`InterfaceId` is not removed from resolved `ScopedIpV4` in case the interface is removed (goes down)

[hrzlgnm]

While playing around more with the addition of interface_ids to ScopedIpV4 I found another issue. Having two interfaces connected to the same network. One ethernet and one WiFi interface. After some services were discovered, I've disabled the WiFi interface. I expected the interface_ids to also be updated at some point, but no new resolved events were observed with the wlan0 interface removed.

[hrzlgnm]

I also checked that for the IPv6 addrs, and those that were scoped to the WiFi interface were also not removed.

Output of avahi-browse:

avahi-browse -pr _adb._tcp
+;wlan0;IPv6;adb-1323823001290;_adb._tcp;local
+;enp12s0;IPv6;adb-1323823001290;_adb._tcp;local
+;enp12s0;IPv4;adb-1323823001290;_adb._tcp;local
=;enp12s0;IPv6;adb-1323823001290;_adb._tcp;local;Android-2.local;192.168.178.22;5555;
=;enp12s0;IPv4;adb-1323823001290;_adb._tcp;local;Android-2.local;192.168.178.22;5555;
+;wlan0;IPv4;adb-1323823001290;_adb._tcp;local
=;wlan0;IPv6;adb-1323823001290;_adb._tcp;local;Android-2.local;192.168.178.22;5555;
=;wlan0;IPv4;adb-1323823001290;_adb._tcp;local;Android-2.local;192.168.178.22;5555;
-;wlan0;IPv4;adb-1323823001290;_adb._tcp;local

vs.

Output of example query:

cargo run --example=query _adb._tcp
    Finished [`dev` profile [unoptimized + debuginfo]](https://doc.rust-lang.org/cargo/reference/profiles.html#default-profiles) target(s) in 0.01s
     Running `target/debug/examples/query _adb._tcp`
At 301.456µs: SearchStarted("_adb._tcp.local. on 3 interfaces [enp12s0 (2), wlan0 (4), lo (1)]")
At 94.909852ms: ServiceFound("_adb._tcp.local.", "adb-1323823001290._adb._tcp.local.")
At 94.94952ms: Resolved a new service: adb-1323823001290._adb._tcp.local.
 host: Android-2.local.
 port: 5555
 Address: fe80::a893:5e1d:3b17:a3c0%enp12s0
 Address: 192.168.178.22 via [InterfaceId { name: "enp12s0", index: 2 }]
At 169.307206ms: Resolved a new service: adb-1323823001290._adb._tcp.local.
 host: Android-2.local.
 port: 5555
 Address: fe80::a893:5e1d:3b17:a3c0%wlan0
 Address: fe80::a893:5e1d:3b17:a3c0%enp12s0
 Address: 192.168.178.22 via [InterfaceId { name: "wlan0", index: 4 }, InterfaceId { name: "enp12s0", index: 2 }]
At 1.00086654s: SearchStarted("_adb._tcp.local. on 3 interfaces [enp12s0 (2), wlan0 (4), lo (1)]")
At 3.002857092s: SearchStarted("_adb._tcp.local. on 3 interfaces [enp12s0 (2), wlan0 (4), lo (1)]")
At 7.004137181s: SearchStarted("_adb._tcp.local. on 3 interfaces [enp12s0 (2), wlan0 (4), lo (1)]")
At 15.004294571s: SearchStarted("_adb._tcp.local. on 3 interfaces [enp12s0 (2), wlan0 (4), lo (1)]")
At 31.004719639s: SearchStarted("_adb._tcp.local. on 2 interfaces [enp12s0 (2), lo (1)]")
At 63.007096377s: SearchStarted("_adb._tcp.local. on 2 interfaces [enp12s0 (2), lo (1)]")
At 127.008848641s: SearchStarted("_adb._tcp.local. on 2 interfaces [enp12s0 (2), lo (1)]")

I guess we also need to remove the DNS records associated with an interface that is removed, in case an interface is removed.

[hrzlgnm]

I've slightly modified the query example so it prints the via interface_ids additionally for debugging.

[keepsimple1]

Thanks for testing it out! I think it's a valid issue. Will work on a fix.

[keepsimple1]

@hrzlgnm I've created a PR #448 to fix this issue. Could you please give it a try? Thanks!

[hrzlgnm]

@hrzlgnm I've created a PR #448 to fix this issue. Could you please give it a try? Thanks!

Tested both ways, enabling and disabling another interface during browsing:

It works in general.

