group == group[i])
+                                       break;
+                       if (i == n)
+                               continue;
+               }
+               tmp = &list;
+               while (*tmp &&
+                      (pf->attr.group > (*tmp)->attr.group ||
+                       (pf->attr.group == (*tmp)->attr.group &&
+                        pf->attr.priority > (*tmp)->attr.priority) ||
+                       (pf->attr.group == (*tmp)->attr.group &&
+                        pf->attr.priority == (*tmp)->attr.priority &&
+                        pf->id > (*tmp)->id)))
+                       tmp = &(*tmp)->tmp;
+               pf->tmp = *tmp;
+               *tmp = pf;
+       }
+       printf("ID\tGroup\tPrio\tAttr\tRule\n");
+       for (pf = list; pf != NULL; pf = pf->tmp) {
+               const struct rte_flow_item *item = pf->pattern;
+               const struct rte_flow_action *action = pf->actions;
+
+               printf("%" PRIu32 "\t%" PRIu32 "\t%" PRIu32 "\t%c%c\t",
+                      pf->id,
+                      pf->attr.group,
+                      pf->attr.priority,
+                      pf->attr.ingress ? 'i' : '-',
+                      pf->attr.egress ? 'e' : '-');
+               while (item->type != RTE_FLOW_ITEM_TYPE_END) {
+                       if (item->type != RTE_FLOW_ITEM_TYPE_VOID)
+                               printf("%s ", flow_item[item->type].name);
+                       ++item;
+               }
+               printf("=>");
+               while (action->type != RTE_FLOW_ACTION_TYPE_END) {
+                       if (action->type != RTE_FLOW_ACTION_TYPE_VOID)
+                               printf(" %s", flow_action[action->type].name);
+                       ++action;
+               }
+               printf("\n");
+       }
+}
+
 /*
  * RX/TX ring descriptors display functions.
  */
index 57e6ae2..dd67ebf 100644 (file)
@@ -70,6 +70,7 @@
 #include <rte_sctp.h>
 #include <rte_prefetch.h>
 #include <rte_string_fns.h>
+#include <rte_flow.h>
 #include "testpmd.h"
 
 #define IP_DEFTTL  64   /* from RFC 1340. */
index b13ff89..13b4f90 100644 (file)
@@ -68,6 +68,7 @@
 #include <rte_tcp.h>
 #include <rte_udp.h>
 #include <rte_string_fns.h>
+#include <rte_flow.h>
 
 #include "testpmd.h"
 
index 6a4e750..f25a8f5 100644 (file)
@@ -61,6 +61,7 @@
 #include <rte_ip.h>
 #include <rte_icmp.h>
 #include <rte_string_fns.h>
+#include <rte_flow.h>
 
 #include "testpmd.h"
 
index 0d3b37a..51170ee 100644 (file)
@@ -34,6 +34,7 @@
 
 #include <rte_cycles.h>
 #include <rte_ethdev.h>
+#include <rte_flow.h>
 
 #include "testpmd.h"
 
index 26936b7..15cb4a2 100644 (file)
@@ -64,6 +64,7 @@
 #include <rte_ether.h>
 #include <rte_ethdev.h>
 #include <rte_string_fns.h>
+#include <rte_flow.h>
 
 #include "testpmd.h"
 
index 86e01de..d361db1 100644 (file)
@@ -65,6 +65,7 @@
 #include <rte_ethdev.h>
 #include <rte_ip.h>
 #include <rte_string_fns.h>
+#include <rte_flow.h>
 
 #include "testpmd.h"
 
index 36e139f..f996039 100644 (file)
@@ -65,6 +65,7 @@
 #include <rte_ethdev.h>
 #include <rte_ip.h>
 #include <rte_string_fns.h>
+#include <rte_flow.h>
 
 #include "testpmd.h"
 
index 08e5a76..28db8cd 100644 (file)
@@ -76,6 +76,7 @@
 #ifdef RTE_LIBRTE_PMD_BOND
 #include <rte_eth_bond.h>
 #endif
+#include <rte_flow.h>
 
