1807 report: updated dpdk performance overview section. - csit

diff options

author	Maciek Konstantynowicz <mkonstan@cisco.com>	2018-08-14 19:38:08 +0100
committer	Maciek Konstantynowicz <mkonstan@cisco.com>	2018-08-15 09:29:30 +0000
commit	14f5a52b235fc196952c85e127a57821e325d7bc (patch)
tree	3ae759e0ff2a76dba3dc525ff9d35641e9dcbba6 /tests/vpp/perf/ip4/10ge2p1x520-ethip4-ip4base-ndrpdr.robot
parent	c2ce046f71f7de1f247be68513e3d4e10c8c2a04 (diff)

1807 report: updated dpdk performance overview section.

Change-Id: I48e6e5b5750bcb4b3b7b6eb82c351bed1de08c73 Signed-off-by: Maciek Konstantynowicz <mkonstan@cisco.com>

Diffstat (limited to 'tests/vpp/perf/ip4/10ge2p1x520-ethip4-ip4base-ndrpdr.robot')

0 files changed, 0 insertions, 0 deletions

/* * Copyright (c) 2018 Cisco and/or its affiliates. * Licensed under the Apache License, Version 2.0 (the "License"); * you may not use this file except in compliance with the License. * You may obtain a copy of the License at: * * http://www.apache.org/licenses/LICENSE-2.0 * * Unless required by applicable law or agreed to in writing, software * distributed under the License is distributed on an "AS IS" BASIS, * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. * See the License for the specific language governing permissions and * limitations under the License. */ #include <vppinfra/error.h> #include <vppinfra/hash.h> #include <vnet/vnet.h> #include <vnet/ip/ip.h> #include <vnet/ethernet/ethernet.h> #include <vnet/interface_output.h> #include <vnet/vxlan-gbp/vxlan_gbp.h> #include <vnet/qos/qos_types.h> #include <vnet/adj/rewrite.h> /* Statistics (not all errors) */ #define foreach_vxlan_gbp_encap_error \ _(ENCAPSULATED, "good packets encapsulated") static char *vxlan_gbp_encap_error_strings[] = { #define _(sym,string) string, foreach_vxlan_gbp_encap_error #undef _ }; typedef enum { #define _(sym,str) VXLAN_GBP_ENCAP_ERROR_##sym, foreach_vxlan_gbp_encap_error #undef _ VXLAN_GBP_ENCAP_N_ERROR, } vxlan_gbp_encap_error_t; typedef enum { VXLAN_GBP_ENCAP_NEXT_DROP, VXLAN_GBP_ENCAP_N_NEXT, } vxlan_gbp_encap_next_t; typedef struct { u32 tunnel_index; u32 vni; u16 sclass; u8 flags; } vxlan_gbp_encap_trace_t; #ifndef CLIB_MARCH_VARIANT u8 * format_vxlan_gbp_encap_trace (u8 * s, va_list * args) { CLIB_UNUSED (vlib_main_t * vm) = va_arg (*args, vlib_main_t *); CLIB_UNUSED (vlib_node_t * node) = va_arg (*args, vlib_node_t *); vxlan_gbp_encap_trace_t *t = va_arg (*args, vxlan_gbp_encap_trace_t *); s = format (s, "VXLAN_GBP encap to vxlan_gbp_tunnel%d vni %d sclass %d flags %U", t->tunnel_index, t->vni, t->sclass, format_vxlan_gbp_header_gpflags, t->flags); return s; } #endif /* CLIB_MARCH_VARIANT */ always_inline uword vxlan_gbp_encap_inline (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame, u8 is_ip4, u8 csum_offload) { u32 n_left_from, next_index, *from, *to_next; vxlan_gbp_main_t *vxm = &vxlan_gbp_main; vnet_main_t *vnm = vxm->vnet_main; vnet_interface_main_t *im = &vnm->interface_main; vlib_combined_counter_main_t *tx_counter = im->combined_sw_if_counters + VNET_INTERFACE_COUNTER_TX; u32 pkts_encapsulated = 0; u32 thread_index = vlib_get_thread_index (); u32 sw_if_index0 = 0, sw_if_index1 = 0; u32 next0 = 0, next1 = 0; vxlan_gbp_tunnel_t *t0 = NULL, *t1 = NULL; index_t dpoi_idx0 = INDEX_INVALID, dpoi_idx1 = INDEX_INVALID; vlib_buffer_t *bufs[VLIB_FRAME_SIZE], **b = bufs; from = vlib_frame_vector_args (from_frame); n_left_from = from_frame->n_vectors; vlib_get_buffers (vm, from, bufs, n_left_from); next_index = node->cached_next_index; STATIC_ASSERT_SIZEOF (ip6_vxlan_gbp_header_t, 56); STATIC_ASSERT_SIZEOF (ip4_vxlan_gbp_header_t, 36); u8 const underlay_hdr_len = is_ip4 ? sizeof (ip4_vxlan_gbp_header_t) : sizeof (ip6_vxlan_gbp_header_t); u16 const l3_len = is_ip4 ? sizeof (ip4_header_t) : sizeof (ip6_header_t); u32 const csum_flags = is_ip4 ? VNET_BUFFER_F_IS_IP4 | VNET_BUFFER_F_L3_HDR_OFFSET_VALID | VNET_BUFFER_F_L4_HDR_OFFSET_VALID : VNET_BUFFER_F_IS_IP6 | VNET_BUFFER_F_L3_HDR_OFFSET_VALID | VNET_BUFFER_F_L4_HDR_OFFSET_VALID; u32 const outer_packet_csum_offload_flags = is_ip4 ? VNET_BUFFER_OFFLOAD_F_IP_CKSUM | VNET_BUFFER_OFFLOAD_F_UDP_CKSUM : VNET_BUFFER_OFFLOAD_F_UDP_CKSUM; u32 const inner_packet_removed_flags = VNET_BUFFER_F_IS_IP4 | VNET_BUFFER_F_IS_IP6 | VNET_BUFFER_F_L2_HDR_OFFSET_VALID | VNET_BUFFER_F_L3_HDR_OFFSET_VALID | VNET_BUFFER_F_L4_HDR_OFFSET_VALID; while (n_left_from > 0) { u32 n_left_to_next; vlib_get_next_frame (vm, node, next_index, to_next, n_left_to_next); while (n_left_from >= 4 && n_left_to_next >= 2) { /* Prefetch next iteration. */ { vlib_buffer_t *p2, *p3; p2 = vlib_get_buffer (vm, from[2]); p3 = vlib_get_buffer (vm, from[3]); vlib_prefetch_buffer_header (p2, LOAD); vlib_prefetch_buffer_header (p3, LOAD); CLIB_PREFETCH (b[2]->data - CLIB_CACHE_LINE_BYTES, 2 * CLIB_CACHE_LINE_BYTES, LOAD); CLIB_PREFETCH (b[3]->data - CLIB_CACHE_LINE_BYTES, 2 * CLIB_CACHE_LINE_BYTES, LOAD); } u32 bi0 = to_next[0] = from[0]; u32 bi1 = to_next[1] = from[1]; from += 2; to_next += 2; n_left_to_next -= 2; n_left_from -= 2; u32 or_flags = b[0]->flags | b[1]->flags; if (csum_offload && (or_flags & VNET_BUFFER_F_OFFLOAD)) { /* Only calculate the non-GSO packet csum offload */ if ((b[0]->flags & VNET_BUFFER_F_GSO) == 0) { vnet_calc_checksums_inline (vm, b[0], b[0]->flags & VNET_BUFFER_F_IS_IP4, b[0]->flags & VNET_BUFFER_F_IS_IP6); b[0]->flags &= ~inner_packet_removed_flags; } if ((b[1]->flags & VNET_BUFFER_F_GSO) == 0) { vnet_calc_checksums_inline (vm, b[1], b[1]->flags & VNET_BUFFER_F_IS_IP4, b[1]->flags & VNET_BUFFER_F_IS_IP6); b[1]->flags &= ~inner_packet_removed_flags; } } u32 flow_hash0 = vnet_l2_compute_flow_hash (b[0]); u32 flow_hash1 = vnet_l2_compute_flow_hash (b[1]); /* Get next node