node: add IPv4 lookup for x86
authorPavan Nikhilesh <pbhagavatula@marvell.com>
Sat, 11 Apr 2020 14:14:20 +0000 (19:44 +0530)
committerThomas Monjalon <thomas@monjalon.net>
Tue, 5 May 2020 21:41:05 +0000 (23:41 +0200)
Add IPv4 lookup process function for ip4_lookup
rte_node. This node performs LPM lookup using x86_64
vector supported RTE_LPM API on every packet received
and forwards it to a next node that is identified by
lookup result.

Signed-off-by: Pavan Nikhilesh <pbhagavatula@marvell.com>
Signed-off-by: Nithin Dabilpuram <ndabilpuram@marvell.com>
Signed-off-by: Kiran Kumar K <kirankumark@marvell.com>
lib/librte_node/ip4_lookup.c
lib/librte_node/ip4_lookup_sse.h [new file with mode: 0644]

index 8ec4f06..39c8d8d 100644 (file)
@@ -30,6 +30,8 @@ struct ip4_lookup_node_main {
 
 #if defined(RTE_MACHINE_CPUFLAG_NEON)
 #include "ip4_lookup_neon.h"
+#elif defined(RTE_ARCH_X86)
+#include "ip4_lookup_sse.h"
 #else
 
 static uint16_t
diff --git a/lib/librte_node/ip4_lookup_sse.h b/lib/librte_node/ip4_lookup_sse.h
new file mode 100644 (file)
index 0000000..a071cc5
--- /dev/null
@@ -0,0 +1,244 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(C) 2020 Marvell International Ltd.
+ */
+
+#ifndef __INCLUDE_IP4_LOOKUP_SSE_H__
+#define __INCLUDE_IP4_LOOKUP_SSE_H__
+
+/* X86 SSE */
+static uint16_t
+ip4_lookup_node_process(struct rte_graph *graph, struct rte_node *node,
+                       void **objs, uint16_t nb_objs)
+{
+       struct rte_mbuf *mbuf0, *mbuf1, *mbuf2, *mbuf3, **pkts;
+       rte_edge_t next0, next1, next2, next3, next_index;
+       struct rte_ipv4_hdr *ipv4_hdr;
+       uint32_t ip0, ip1, ip2, ip3;
+       void **to_next, **from;
+       uint16_t last_spec = 0;
+       uint16_t n_left_from;
+       struct rte_lpm *lpm;
+       uint16_t held = 0;
+       uint32_t drop_nh;
+       rte_xmm_t dst;
+       __m128i dip; /* SSE register */
+       int rc, i;
+
+       /* Speculative next */
+       next_index = RTE_NODE_IP4_LOOKUP_NEXT_REWRITE;
+       /* Drop node */
+       drop_nh = ((uint32_t)RTE_NODE_IP4_LOOKUP_NEXT_PKT_DROP) << 16;
+
+       /* Get socket specific LPM from ctx */
+       lpm = *((struct rte_lpm **)node->ctx);
+
+       pkts = (struct rte_mbuf **)objs;
+       from = objs;
+       n_left_from = nb_objs;
+
+       if (n_left_from >= 4) {
+               for (i = 0; i < 4; i++)
+                       rte_prefetch0(rte_pktmbuf_mtod_offset(pkts[i], void *,
+                                               sizeof(struct rte_ether_hdr)));
+       }
+
+       /* Get stream for the speculated next node */
+       to_next = rte_node_next_stream_get(graph, node, next_index, nb_objs);
+       while (n_left_from >= 4) {
+               /* Prefetch next-next mbufs */
+               if (likely(n_left_from > 11)) {
+                       rte_prefetch0(pkts[8]);
+                       rte_prefetch0(pkts[9]);
+                       rte_prefetch0(pkts[10]);
+                       rte_prefetch0(pkts[11]);
+               }
+
+               /* Prefetch next mbuf data */
+               if (likely(n_left_from > 7)) {
+                       rte_prefetch0(rte_pktmbuf_mtod_offset(pkts[4], void *,
+                                               sizeof(struct rte_ether_hdr)));
+                       rte_prefetch0(rte_pktmbuf_mtod_offset(pkts[5], void *,
+                                               sizeof(struct rte_ether_hdr)));
+                       rte_prefetch0(rte_pktmbuf_mtod_offset(pkts[6], void *,
+                                               sizeof(struct rte_ether_hdr)));
+                       rte_prefetch0(rte_pktmbuf_mtod_offset(pkts[7], void *,
+                                               sizeof(struct rte_ether_hdr)));