There is an issue with enabling an interface does yield new resolve events with a high delay, roughly at 90s which looks like TTL expiry triggered re-query of those A records, not the interface addition:

At 583.8µs: SearchStarted("_adb._tcp.local. on 7 interfaces [lo (1), br-9188c292b15e (10), veth33f9234 (14), vetha3f1653 (13), vetha265792 (16), veth9caf94e (15), enx00e04c683881 (3)]")
At 89.546411ms: ServiceFound("_adb._tcp.local.", "adb-1323823001290._adb._tcp.local.")
At 89.630721ms: Resolved a new service: adb-1323823001290._adb._tcp.local.
 host: Android-2.local.
 port: 5555
 Address: 192.168.178.22 (IPv4) via [InterfaceId { name: "enx00e04c683881", index: 3 }]
 Address: fe80::a893:5e1d:3b17:a3c0%enx00e04c683881
At 1.000920366s: SearchStarted("_adb._tcp.local. on 7 interfaces [lo (1), br-9188c292b15e (10), veth33f9234 (14), vetha3f1653 (13), vetha265792 (16), veth9caf94e (15), enx00e04c683881 (3)]")
At 3.002060759s: SearchStarted("_adb._tcp.local. on 7 interfaces [lo (1), br-9188c292b15e (10), veth33f9234 (14), vetha3f1653 (13), vetha265792 (16), veth9caf94e (15), enx00e04c683881 (3)]")
At 7.003429412s: SearchStarted("_adb._tcp.local. on 7 interfaces [lo (1), br-9188c292b15e (10), veth33f9234 (14), vetha3f1653 (13), vetha265792 (16), veth9caf94e (15), enx00e04c683881 (3)]")
At 9.016366181s: SearchStarted("_adb._tcp.local. on 8 interfaces [veth33f9234 (14), enx00e04c683881 (3), lo (1), br-9188c292b15e (10), vetha3f1653 (13), wlo1 (4), veth9caf94e (15), vetha265792 (16)]")
At 10.004586306s: SearchStarted("_adb._tcp.local. on 8 interfaces [veth33f9234 (14), enx00e04c683881 (3), lo (1), br-9188c292b15e (10), vetha3f1653 (13), wlo1 (4), veth9caf94e (15), vetha265792 (16)]")
At 11.004330181s: SearchStarted("_adb._tcp.local. on 8 interfaces [veth33f9234 (14), enx00e04c683881 (3), lo (1), br-9188c292b15e (10), vetha3f1653 (13), wlo1 (4), veth9caf94e (15), vetha265792 (16)]")
At 11.005363536s: SearchStarted("_adb._tcp.local. on 8 interfaces [veth33f9234 (14), enx00e04c683881 (3), lo (1), br-9188c292b15e (10), vetha3f1653 (13), wlo1 (4), veth9caf94e (15), vetha265792 (16)]")
At 96.135851641s: Resolved a new service: adb-1323823001290._adb._tcp.local.
 host: Android-2.local.
 port: 5555
 Address: 192.168.178.22 (IPv4) via [InterfaceId { name: "wlo1", index: 4 }, InterfaceId { name: "enx00e04c683881", index: 3 }]
 Address: fe80::a893:5e1d:3b17:a3c0%enx00e04c683881
 Address: fe80::a893:5e1d:3b17:a3c0%wlo1

We probably need to add a re-query for the added interface too.

The interface removal works quite well, if one sets the ip check interval to 1s.

Continuation from the query above:

At 139.006020319s: SearchStarted("_adb._tcp.local. on 8 interfaces [veth33f9234 (14), enx00e04c683881 (3), lo (1), br-9188c292b15e (10), vetha3f1653 (13), wlo1 (4), veth9caf94e (15), vetha265792 (16)]")
At 165.1369085s: Resolved a new service: adb-1323823001290._adb._tcp.local.
 host: Android-2.local.
 port: 5555
 Address: fe80::a893:5e1d:3b17:a3c0%enx00e04c683881
 Address: 192.168.178.22 (IPv4) via [InterfaceId { name: "enx00e04c683881", index: 3 }]

[keepsimple1]

Thanks!

From this log:

At 9.016366181s: SearchStarted("_adb._tcp.local. on 8 interfaces [veth33f9234 (14), enx00e04c683881 (3), lo (1), br-9188c292b15e (10), vetha3f1653 (13), wlo1 (4), veth9caf94e (15), vetha265792 (16)]")

We are re-querying with the added interface wlo1. But for some reason, it didn't re-trigger ServiceResolved. With a quick look I didn't find the root cause, will try to find a testbed that could reproduce. If you could enable debug logs and collect the log for this case, that will be great.