 #include "testpmd.h"
 
index fff815c..cf00576 100644 (file)
@@ -67,6 +67,7 @@
 #include <rte_ip.h>
 #include <rte_udp.h>
 #include <rte_net.h>
+#include <rte_flow.h>
 
 #include "testpmd.h"
 
index a0332c2..bfb2f8e 100644 (file)
@@ -78,6 +78,7 @@
 #ifdef RTE_LIBRTE_PDUMP
 #include <rte_pdump.h>
 #endif
+#include <rte_flow.h>
 
 #include "testpmd.h"
 
@@ -1545,6 +1546,8 @@ close_port(portid_t pid)
                        continue;
                }
 
+               if (port->flow_list)
+                       port_flow_flush(pi);
                rte_eth_dev_close(pi);
 
                if (rte_atomic16_cmpset(&(port->port_status),
@@ -1599,6 +1602,9 @@ detach_port(uint8_t port_id)
                return;
        }
 
+       if (ports[port_id].flow_list)
+               port_flow_flush(port_id);
+
        if (rte_eth_dev_detach(port_id, name))
                return;
 
index 9c1e703..22ce2d6 100644 (file)
@@ -144,6 +144,19 @@ struct fwd_stream {
 /** Insert double VLAN header in forward engine */
 #define TESTPMD_TX_OFFLOAD_INSERT_QINQ       0x0080
 
+/** Descriptor for a single flow. */
+struct port_flow {
+       size_t size; /**< Allocated space including data[]. */
+       struct port_flow *next; /**< Next flow in list. */
+       struct port_flow *tmp; /**< Temporary linking. */
+       uint32_t id; /**< Flow rule ID. */
+       struct rte_flow *flow; /**< Opaque flow object returned by PMD. */
+       struct rte_flow_attr attr; /**< Attributes. */
+       struct rte_flow_item *pattern; /**< Pattern. */
+       struct rte_flow_action *actions; /**< Actions. */
+       uint8_t data[]; /**< Storage for pattern/actions. */
+};
+
 /**
  * The data structure associated with each port.
  */
@@ -177,6 +190,7 @@ struct rte_port {
        struct ether_addr       *mc_addr_pool; /**< pool of multicast addrs */
        uint32_t                mc_addr_nb; /**< nb. of addr. in mc_addr_pool */
        uint8_t                 slave_flag; /**< bonding slave port */
+       struct port_flow        *flow_list; /**< Associated flows. */
 };
 
 extern portid_t __rte_unused
@@ -504,6 +518,19 @@ void port_reg_bit_field_set(portid_t port_id, uint32_t reg_off,
                            uint8_t bit1_pos, uint8_t bit2_pos, uint32_t value);
 void port_reg_display(portid_t port_id, uint32_t reg_off);
 void port_reg_set(portid_t port_id, uint32_t reg_off, uint32_t value);
+int port_flow_validate(portid_t port_id,
+                      const struct rte_flow_attr *attr,
+                      const struct rte_flow_item *pattern,
+                      const struct rte_flow_action *actions);
+int port_flow_create(portid_t port_id,
+                    const struct rte_flow_attr *attr,
+                    const struct rte_flow_item *pattern,
+                    const struct rte_flow_action *actions);
+int port_flow_destroy(portid_t port_id, uint32_t n, const uint32_t *rule);
+int port_flow_flush(portid_t port_id);
+int port_flow_query(portid_t port_id, uint32_t rule,
+                   enum rte_flow_action_type action);
+void port_flow_list(portid_t port_id, uint32_t n, const uint32_t *group);
 
 void rx_ring_desc_display(portid_t port_id, queueid_t rxq_id, uint16_t rxd_id);
 void tx_ring_desc_display(portid_t port_id, queueid_t txq_id, uint16_t txd_id);
index 8513a06..e996f35 100644 (file)
@@ -68,6 +68,7 @@
 #include <rte_tcp.h>
 #include <rte_udp.h>
 #include <rte_string_fns.h>
+#include <rte_flow.h>
 
 #include "testpmd.h"