On 2026-08-25 17:22:22, Stuart Henderson wrote: > no need to guess, just build from source so you know what's in it
I did that to test Theo's diff. Theo then posted "wait" as he saw kern_pledge.c needed patching to allow SIOCGIFDATA in pledge "route". So I waited for a newer snapshot with the patched kernel to retest. The original pledge issue in my bug report is fixed only to be followed by another pledge issue: pledge "inet", syscall 105. I was curious what should the "-d" debug output supposed to look like So I removed the pledge line, recompiled, and tested. lldpd[96942]: route message: 26 bytes lldpd[96942]: tap0 index 21: attached lldpd[96942]: route message: 26 bytes lldpd[96942]: tap0 index 21: del membership: Device not configured lldpd[96942]: tap0 index 21: detached lldpd[96942]: route message: 26 bytes lldpd[96942]: tap0 index 22: attached lldpd[96942]: route message: 26 bytes lldpd[96942]: tap0 index 22: del membership: Device not configured lldpd[96942]: tap0 index 22: detached Ok, why is lldpd trying to "del membership" on an interface already destroyed? I see why Theo said "I'm surprised, it is like this code was never tested." lldpd still has issues (pledge and del membership) with virtual interfaces. Runs ok as long as interfaces stay the same but dies if interfaces change.