index and adj index from tunnel next_dpo */ if (sw_if_index0 != vnet_buffer (b[0])->sw_if_index[VLIB_TX]) { sw_if_index0 = vnet_buffer (b[0])->sw_if_index[VLIB_TX]; vnet_hw_interface_t *hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0); t0 = &vxm->tunnels[hi0->dev_instance]; /* Note: change to always set next0 if it may set to drop */ next0 = t0->next_dpo.dpoi_next_node; dpoi_idx0 = t0->next_dpo.dpoi_index; } /* Get next node index and adj index from tunnel next_dpo */ if (sw_if_index1 != vnet_buffer (b[1])->sw_if_index[VLIB_TX]) { if (sw_if_index0 == vnet_buffer (b[1])->sw_if_index[VLIB_TX]) { sw_if_index1 = sw_if_index0; t1 = t0; next1 = next0; dpoi_idx1 = dpoi_idx0; } else { sw_if_index1 = vnet_buffer (b[1])->sw_if_index[VLIB_TX]; vnet_hw_interface_t *hi1 = vnet_get_sup_hw_interface (vnm, sw_if_index1); t1 = &vxm->tunnels[hi1->dev_instance]; /* Note: change to always set next1 if it may set to drop */ next1 = t1->next_dpo.dpoi_next_node; dpoi_idx1 = t1->next_dpo.dpoi_index; } } vnet_buffer (b[0])->ip.adj_index[VLIB_TX] = dpoi_idx0; vnet_buffer (b[1])->ip.adj_index[VLIB_TX] = dpoi_idx1; ASSERT (t0->rewrite_header.data_bytes == underlay_hdr_len); ASSERT (t1->rewrite_header.data_bytes == underlay_hdr_len); vnet_rewrite_two_headers (*t0, *t1, vlib_buffer_get_current (b[0]), vlib_buffer_get_current (b[1]), underlay_hdr_len); vlib_buffer_advance (b[0], -underlay_hdr_len); vlib_buffer_advance (b[1], -underlay_hdr_len); u32 len0 = vlib_buffer_length_in_chain (vm, b[0]); u32 len1 = vlib_buffer_length_in_chain (vm, b[1]); u16 payload_l0 = clib_host_to_net_u16 (len0 - l3_len); u16 payload_l1 = clib_host_to_net_u16 (len1 - l3_len); void *underlay0 = vlib_buffer_get_current (b[0]); void *underlay1 = vlib_buffer_get_current (b[1]); ip4_header_t *ip4_0, *ip4_1; qos_bits_t ip4_0_tos = 0, ip4_1_tos = 0; ip6_header_t *ip6_0, *ip6_1; udp_header_t *udp0, *udp1; vxlan_gbp_header_t *vxlan_gbp0, *vxlan_gbp1; u8 *l3_0, *l3_1; if (is_ip4) { ip4_vxlan_gbp_header_t *hdr0 = underlay0; ip4_vxlan_gbp_header_t *hdr1 = underlay1; /* Fix the IP4 checksum and length */ ip4_0 = &hdr0->ip4; ip4_1 = &hdr1->ip4; ip4_0->length = clib_host_to_net_u16 (len0); ip4_1->length = clib_host_to_net_u16 (len1); if (PREDICT_FALSE (b[0]->flags & VNET_BUFFER_F_QOS_DATA_VALID)) { ip4_0_tos = vnet_buffer2 (b[0])->qos.bits; ip4_0->tos = ip4_0_tos; } if (PREDICT_FALSE (b[1]->flags & VNET_BUFFER_F_QOS_DATA_VALID)) { ip4_1_tos = vnet_buffer2 (b[1])->qos.bits; ip4_1->tos = ip4_1_tos; } l3_0 = (u8 *) ip4_0; l3_1 = (u8 *) ip4_1; udp0 = &hdr0->udp; udp1 = &hdr1->udp; vxlan_gbp0 = &hdr0->vxlan_gbp; vxlan_gbp1 = &hdr1->vxlan_gbp; } else /* ipv6 */ { ip6_vxlan_gbp_header_t *hdr0 = underlay0; ip6_vxlan_gbp_header_t *hdr1 = underlay1; /* Fix IP6 payload length */ ip6_0 = &hdr0->ip6; ip6_1 = &hdr1->ip6; ip6_0->payload_length = payload_l0; ip6_1->payload_length = payload_l1; l3_0 = (u8 *) ip6_0; l3_1 = (u8 *) ip6_1; udp0 = &hdr0->udp; udp1 = &hdr1->udp; vxlan_gbp0 = &hdr0->vxlan_gbp; vxlan_gbp1 = &hdr1->vxlan_gbp; } /* Fix UDP length and set source port */ udp0->length = payload_l0; udp0->src_port = flow_hash0; udp1->length = payload_l1; udp1->src_port = flow_hash1; /* set source class and gpflags */ vxlan_gbp0->gpflags = vnet_buffer2 (b[0])->gbp.flags; vxlan_gbp1->gpflags = vnet_buffer2 (b[1])->gbp.flags; vxlan_gbp0->sclass = clib_host_to_net_u16 (vnet_buffer2 (b[0])->gbp.sclass); vxlan_gbp1->sclass = clib_host_to_net_u16 (vnet_buffer2 (b[1])->gbp.sclass); if (csum_offload) { b[0]->flags |= csum_flags; vnet_buffer (b[0])->l3_hdr_offset = l3_0 - b[0]->data; vnet_buffer (b[0])->l4_hdr_offset = (u8 *) udp0 - b[0]->data; vnet_buffer_offload_flags_set (b[0], outer_packet_csum_offload_flags); b[1]->flags |= csum_flags; vnet_buffer (b[1])->l3_hdr_offset = l3_1 - b[1]->data; vnet_buffer (b[1])->l4_hdr_offset = (u8 *) udp1 - b[1]->data; vnet_buffer_offload_flags_set (b[1], outer_packet_csum_offload_flags); } /* IPv4 UDP checksum only if checksum offload is used */ else if (is_ip4) { ip_csum_t sum0 = ip4_0->checksum; sum0 = ip_csum_update (sum0, 0, ip4_0->length, ip4_header_t, length /* changed member */ ); if (PREDICT_FALSE (ip4_0_tos)) { sum0 = ip_csum_update (sum0, 0, ip4_0_tos, ip4_header_t, tos /* changed member */ ); } ip4_0->checksum = ip_csum_fold (sum0); ip_csum_t sum1 = ip4_1->checksum; sum1 = ip_csum_update (sum1, 0, ip4_1->length, ip4_header_t, length /* changed member */ ); if (PREDICT_FALSE (ip4_1_tos)) { sum1 = ip_csum_update (sum1, 0, ip4_1_tos, ip4_header_t, tos /* changed member */ ); } ip4_1->checksum = ip_csum_fold (sum1); } /* IPv6 UDP checksum is mandatory */ else { int bogus = 0; udp0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b[0], ip6_0, &bogus); ASSERT (bogus == 0); if (udp0->checksum == 0) udp0->checksum = 0xffff; udp1->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b[1], ip6_1, &bogus); ASSERT (bogus == 0); if (udp1->checksum == 0) udp1->checksum = 0xffff; } /* save inner packet flow_hash for load-balance node */ vnet_buffer (b[0])->ip.flow_hash = flow_hash0; vnet_buffer (b[1])->ip.flow_hash = flow_hash1; vlib_increment_combined_counter (tx_counter, thread_index, sw_if_index0, 1, len0); vlib_increment_combined_counter (tx_counter, thread_index, sw_if_index1, 1, len1); pkts_encapsulated += 2; if (PREDICT_FALSE (b[0]->flags & VLIB_BUFFER_IS_TRACED)) { vxlan_gbp_encap_trace_t *tr = vlib_add_trace (vm, node, b[0], sizeof (*tr)); tr->tunnel_index = t0 - vxm->tunnels; tr->vni = t0->vni; tr->sclass = vnet_buffer2 (b[0])->gbp.sclass; tr->flags = vnet_buffer2 (b[0])->gbp.flags; } if (PREDICT_FALSE (b[1]->flags & VLIB_BUFFER_IS_TRACED)) { vxlan_gbp_encap_trace_t *tr = vlib_add_trace (vm, node, b[1], sizeof (*tr)); tr->tunnel_index = t1 - vxm->tunnels; tr->vni = t1->vni; tr->sclass = vnet_buffer2 (b[1])->gbp.sclass; tr->flags = vnet_buffer2 (b[1])->gbp.flags; } b += 2; vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next, n_left_to_next, bi0, bi1, next0, next1); } while (n_left_from > 0 && n_left_to_next > 0) { u32 bi0 = to_next[0] = from[0]; from += 1; to_next += 1; n_left_from -= 1; n_left_to_next -= 1; if (csum_offload && (b[0]->flags & VNET_BUFFER_F_OFFLOAD)) { /* Only calculate the non-GSO packet csum offload */ if ((b[0]->flags & VNET_BUFFER_F_GSO) == 0) { vnet_calc_checksums_inline (vm, b[0], b[0]->flags & VNET_BUFFER_F_IS_IP4, b[0]->flags & VNET_BUFFER_F_IS_IP6); b[0]->flags &= ~inner_packet_removed_flags; } } u32 flow_hash0 = vnet_l2_compute_flow_hash (b[0]); /* Get next node index and adj index from tunnel next_dpo */ if (sw_if_index0 != vnet_buffer (b[0])->sw_if_index[VLIB_TX]) { sw_if_index0 = vnet_buffer (b[0])->sw_if_index[VLIB_TX]; vnet_hw_interface_t *hi0 = vnet_get_sup_hw_interface (vnm, sw_if_index0); t0 = &vxm->tunnels[hi0->dev_instance]; /* Note: change to always set next0 if it may be set to drop */ next0 = t0->next_dpo.dpoi_next_node; dpoi_idx0 = t0->next_dpo.dpoi_index; } vnet_buffer (b[0])->ip.adj_index[VLIB_TX] = dpoi_idx0; ASSERT (t0->rewrite_header.data_bytes == underlay_hdr_len); vnet_rewrite_one_header (*t0, vlib_buffer_get_current (b[0]), underlay_hdr_len); vlib_buffer_advance (b[0], -underlay_hdr_len); void *underlay0 = vlib_buffer_get_current (b[0]); u32 len0 = vlib_buffer_length_in_chain (vm, b[0]); u16 payload_l0 = clib_host_to_net_u16 (len0 - l3_len); vxlan_gbp_header_t *vxlan_gbp0; udp_header_t *udp0; ip4_header_t *ip4_0; qos_bits_t ip4_0_tos = 0; ip6_header_t *ip6_0; u8 *l3_0; if (is_ip4) { ip4_vxlan_gbp_header_t *hdr = underlay0; /* Fix the IP4 checksum and length */ ip4_0 = &hdr->ip4; ip4_0->length = clib_host_to_net_u16 (len0); if (PREDICT_FALSE (b[0]->flags & VNET_BUFFER_F_QOS_DATA_VALID)) { ip4_0_tos = vnet_buffer2 (b[0])->qos.bits; ip4_0->tos = ip4_0_tos; } l3_0 = (u8 *) ip4_0; udp0 = &hdr->udp; vxlan_gbp0 = &hdr->vxlan_gbp; } else /* ip6 path */ { ip6_vxlan_gbp_header_t *hdr = underlay0; /* Fix IP6 payload length */ ip6_0 = &hdr->ip6; ip6_0->payload_length = payload_l0; l3_0 = (u8 *) ip6_0; udp0 = &hdr->udp; vxlan_gbp0 = &hdr->vxlan_gbp; } /* Fix UDP length and set source port */ udp0->length = payload_l0; udp0->src_port = flow_hash0; /* set source class and gpflags */ vxlan_gbp0->gpflags = vnet_buffer2 (b[0])->gbp.flags; vxlan_gbp0->sclass = clib_host_to_net_u16 (vnet_buffer2 (b[0])->gbp.sclass); if (csum_offload) { b[0]->flags |= csum_flags; vnet_buffer (b[0])->l3_hdr_offset = l3_0 - b[0]->data; vnet_buffer (b[0])->l4_hdr_offset = (u8 *) udp0 - b[0]->data; vnet_buffer_offload_flags_set (b[0], outer_packet_csum_offload_flags); } /* IPv4 UDP checksum only if checksum offload is used */ else if (is_ip4) { ip_csum_t sum0 = ip4_0->checksum; sum0 = ip_csum_update (sum0, 0, ip4_0->length, ip4_header_t, length /* changed member */ ); if (PREDICT_FALSE (ip4_0_tos)) { sum0 = ip_csum_update (sum0, 0, ip4_0_tos, ip4_header_t, tos /* changed member */ ); } ip4_0->checksum = ip_csum_fold (sum0); } /* IPv6 UDP checksum is mandatory */ else { int bogus = 0; udp0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b[0], ip6_0, &bogus); ASSERT (bogus == 0); if (udp0->checksum == 0) udp0->checksum = 0xffff; } /* save inner packet flow_hash for load-balance node */ vnet_buffer (b[0])->ip.flow_hash = flow_hash0; vlib_increment_combined_counter (tx_counter, thread_index, sw_if_index0, 1, len0); pkts_encapsulated++; if (PREDICT_FALSE (b[0]->flags & VLIB_BUFFER_IS_TRACED)) { vxlan_gbp_encap_trace_t *tr = vlib_add_trace (vm, node, b[0], sizeof (*tr)); tr->tunnel_index = t0 - vxm->tunnels; tr->vni = t0->vni; tr->sclass = vnet_buffer2 (b[0])->gbp.sclass; tr->flags = vnet_buffer2 (b[0])->gbp.flags; } b += 1; vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, n_left_to_next, bi0, next0); } vlib_put_next_frame (vm, node, next_index, n_left_to_next); } /* Do we still need this now that tunnel tx stats is kept? */ vlib_node_increment_counter (vm, node->node_index, VXLAN_GBP_ENCAP_ERROR_ENCAPSULATED, pkts_encapsulated); return from_frame->n_vectors; } VLIB_NODE_FN (vxlan4_gbp_encap_node) (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame) { /* Disable chksum offload as setup overhead in tx node is not worthwhile for ip4 header checksum only, unless udp checksum is also required */ return vxlan_gbp_encap_inline (vm, node, from_frame, /* is_ip4 */ 1, /* csum_offload */ 0); } VLIB_NODE_FN (vxlan6_gbp_encap_node) (vlib_main_t * vm, vlib_node_runtime_t * node, vlib_frame_t * from_frame) { /* Enable checksum offload for ip6 as udp checksum is mandatory, */ return vxlan_gbp_encap_inline (vm, node, from_frame, /* is_ip4 */ 0, /* csum_offload */ 1); } /* *INDENT-OFF* */ VLIB_REGISTER_NODE (vxlan4_gbp_encap_node) = { .name = "vxlan4-gbp-encap", .vector_size = sizeof (u32), .format_trace = format_vxlan_gbp_encap_trace, .type = VLIB_NODE_TYPE_INTERNAL, .n_errors = ARRAY_LEN (vxlan_gbp_encap_error_strings), .error_strings = vxlan_gbp_encap_error_strings, .n_next_nodes = VXLAN_GBP_ENCAP_N_NEXT, .next_nodes = { [VXLAN_GBP_ENCAP_NEXT_DROP] = "error-drop", }, }; VLIB_REGISTER_NODE (vxlan6_gbp_encap_node) = { .name = "vxlan6-gbp-encap", .vector_size = sizeof (u32), .format_trace = format_vxlan_gbp_encap_trace, .type = VLIB_NODE_TYPE_INTERNAL, .n_errors = ARRAY_LEN (vxlan_gbp_encap_error_strings), .error_strings = vxlan_gbp_encap_error_strings, .n_next_nodes = VXLAN_GBP_ENCAP_N_NEXT, .next_nodes = { [VXLAN_GBP_ENCAP_NEXT_DROP] = "error-drop", }, }; /* *INDENT-ON* */ /* * fd.io coding-style-patch-verification: ON * * Local Variables: * eval: (c-set-style "gnu") * End: */


context:
space:
mode: