Repository navigation
Wired Offload
Julius Bairaktaris edited this page Oct 19, 2026
·
2 revisions
qca_ppe programs the PPE through the kernel's standard offload
interfaces. No firmware is involved.
-
Routed traffic (IPv4/IPv6, NAT, PPPoE, 802.1Q VLAN) goes through the
fw4 flowtable. It needs hardware flow offloading, which the image turns
on: Network → Firewall → Hardware flow offloading, or
flow_offloading_hw='1'infirewall.@defaults[0]. - Bridged traffic goes through switchdev: FDB, VLAN filtering, STP, LAG and IGMP/MLD snooping. Always on.
-
tc and ethtool:
tbf,ets,prio,mqprio,red,gred,tc flowerand n-tuple rules,matchallpolice and mirroring,dcb apppriority maps.
The Wi-Fi offload takes routed flows from the same flowtable, so with hardware flow offloading off, routed Wi-Fi flows are not offloaded either.
mount -t debugfs none /sys/kernel/debug 2>/dev/null
grep -c HW_OFFLOAD /proc/net/nf_conntrack # connections in hardware
cat /sys/kernel/debug/*.ppe/offload # offloaded and rejected flows
cat /sys/kernel/debug/*.ppe/flows # the PPE flow table[OFFLOAD] without [HW_OFFLOAD] means the software flowtable only. Short
connections may end before they reach hardware.