+               }
+
+               mbuf0 = pkts[0];
+               mbuf1 = pkts[1];
+               mbuf2 = pkts[2];
+               mbuf3 = pkts[3];
+
+               pkts += 4;
+               n_left_from -= 4;
+
+               /* Extract DIP of mbuf0 */
+               ipv4_hdr = rte_pktmbuf_mtod_offset(mbuf0, struct rte_ipv4_hdr *,
+                                               sizeof(struct rte_ether_hdr));
+               ip0 = ipv4_hdr->dst_addr;
+               /* Extract cksum, ttl as ipv4 hdr is in cache */
+               node_mbuf_priv1(mbuf0)->cksum = ipv4_hdr->hdr_checksum;
+               node_mbuf_priv1(mbuf0)->ttl = ipv4_hdr->time_to_live;
+
+               /* Extract DIP of mbuf1 */
+               ipv4_hdr = rte_pktmbuf_mtod_offset(mbuf1, struct rte_ipv4_hdr *,
+                                               sizeof(struct rte_ether_hdr));
+               ip1 = ipv4_hdr->dst_addr;
+               /* Extract cksum, ttl as ipv4 hdr is in cache */
+               node_mbuf_priv1(mbuf1)->cksum = ipv4_hdr->hdr_checksum;
+               node_mbuf_priv1(mbuf1)->ttl = ipv4_hdr->time_to_live;
+
+               /* Extract DIP of mbuf2 */
+               ipv4_hdr = rte_pktmbuf_mtod_offset(mbuf2, struct rte_ipv4_hdr *,
+                                               sizeof(struct rte_ether_hdr));
+               ip2 = ipv4_hdr->dst_addr;
+               /* Extract cksum, ttl as ipv4 hdr is in cache */
+               node_mbuf_priv1(mbuf2)->cksum = ipv4_hdr->hdr_checksum;
+               node_mbuf_priv1(mbuf2)->ttl = ipv4_hdr->time_to_live;
+
+               /* Extract DIP of mbuf3 */
+               ipv4_hdr = rte_pktmbuf_mtod_offset(mbuf3, struct rte_ipv4_hdr *,
+                                               sizeof(struct rte_ether_hdr));
+               ip3 = ipv4_hdr->dst_addr;
+
+               /* Prepare for lookup x4 */
+               dip = _mm_set_epi32(ip3, ip2, ip1, ip0);
+
+               /* Byte swap 4 IPV4 addresses. */
+               const __m128i bswap_mask = _mm_set_epi8(
+                       12, 13, 14, 15, 8, 9, 10, 11, 4, 5, 6, 7, 0, 1, 2, 3);
+               dip = _mm_shuffle_epi8(dip, bswap_mask);
+
+               /* Extract cksum, ttl as ipv4 hdr is in cache */
+               node_mbuf_priv1(mbuf3)->cksum = ipv4_hdr->hdr_checksum;
+               node_mbuf_priv1(mbuf3)->ttl = ipv4_hdr->time_to_live;
+
+               /* Perform LPM lookup to get NH and next node */
+               rte_lpm_lookupx4(lpm, dip, dst.u32, drop_nh);
+
+               /* Extract next node id and NH */
+               node_mbuf_priv1(mbuf0)->nh = dst.u32[0] & 0xFFFF;
+               next0 = (dst.u32[0] >> 16);
+
+               node_mbuf_priv1(mbuf1)->nh = dst.u32[1] & 0xFFFF;
+               next1 = (dst.u32[1] >> 16);
+
+               node_mbuf_priv1(mbuf2)->nh = dst.u32[2] & 0xFFFF;
+               next2 = (dst.u32[2] >> 16);
+
+               node_mbuf_priv1(mbuf3)->nh = dst.u32[3] & 0xFFFF;
+               next3 = (dst.u32[3] >> 16);
+
+               /* Enqueue four to next node */
+               rte_edge_t fix_spec =
+                       (next_index ^ next0) | (next_index ^ next1) |
+                       (next_index ^ next2) | (next_index ^ next3);
+
+               if (unlikely(fix_spec)) {
+                       /* Copy things successfully speculated till now */
+                       rte_memcpy(to_next, from, last_spec * sizeof(from[0]));
+                       from += last_spec;
+                       to_next += last_spec;
+                       held += last_spec;
+                       last_spec = 0;
+
+                       /* Next0 */
+                       if (next_index == next0) {
+                               to_next[0] = from[0];
+                               to_next++;
+                               held++;
+                       } else {
+                               rte_node_enqueue_x1(graph, node, next0,
+                                                   from[0]);
+                       }
+
+                       /* Next1 */
+                       if (next_index == next1) {
+                               to_next[0] = from[1];
+                               to_next++;
+                               held++;
+                       } else {
+                               rte_node_enqueue_x1(graph, node, next1,
+                                                   from[1]);
+                       }
+
+                       /* Next2 */
+                       if (next_index == next2) {
+                               to_next[0] = from[2];
+                               to_next++;
+                               held++;
+                       } else {
+                               rte_node_enqueue_x1(graph, node, next2,
+                                                   from[2]);
+                       }
+
+                       /* Next3 */
+                       if (next_index == next3) {
+                               to_next[0] = from[3];
+                               to_next++;
+                               held++;
+                       } else {
+                               rte_node_enqueue_x1(graph, node, next3,
+                                                   from[3]);
+                       }
+
+                       from += 4;
+
+               } else {
+                       last_spec += 4;
+               }
+       }
+
+       while (n_left_from > 0) {
+               uint32_t next_hop;
+
+               mbuf0 = pkts[0];
+
+               pkts += 1;
+               n_left_from -= 1;
+
+               /* Extract DIP of mbuf0 */
+               ipv4_hdr = rte_pktmbuf_mtod_offset(mbuf0, struct rte_ipv4_hdr *,
+                                               sizeof(struct rte_ether_hdr));
+               /* Extract cksum, ttl as ipv4 hdr is in cache */
+               node_mbuf_priv1(mbuf0)->cksum = ipv4_hdr->hdr_checksum;
+               node_mbuf_priv1(mbuf0)->ttl = ipv4_hdr->time_to_live;
+
+               rc = rte_lpm_lookup(lpm, rte_be_to_cpu_32(ipv4_hdr->dst_addr),
+                                   &next_hop);
+               next_hop = (rc == 0) ? next_hop : drop_nh;
+
+               node_mbuf_priv1(mbuf0)->nh = next_hop & 0xFFFF;
+               next0 = (next_hop >> 16);
+
+               if (unlikely(next_index ^ next0)) {
+                       /* Copy things successfully speculated till now */
+                       rte_memcpy(to_next, from, last_spec * sizeof(from[0]));
+                       from += last_spec;
+                       to_next += last_spec;
+                       held += last_spec;
+                       last_spec = 0;
+
+                       rte_node_enqueue_x1(graph, node, next0, from[0]);
+                       from += 1;
+               } else {
+                       last_spec += 1;
+               }
+       }
+
+       /* !!! Home run !!! */
+       if (likely(last_spec == nb_objs)) {
+               rte_node_next_stream_move(graph, node, next_index);
+               return nb_objs;
+       }
+
+       held += last_spec;
+       /* Copy things successfully speculated till now */
+       rte_memcpy(to_next, from, last_spec * sizeof(from[0]));
+       rte_node_next_stream_put(graph, node, next_index, held);
+
+       return nb_objs;
+}
+
+#endif /* __INCLUDE_IP4_LOOKUP_SSE_H__ */