summaryrefslogtreecommitdiffstats
path: root/VNFs
diff options
context:
space:
mode:
authorXavier Simonart <xavier.simonart@intel.com>2018-09-04 11:58:36 +0200
committerXavier Simonart <xavier.simonart@intel.com>2018-09-14 14:30:36 +0000
commit3afd384b7e68a3cd29f4d2cdf39785f8441d0b6f (patch)
treea7a2e5705dd088b1c95a3de6afcf6271993153a2 /VNFs
parenta3a7fbb44e261987e2cd800fb049c20b16381e4d (diff)
Improve l3fwd performance
JIRA: SAMPLEVNF-128 Improve performance of l3fwd use-case. L3fwd did not disable TX offload flags, which resulted in non-vector mode to be used in pmd in some cases like bare-metal. Note that in some cases such as virtio, vector mode was already used as OVS reports that it does not support TX-offload, hence TX-offload was already disabled on virtio. Using vector mode usually brings higher performance. Change-Id: I102b846d604a9cf28c6b7dae8bedbe918ccfcf6c Signed-off-by: Xavier Simonart <xavier.simonart@intel.com>
Diffstat (limited to 'VNFs')
-rw-r--r--VNFs/DPPD-PROX/handle_routing.c12
1 files changed, 3 insertions, 9 deletions
diff --git a/VNFs/DPPD-PROX/handle_routing.c b/VNFs/DPPD-PROX/handle_routing.c
index 9dd45ed8..8cdc8427 100644
--- a/VNFs/DPPD-PROX/handle_routing.c
+++ b/VNFs/DPPD-PROX/handle_routing.c
@@ -48,7 +48,6 @@ struct task_routing {
struct lcore_cfg *lconf;
struct rte_lpm *ipv4_lpm;
struct next_hop *next_hops;
- int offload_crc;
uint32_t number_free_rules;
uint16_t qinq_tag;
uint32_t marking[4];
@@ -144,9 +143,6 @@ static void init_task_routing(struct task_base *tbase, struct task_args *targ)
}
struct prox_port_cfg *port = find_reachable_port(targ);
- if (port) {
- task->offload_crc = port->capabilities.tx_offload_cksum;
- }
targ->lconf->ctrl_func_m[targ->task] = routing_update;
targ->lconf->ctrl_timeout = freq_to_tsc(20);
@@ -186,11 +182,9 @@ static void set_l2(struct task_routing *task, struct rte_mbuf *mbuf, uint8_t nh_
*((uint64_t *)(&peth->s_addr)) = task->src_mac[task->next_hops[nh_idx].mac_port.out_idx];
}
-static void set_l2_mpls(struct task_routing *task, struct rte_mbuf *mbuf, uint8_t nh_idx, uint16_t l2_len)
+static void set_l2_mpls(struct task_routing *task, struct rte_mbuf *mbuf, uint8_t nh_idx)
{
struct ether_hdr *peth = (struct ether_hdr *)rte_pktmbuf_prepend(mbuf, sizeof(struct mpls_hdr));
- l2_len += sizeof(struct mpls_hdr);
- prox_ip_cksum(mbuf, (struct ipv4_hdr *)((uint8_t *)peth + l2_len), l2_len, sizeof(struct ipv4_hdr), task->offload_crc);
*((uint64_t *)(&peth->d_addr)) = task->next_hops[nh_idx].mac_port_8bytes;
*((uint64_t *)(&peth->s_addr)) = task->src_mac[task->next_hops[nh_idx].mac_port.out_idx];
@@ -255,7 +249,7 @@ static uint8_t route_ipv4(struct task_routing *task, uint8_t *beg, uint32_t ip_o
rte_pktmbuf_trim(mbuf, padlen);
}
- set_l2_mpls(task, mbuf, next_hop_index, ip_offset);
+ set_l2_mpls(task, mbuf, next_hop_index);
}
else {
set_l2(task, mbuf, next_hop_index);
@@ -311,7 +305,7 @@ static struct task_init task_init_routing = {
.mode_str = "routing",
.init = init_task_routing,
.handle = handle_routing_bulk,
- .flag_features = TASK_FEATURE_ROUTING,
+ .flag_features = TASK_FEATURE_ROUTING|TASK_FEATURE_TXQ_FLAGS_NOOFFLOADS,
.size = sizeof(struct task_routing)
};