[dpdk-dev] Breakage in l3fwd example with new event main loop

Stephen Hemminger stephen at networkplumber.org
Fri May 22 01:09:40 CEST 2020

The l3fwd implement to do events has lots of copy/paste code.
Much of this must be never tested because it has compile errors.


static __rte_always_inline uint16_t
lpm_process_event_pkt(const struct lcore_conf *lconf, struct rte_mbuf *mbuf)
	mbuf->port = lpm_get_dst_port(lconf, mbuf, mbuf->port);

#if defined RTE_ARCH_X86 || defined RTE_MACHINE_CPUFLAG_NEON \
	|| defined RTE_ARCH_PPC_64
	process_packet(mbuf, &mbuf->port);

	struct rte_ether_hdr *eth_hdr = rte_pktmbuf_mtod(mbuf,
			struct rte_ether_hdr *);
#ifdef DO_RFC_1812_CHECKS
	struct rte_ipv4_hdr *ipv4_hdr;
	if (RTE_ETH_IS_IPV4_HDR(mbuf->packet_type)) {
		/* Handle IPv4 headers.*/
		ipv4_hdr = rte_pktmbuf_mtod_offset(mbuf,
				struct rte_ipv4_hdr *,
				sizeof(struct rte_ether_hdr));

		if (is_valid_ipv4_pkt(ipv4_hdr, mbuf->pkt_len)
				< 0) {
			mbuf->port = BAD_PORT;
			continue; <<<<<<<<< not in loop?

If the checks are enabled, the code doesn't compile because continue
is not in a loop.

Which leads to a number of observations:
   - copy/paste is bad, see DNR principle. Better to make it an inline
   - #ifdef are evil since no one ever tests the other half
   - why did CI not catch this?
   - do you ever build on other architectures?

Please fix!

More information about the dev mailing list