diff options
author | Zhiyong Yang <zhiyong.yang@intel.com> | 2019-05-20 01:31:38 -0400 |
---|---|---|
committer | Damjan Marion <dmarion@me.com> | 2019-05-27 10:06:28 +0000 |
commit | a80d9cfcac02f59c15a57bd6a9a73a523e3499fc (patch) | |
tree | befed138a591e9d287b553f05beca9e5c1bb9c8d /src | |
parent | a462c070ab9bf593d7ed3c454ea3267929550a62 (diff) |
geneve: encap performance enhancement
This patch can help save 2.7 clocks/pkt from 51.5 to 48.5
clocks/pkt on Skylake server.
Change-Id: I10173c8a147a0e54f925c7841c26f133eb75cbed
Signed-off-by: Zhiyong Yang <zhiyong.yang@intel.com>
Diffstat (limited to 'src')
-rw-r--r-- | src/vnet/geneve/encap.c | 111 |
1 files changed, 53 insertions, 58 deletions
diff --git a/src/vnet/geneve/encap.c b/src/vnet/geneve/encap.c index 3d98f9e2ba6..8e59aea3626 100644 --- a/src/vnet/geneve/encap.c +++ b/src/vnet/geneve/encap.c @@ -66,9 +66,11 @@ geneve_encap_inline (vlib_main_t * vm, u32 next0 = 0, next1 = 0; vnet_hw_interface_t *hi0, *hi1; geneve_tunnel_t *t0 = NULL, *t1 = NULL; + 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; stats_sw_if_index = node->runtime_data[0]; @@ -83,7 +85,6 @@ geneve_encap_inline (vlib_main_t * vm, while (n_left_from >= 4 && n_left_to_next >= 2) { u32 bi0, bi1; - vlib_buffer_t *b0, *b1; u32 flow_hash0, flow_hash1; u32 len0, len1; ip4_header_t *ip4_0, *ip4_1; @@ -98,16 +99,13 @@ geneve_encap_inline (vlib_main_t * vm, /* Prefetch next iteration. */ { - vlib_buffer_t *p2, *p3; + vlib_prefetch_buffer_header (b[2], LOAD); + vlib_prefetch_buffer_header (b[3], LOAD); - 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 (p2->data, 2 * CLIB_CACHE_LINE_BYTES, LOAD); - CLIB_PREFETCH (p3->data, 2 * CLIB_CACHE_LINE_BYTES, 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); } bi0 = from[0]; @@ -119,16 +117,13 @@ geneve_encap_inline (vlib_main_t * vm, n_left_to_next -= 2; n_left_from -= 2; - b0 = vlib_get_buffer (vm, bi0); - b1 = vlib_get_buffer (vm, bi1); - - flow_hash0 = vnet_l2_compute_flow_hash (b0); - flow_hash1 = vnet_l2_compute_flow_hash (b1); + flow_hash0 = vnet_l2_compute_flow_hash (b[0]); + 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 (b0)->sw_if_index[VLIB_TX]) + if (sw_if_index0 != vnet_buffer (b[0])->sw_if_index[VLIB_TX]) { - sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_TX]; + sw_if_index0 = vnet_buffer (b[0])->sw_if_index[VLIB_TX]; 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 */ @@ -137,12 +132,12 @@ geneve_encap_inline (vlib_main_t * vm, ASSERT (t0 != NULL); - vnet_buffer (b0)->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index; + vnet_buffer (b[0])->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index; /* Get next node index and adj index from tunnel next_dpo */ - if (sw_if_index1 != vnet_buffer (b1)->sw_if_index[VLIB_TX]) + if (sw_if_index1 != vnet_buffer (b[1])->sw_if_index[VLIB_TX]) { - sw_if_index1 = vnet_buffer (b1)->sw_if_index[VLIB_TX]; + sw_if_index1 = vnet_buffer (b[1])->sw_if_index[VLIB_TX]; 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 be set to drop */ @@ -151,11 +146,11 @@ geneve_encap_inline (vlib_main_t * vm, ASSERT (t1 != NULL); - vnet_buffer (b1)->ip.adj_index[VLIB_TX] = t1->next_dpo.dpoi_index; + vnet_buffer (b[1])->ip.adj_index[VLIB_TX] = t1->next_dpo.dpoi_index; /* Apply the rewrite string. $$$$ vnet_rewrite? */ - vlib_buffer_advance (b0, -(word) _vec_len (t0->rewrite)); - vlib_buffer_advance (b1, -(word) _vec_len (t1->rewrite)); + vlib_buffer_advance (b[0], -(word) _vec_len (t0->rewrite)); + vlib_buffer_advance (b[1], -(word) _vec_len (t1->rewrite)); if (is_ip4) { @@ -171,8 +166,8 @@ geneve_encap_inline (vlib_main_t * vm, ASSERT (vec_len (t0->rewrite) == ip4_geneve_header_total_len0); ASSERT (vec_len (t1->rewrite) == ip4_geneve_header_total_len1); - ip4_0 = vlib_buffer_get_current (b0); - ip4_1 = vlib_buffer_get_current (b1); + ip4_0 = vlib_buffer_get_current (b[0]); + ip4_1 = vlib_buffer_get_current (b[1]); /* Copy the fixed header */ copy_dst0 = (u64 *) ip4_0; @@ -197,14 +192,14 @@ geneve_encap_inline (vlib_main_t * vm, /* Fix the IP4 checksum and length */ sum0 = ip4_0->checksum; new_l0 = /* old_l0 always 0, see the rewrite setup */ - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b[0])); sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, length /* changed member */ ); ip4_0->checksum = ip_csum_fold (sum0); ip4_0->length = new_l0; sum1 = ip4_1->checksum; new_l1 = /* old_l1 always 0, see the rewrite setup */ - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1)); + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b[1])); sum1 = ip_csum_update (sum1, old_l1, new_l1, ip4_header_t, length /* changed member */ ); ip4_1->checksum = ip_csum_fold (sum1); @@ -213,13 +208,13 @@ geneve_encap_inline (vlib_main_t * vm, /* Fix UDP length and set source port */ udp0 = (udp_header_t *) (ip4_0 + 1); new_l0 = - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b[0]) - sizeof (*ip4_0)); udp0->length = new_l0; udp0->src_port = flow_hash0; udp1 = (udp_header_t *) (ip4_1 + 1); new_l1 = - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) - + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b[1]) - sizeof (*ip4_1)); udp1->length = new_l1; udp1->src_port = flow_hash1; @@ -240,8 +235,8 @@ geneve_encap_inline (vlib_main_t * vm, ASSERT (vec_len (t0->rewrite) == ip6_geneve_header_total_len0); ASSERT (vec_len (t1->rewrite) == ip6_geneve_header_total_len1); - ip6_0 = vlib_buffer_get_current (b0); - ip6_1 = vlib_buffer_get_current (b1); + ip6_0 = vlib_buffer_get_current (b[0]); + ip6_1 = vlib_buffer_get_current (b[1]); /* Copy the fixed header */ copy_dst0 = (u64 *) ip6_0; @@ -257,11 +252,11 @@ geneve_encap_inline (vlib_main_t * vm, #undef _ /* Fix IP6 payload length */ new_l0 = - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b[0]) - sizeof (*ip6_0)); ip6_0->payload_length = new_l0; new_l1 = - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b1) + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b[1]) - sizeof (*ip6_1)); ip6_1->payload_length = new_l1; @@ -274,13 +269,13 @@ geneve_encap_inline (vlib_main_t * vm, udp1->src_port = flow_hash1; /* IPv6 UDP checksum is mandatory */ - udp0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0, + 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, b1, + udp1->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b[1], ip6_1, &bogus); ASSERT (bogus == 0); @@ -289,8 +284,8 @@ geneve_encap_inline (vlib_main_t * vm, } pkts_encapsulated += 2; - len0 = vlib_buffer_length_in_chain (vm, b0); - len1 = vlib_buffer_length_in_chain (vm, b1); + len0 = vlib_buffer_length_in_chain (vm, b[0]); + len1 = vlib_buffer_length_in_chain (vm, b[1]); stats_n_packets += 2; stats_n_bytes += len0 + len1; @@ -325,21 +320,22 @@ geneve_encap_inline (vlib_main_t * vm, } } - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + if (PREDICT_FALSE (b[0]->flags & VLIB_BUFFER_IS_TRACED)) { geneve_encap_trace_t *tr = - vlib_add_trace (vm, node, b0, sizeof (*tr)); + vlib_add_trace (vm, node, b[0], sizeof (*tr)); tr->tunnel_index = t0 - vxm->tunnels; tr->vni = t0->vni; } - if (PREDICT_FALSE (b1->flags & VLIB_BUFFER_IS_TRACED)) + if (PREDICT_FALSE (b[1]->flags & VLIB_BUFFER_IS_TRACED)) { geneve_encap_trace_t *tr = - vlib_add_trace (vm, node, b1, sizeof (*tr)); + vlib_add_trace (vm, node, b[1], sizeof (*tr)); tr->tunnel_index = t1 - vxm->tunnels; tr->vni = t1->vni; } + b += 2; vlib_validate_buffer_enqueue_x2 (vm, node, next_index, to_next, n_left_to_next, @@ -349,7 +345,6 @@ geneve_encap_inline (vlib_main_t * vm, while (n_left_from > 0 && n_left_to_next > 0) { u32 bi0; - vlib_buffer_t *b0; u32 flow_hash0; u32 len0; ip4_header_t *ip4_0; @@ -367,23 +362,21 @@ geneve_encap_inline (vlib_main_t * vm, n_left_from -= 1; n_left_to_next -= 1; - b0 = vlib_get_buffer (vm, bi0); - - flow_hash0 = vnet_l2_compute_flow_hash (b0); + 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 (b0)->sw_if_index[VLIB_TX]) + if (sw_if_index0 != vnet_buffer (b[0])->sw_if_index[VLIB_TX]) { - sw_if_index0 = vnet_buffer (b0)->sw_if_index[VLIB_TX]; + sw_if_index0 = vnet_buffer (b[0])->sw_if_index[VLIB_TX]; 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; } - vnet_buffer (b0)->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index; + vnet_buffer (b[0])->ip.adj_index[VLIB_TX] = t0->next_dpo.dpoi_index; /* Apply the rewrite string. $$$$ vnet_rewrite? */ - vlib_buffer_advance (b0, -(word) _vec_len (t0->rewrite)); + vlib_buffer_advance (b[0], -(word) _vec_len (t0->rewrite)); if (is_ip4) { @@ -396,7 +389,7 @@ geneve_encap_inline (vlib_main_t * vm, #endif ASSERT (vec_len (t0->rewrite) == ip4_geneve_header_total_len0); - ip4_0 = vlib_buffer_get_current (b0); + ip4_0 = vlib_buffer_get_current (b[0]); /* Copy the fixed header */ copy_dst0 = (u64 *) ip4_0; @@ -413,7 +406,7 @@ geneve_encap_inline (vlib_main_t * vm, /* Fix the IP4 checksum and length */ sum0 = ip4_0->checksum; new_l0 = /* old_l0 always 0, see the rewrite setup */ - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0)); + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b[0])); sum0 = ip_csum_update (sum0, old_l0, new_l0, ip4_header_t, length /* changed member */ ); ip4_0->checksum = ip_csum_fold (sum0); @@ -422,7 +415,7 @@ geneve_encap_inline (vlib_main_t * vm, /* Fix UDP length and set source port */ udp0 = (udp_header_t *) (ip4_0 + 1); new_l0 = - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) - + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b[0]) - sizeof (*ip4_0)); udp0->length = new_l0; udp0->src_port = flow_hash0; @@ -441,7 +434,7 @@ geneve_encap_inline (vlib_main_t * vm, #endif ASSERT (vec_len (t0->rewrite) == ip6_geneve_header_total_len0); - ip6_0 = vlib_buffer_get_current (b0); + ip6_0 = vlib_buffer_get_current (b[0]); /* Copy the fixed header */ copy_dst0 = (u64 *) ip6_0; copy_src0 = (u64 *) t0->rewrite; @@ -451,7 +444,7 @@ geneve_encap_inline (vlib_main_t * vm, #undef _ /* Fix IP6 payload length */ new_l0 = - clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b0) + clib_host_to_net_u16 (vlib_buffer_length_in_chain (vm, b[0]) - sizeof (*ip6_0)); ip6_0->payload_length = new_l0; @@ -461,7 +454,7 @@ geneve_encap_inline (vlib_main_t * vm, udp0->src_port = flow_hash0; /* IPv6 UDP checksum is mandatory */ - udp0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b0, + udp0->checksum = ip6_tcp_udp_icmp_compute_checksum (vm, b[0], ip6_0, &bogus); ASSERT (bogus == 0); @@ -470,7 +463,7 @@ geneve_encap_inline (vlib_main_t * vm, } pkts_encapsulated++; - len0 = vlib_buffer_length_in_chain (vm, b0); + len0 = vlib_buffer_length_in_chain (vm, b[0]); stats_n_packets += 1; stats_n_bytes += len0; @@ -492,13 +485,15 @@ geneve_encap_inline (vlib_main_t * vm, stats_sw_if_index = sw_if_index0; } - if (PREDICT_FALSE (b0->flags & VLIB_BUFFER_IS_TRACED)) + if (PREDICT_FALSE (b[0]->flags & VLIB_BUFFER_IS_TRACED)) { geneve_encap_trace_t *tr = - vlib_add_trace (vm, node, b0, sizeof (*tr)); + vlib_add_trace (vm, node, b[0], sizeof (*tr)); tr->tunnel_index = t0 - vxm->tunnels; tr->vni = t0->vni; } + b += 1; + vlib_validate_buffer_enqueue_x1 (vm, node, next_index, to_next, n_left_to_next, bi0, next0); |