[dpdk-dev] net/bonding: strengthen the judgment of lacp packets
Checks
Commit Message
When the nic does not support vlan rx offload may be wrong, resulting in
lacp packets will not be processed.
---
drivers/net/bonding/rte_eth_bond_pmd.c | 7 ++++---
1 file changed, 4 insertions(+), 3 deletions(-)
@@ -125,11 +125,12 @@
}
static inline uint8_t
-is_lacp_packets(uint16_t ethertype, uint8_t subtype, uint16_t vlan_tci)
+is_lacp_packets(uint16_t ethertype, uint8_t subtype, struct rte_mbuf *mbuf)
{
const uint16_t ether_type_slow_be = rte_be_to_cpu_16(ETHER_TYPE_SLOW);
- return !vlan_tci && (ethertype == ether_type_slow_be &&
+ return !((mbuf->ol_flags & PKT_RX_VLAN_PKT) ? mbuf->vlan_tci : 0) &&
+ (ethertype == ether_type_slow_be &&
(subtype == SLOW_SUBTYPE_MARKER || subtype == SLOW_SUBTYPE_LACP));
}
@@ -444,7 +445,7 @@
/* Remove packet from array if it is slow packet or slave is not
* in collecting state or bonding interface is not in promiscuous
* mode and packet address does not match. */
- if (unlikely(is_lacp_packets(hdr->ether_type, subtype, bufs[j]->vlan_tci) ||
+ if (unlikely(is_lacp_packets(hdr->ether_type, subtype, bufs[j]) ||
!collecting || (!promisc &&
!is_multicast_ether_addr(&hdr->d_addr) &&
!is_same_ether_addr(&bond_mac, &hdr->d_addr)))) {