Hi Aaron,
I tested this series combined with your kernel sockmap branch
(sockmap_2026_feb, built as a scratch kernel based on RHEL 9.8 /
5.14.0-611.el9_7) on a single-node x86_64 machine.
Tests performed:
- ODP action round-trip parsing (valid and invalid cases): pass
- Socket action generation via ofproto/trace for TCP (IPv4/IPv6): pass
- Non-TCP exclusion (ICMP, UDP): pass
- socket_lookup enable/disable per port: pass
- socket_lookup with group recirculation: pass
- OpenFlow regression with socket_lookup enabled: pass
- Conntrack regression with socket_lookup enabled: pass
- 2-namespace TCP performance (IPv4 and IPv6): pass
During 1000-namespace scale testing (2000 veth pairs with socket_lookup
enabled on all ports), a kernel WARNING burst was observed in the
console log:
WARNING: CPU: 118 PID: 304284 at net/core/skbuff.c:1000
skb_release_head_state+0x95/0xa0
(185 occurrences within 79ms, followed by BUG: scheduling while
atomic in the OVS upcall handler thread handler2052)
Root cause analysis: ARP table overflow under 2000 veth pairs floods
the OVS netlink upcall path, triggering a pre-existing kernel bug where
skbs with netlink_skb_destructor are consumed without prior skb_orphan().
The same WARNING is reproducible on a stock RHEL 9.8 kernel with stock
OVS 3.7 (without this patch series), confirming it is unrelated to
these changes.
Tested-by: Minxi Hou <[email protected]>
_______________________________________________
dev mailing list
[email protected]
https://mail.openvswitch.org/mailman/listinfo/ovs-dev
