kvargs: use SPDX tags
[dpdk.git] / lib / librte_gro / rte_gro.c
index fa6d7ce..6618f4d 100644 (file)
@@ -1,39 +1,15 @@
-/*-
- *   BSD LICENSE
- *
- *   Copyright(c) 2017 Intel Corporation. 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.
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2017 Intel Corporation
  */
 
 #include <rte_malloc.h>
 #include <rte_mbuf.h>
+#include <rte_cycles.h>
+#include <rte_ethdev.h>
 
 #include "rte_gro.h"
+#include "gro_tcp4.h"
+#include "gro_vxlan_tcp4.h"
 
 typedef void *(*gro_tbl_create_fn)(uint16_t socket_id,
                uint16_t max_flow_num,
@@ -41,15 +17,34 @@ typedef void *(*gro_tbl_create_fn)(uint16_t socket_id,
 typedef void (*gro_tbl_destroy_fn)(void *tbl);
 typedef uint32_t (*gro_tbl_pkt_count_fn)(void *tbl);
 
-static gro_tbl_create_fn tbl_create_fn[RTE_GRO_TYPE_MAX_NUM];
-static gro_tbl_destroy_fn tbl_destroy_fn[RTE_GRO_TYPE_MAX_NUM];
-static gro_tbl_pkt_count_fn tbl_pkt_count_fn[RTE_GRO_TYPE_MAX_NUM];
+static gro_tbl_create_fn tbl_create_fn[RTE_GRO_TYPE_MAX_NUM] = {
+               gro_tcp4_tbl_create, gro_vxlan_tcp4_tbl_create, NULL};
+static gro_tbl_destroy_fn tbl_destroy_fn[RTE_GRO_TYPE_MAX_NUM] = {
+                       gro_tcp4_tbl_destroy, gro_vxlan_tcp4_tbl_destroy,
+                       NULL};
+static gro_tbl_pkt_count_fn tbl_pkt_count_fn[RTE_GRO_TYPE_MAX_NUM] = {
+                       gro_tcp4_tbl_pkt_count, gro_vxlan_tcp4_tbl_pkt_count,
+                       NULL};
+
+#define IS_IPV4_TCP_PKT(ptype) (RTE_ETH_IS_IPV4_HDR(ptype) && \
+               ((ptype & RTE_PTYPE_L4_TCP) == RTE_PTYPE_L4_TCP))
+
+#define IS_IPV4_VXLAN_TCP4_PKT(ptype) (RTE_ETH_IS_IPV4_HDR(ptype) && \
+               ((ptype & RTE_PTYPE_L4_UDP) == RTE_PTYPE_L4_UDP) && \
+               ((ptype & RTE_PTYPE_TUNNEL_VXLAN) == \
+                RTE_PTYPE_TUNNEL_VXLAN) && \
+                ((ptype & RTE_PTYPE_INNER_L4_TCP) == \
+                 RTE_PTYPE_INNER_L4_TCP) && \
+                 (((ptype & RTE_PTYPE_INNER_L3_MASK) & \
+                   (RTE_PTYPE_INNER_L3_IPV4 | \
+                    RTE_PTYPE_INNER_L3_IPV4_EXT | \
+                    RTE_PTYPE_INNER_L3_IPV4_EXT_UNKNOWN)) != 0))
 
 /*
- * GRO context structure, which is used to merge packets. It keeps
- * many reassembly tables of desired GRO types. Applications need to
- * create GRO context objects before using rte_gro_reassemble to
- * perform GRO.
+ * GRO context structure. It keeps the table structures, which are
+ * used to merge packets, for different GRO types. Before using
+ * rte_gro_reassemble(), applications need to create the GRO context
+ * first.
  */
 struct gro_ctx {
        /* GRO types to perform */
@@ -75,7 +70,7 @@ rte_gro_ctx_create(const struct rte_gro_param *param)
                return NULL;
 
        for (i = 0; i < RTE_GRO_TYPE_MAX_NUM; i++) {
-               gro_type_flag = 1 << i;
+               gro_type_flag = 1ULL << i;
                if ((param->gro_types & gro_type_flag) == 0)
                        continue;
 
@@ -107,10 +102,8 @@ rte_gro_ctx_destroy(void *ctx)
        uint64_t gro_type_flag;
        uint8_t i;
 
-       if (gro_ctx == NULL)
-               return;
        for (i = 0; i < RTE_GRO_TYPE_MAX_NUM; i++) {
-               gro_type_flag = 1 << i;
+               gro_type_flag = 1ULL << i;
                if ((gro_ctx->gro_types & gro_type_flag) == 0)
                        continue;
                destroy_tbl_fn = tbl_destroy_fn[i];
@@ -121,29 +114,187 @@ rte_gro_ctx_destroy(void *ctx)
 }
 
 uint16_t
-rte_gro_reassemble_burst(struct rte_mbuf **pkts __rte_unused,
+rte_gro_reassemble_burst(struct rte_mbuf **pkts,
                uint16_t nb_pkts,
-               const struct rte_gro_param *param __rte_unused)
+               const struct rte_gro_param *param)
 {
-       return nb_pkts;
+       /* allocate a reassembly table for TCP/IPv4 GRO */
+       struct gro_tcp4_tbl tcp_tbl;
+       struct gro_tcp4_flow tcp_flows[RTE_GRO_MAX_BURST_ITEM_NUM];
+       struct gro_tcp4_item tcp_items[RTE_GRO_MAX_BURST_ITEM_NUM] = {{0} };
+
+       /* Allocate a reassembly table for VXLAN GRO */
+       struct gro_vxlan_tcp4_tbl vxlan_tbl;
+       struct gro_vxlan_tcp4_flow vxlan_flows[RTE_GRO_MAX_BURST_ITEM_NUM];
+       struct gro_vxlan_tcp4_item vxlan_items[RTE_GRO_MAX_BURST_ITEM_NUM] = {
+               {{0}, 0, 0} };
+
+       struct rte_mbuf *unprocess_pkts[nb_pkts];
+       uint32_t item_num;
+       int32_t ret;
+       uint16_t i, unprocess_num = 0, nb_after_gro = nb_pkts;
+       uint8_t do_tcp4_gro = 0, do_vxlan_gro = 0;
+
+       if (unlikely((param->gro_types & (RTE_GRO_IPV4_VXLAN_TCP_IPV4 |
+                                       RTE_GRO_TCP_IPV4)) == 0))
+               return nb_pkts;
+
+       /* Get the maximum number of packets */
+       item_num = RTE_MIN(nb_pkts, (param->max_flow_num *
+                               param->max_item_per_flow));
+       item_num = RTE_MIN(item_num, RTE_GRO_MAX_BURST_ITEM_NUM);
+
+       if (param->gro_types & RTE_GRO_IPV4_VXLAN_TCP_IPV4) {
+               for (i = 0; i < item_num; i++)
+                       vxlan_flows[i].start_index = INVALID_ARRAY_INDEX;
+
+               vxlan_tbl.flows = vxlan_flows;
+               vxlan_tbl.items = vxlan_items;
+               vxlan_tbl.flow_num = 0;
+               vxlan_tbl.item_num = 0;
+               vxlan_tbl.max_flow_num = item_num;
+               vxlan_tbl.max_item_num = item_num;
+               do_vxlan_gro = 1;
+       }
+
+       if (param->gro_types & RTE_GRO_TCP_IPV4) {
+               for (i = 0; i < item_num; i++)
+                       tcp_flows[i].start_index = INVALID_ARRAY_INDEX;
+
+               tcp_tbl.flows = tcp_flows;
+               tcp_tbl.items = tcp_items;
+               tcp_tbl.flow_num = 0;
+               tcp_tbl.item_num = 0;
+               tcp_tbl.max_flow_num = item_num;
+               tcp_tbl.max_item_num = item_num;
+               do_tcp4_gro = 1;
+       }
+
+       for (i = 0; i < nb_pkts; i++) {
+               /*
+                * The timestamp is ignored, since all packets
+                * will be flushed from the tables.
+                */
+               if (IS_IPV4_VXLAN_TCP4_PKT(pkts[i]->packet_type) &&
+                               do_vxlan_gro) {
+                       ret = gro_vxlan_tcp4_reassemble(pkts[i], &vxlan_tbl, 0);
+                       if (ret > 0)
+                               /* Merge successfully */
+                               nb_after_gro--;
+                       else if (ret < 0)
+                               unprocess_pkts[unprocess_num++] = pkts[i];
+               } else if (IS_IPV4_TCP_PKT(pkts[i]->packet_type) &&
+                               do_tcp4_gro) {
+                       ret = gro_tcp4_reassemble(pkts[i], &tcp_tbl, 0);
+                       if (ret > 0)
+                               /* merge successfully */
+                               nb_after_gro--;
+                       else if (ret < 0)
+                               unprocess_pkts[unprocess_num++] = pkts[i];
+               } else
+                       unprocess_pkts[unprocess_num++] = pkts[i];
+       }
+
+       if (nb_after_gro < nb_pkts) {
+               i = 0;
+               /* Flush all packets from the tables */
+               if (do_vxlan_gro) {
+                       i = gro_vxlan_tcp4_tbl_timeout_flush(&vxlan_tbl,
+                                       0, pkts, nb_pkts);
+               }
+               if (do_tcp4_gro) {
+                       i += gro_tcp4_tbl_timeout_flush(&tcp_tbl, 0,
+                                       &pkts[i], nb_pkts - i);
+               }
+               /* Copy unprocessed packets */
+               if (unprocess_num > 0) {
+                       memcpy(&pkts[i], unprocess_pkts,
+                                       sizeof(struct rte_mbuf *) *
+                                       unprocess_num);
+               }
+       }
+
+       return nb_after_gro;
 }
 
