On Mon, Oct 06, 2025 at 09:22:46AM +0200, Denis Fondras wrote: > Hello, > > Since I upgraded to a recent snapshot (OpenBSD rt-grav-02 7.8 GENERIC.MP#45 > amd64) later last week, one of my ixl interface goes OACTIVE every few hour > and > my BGP peering goes down with holdtimer expired (don't know if that's related > though). > > ixl3: flags=8c43<UP,BROADCAST,RUNNING,OACTIVE,SIMPLEX,MULTICAST> mtu 1500 > lladdr 00:00:00:00:01:03 > index 8 priority 0 llprio 3 > media: Ethernet autoselect (10GbaseLR full-duplex) > status: active > inet6 fe80::200:ff:fe00:103%ixl3 prefixlen 64 scopeid 0x8 > inet6 2001:7f8:xxxx prefixlen 64 > inet 185.xxxx netmask 0xffffff00 broadcast 185.xxxx > > Any idea what can be wrong ? > > Only ixl3 exhibits this behaviour, ixl0-2 seems fine. > > syslog only shows the holdtimer expired. > > I can try to upgrade though the timing is really unfortunate as my second > OpenBSD router (same version, same hardware) is under maintenance :D
I guess your ixl3 gets more bursty traffic. OACTIVE meanst that the TX dma ring is full and the device needs to push out packets before accepting new ones. Not sure how OACTIVE and multiqueue devices work together. That's a question for dlg@ to answer. Please provide netstat -I ixl3 and netstat -e -I ixl3 outputs and the kstat output for its queues. -- :wq Claudio