 uint16_t
-rte_gro_reassemble(struct rte_mbuf **pkts __rte_unused,
+rte_gro_reassemble(struct rte_mbuf **pkts,
                uint16_t nb_pkts,
-               void *ctx __rte_unused)
+               void *ctx)
 {
-       return nb_pkts;
+       struct rte_mbuf *unprocess_pkts[nb_pkts];
+       struct gro_ctx *gro_ctx = ctx;
+       void *tcp_tbl, *vxlan_tbl;
+       uint64_t current_time;
+       uint16_t i, unprocess_num = 0;
+       uint8_t do_tcp4_gro, do_vxlan_gro;
+
+       if (unlikely((gro_ctx->gro_types & (RTE_GRO_IPV4_VXLAN_TCP_IPV4 |
+                                       RTE_GRO_TCP_IPV4)) == 0))
+               return nb_pkts;
+
+       tcp_tbl = gro_ctx->tbls[RTE_GRO_TCP_IPV4_INDEX];
+       vxlan_tbl = gro_ctx->tbls[RTE_GRO_IPV4_VXLAN_TCP_IPV4_INDEX];
+
+       do_tcp4_gro = (gro_ctx->gro_types & RTE_GRO_TCP_IPV4) ==
+               RTE_GRO_TCP_IPV4;
+       do_vxlan_gro = (gro_ctx->gro_types & RTE_GRO_IPV4_VXLAN_TCP_IPV4) ==
+               RTE_GRO_IPV4_VXLAN_TCP_IPV4;
+
+       current_time = rte_rdtsc();
+
+       for (i = 0; i < nb_pkts; i++) {
+               if (IS_IPV4_VXLAN_TCP4_PKT(pkts[i]->packet_type) &&
+                               do_vxlan_gro) {
+                       if (gro_vxlan_tcp4_reassemble(pkts[i], vxlan_tbl,
+                                               current_time) < 0)
+                               unprocess_pkts[unprocess_num++] = pkts[i];
+               } else if (IS_IPV4_TCP_PKT(pkts[i]->packet_type) &&
+                               do_tcp4_gro) {
+                       if (gro_tcp4_reassemble(pkts[i], tcp_tbl,
+                                               current_time) < 0)
+                               unprocess_pkts[unprocess_num++] = pkts[i];
+               } else
+                       unprocess_pkts[unprocess_num++] = pkts[i];
+       }
+       if (unprocess_num > 0) {
+               memcpy(pkts, unprocess_pkts, sizeof(struct rte_mbuf *) *
+                               unprocess_num);
+       }
+
+       return unprocess_num;
 }
 
 uint16_t
-rte_gro_timeout_flush(void *ctx __rte_unused,
-               uint64_t timeout_cycles __rte_unused,
-               uint64_t gro_types __rte_unused,
-               struct rte_mbuf **out __rte_unused,
-               uint16_t max_nb_out __rte_unused)
+rte_gro_timeout_flush(void *ctx,
+               uint64_t timeout_cycles,
+               uint64_t gro_types,
+               struct rte_mbuf **out,
+               uint16_t max_nb_out)
 {
-       return 0;
+       struct gro_ctx *gro_ctx = ctx;
+       uint64_t flush_timestamp;
+       uint16_t num = 0;
+
+       gro_types = gro_types & gro_ctx->gro_types;
+       flush_timestamp = rte_rdtsc() - timeout_cycles;
+
+       if (gro_types & RTE_GRO_IPV4_VXLAN_TCP_IPV4) {
+               num = gro_vxlan_tcp4_tbl_timeout_flush(gro_ctx->tbls[
+                               RTE_GRO_IPV4_VXLAN_TCP_IPV4_INDEX],
+                               flush_timestamp, out, max_nb_out);
+               max_nb_out -= num;
+       }
+
+       /* If no available space in 'out', stop flushing. */
+       if ((gro_types & RTE_GRO_TCP_IPV4) && max_nb_out > 0) {
+               num += gro_tcp4_tbl_timeout_flush(
+                               gro_ctx->tbls[RTE_GRO_TCP_IPV4_INDEX],
+                               flush_timestamp,
+                               &out[num], max_nb_out);
+       }
+
+       return num;
 }
 
 uint64_t
@@ -151,19 +302,20 @@ rte_gro_get_pkt_count(void *ctx)
 {
        struct gro_ctx *gro_ctx = ctx;
        gro_tbl_pkt_count_fn pkt_count_fn;
+       uint64_t gro_types = gro_ctx->gro_types, flag;
        uint64_t item_num = 0;
-       uint64_t gro_type_flag;
        uint8_t i;
 
-       for (i = 0; i < RTE_GRO_TYPE_MAX_NUM; i++) {
-               gro_type_flag = 1 << i;
-               if ((gro_ctx->gro_types & gro_type_flag) == 0)
+       for (i = 0; i < RTE_GRO_TYPE_MAX_NUM && gro_types; i++) {
+               flag = 1ULL << i;
+               if ((gro_types & flag) == 0)
                        continue;
 
+               gro_types ^= flag;
                pkt_count_fn = tbl_pkt_count_fn[i];
-               if (pkt_count_fn == NULL)
-                       continue;
-               item_num += pkt_count_fn(gro_ctx->tbls[i]);
+               if (pkt_count_fn)
+                       item_num += pkt_count_fn(gro_ctx->tbls[i]);
        }
+
        return item_num;
 }