wpa_supplicant gets restarted when changing configs

[bmurphy23]

Each VintageNetWiFi config has the WPASupplicant module listed as a child spec, meaning the underlying wpa_supplicant process gets started and killed as a part of that config's life cycle. However, killing wpa_supplicant has some implications, notably removing any interfaces it cares about. In the case of reconfiguring VintageNetWiFi, this is a superfluous step that potentially causes problems due to the interface being removed at the kernel level for a moment. I think there may also be a race in the Interface state machine that can accidentally trigger a 30-second retry due to wpa_supplicant getting killed when reconfiguring.

I'm thinking that WPASupplicant could be managed at the Interface level as some kind of lifetime-supervised process, instead of at the per-config level. The config's up_cmds could include issuing a RECONFIGURE command over the command interface to make wpa_supplicant pick up and apply the new config. For systems that update their config multiple times at runtime, this ensures that wpa_supplicant is a long-lived process that's not directly tied to a specific configuration.

[fhunleth]

This is by design and important to guarantee that WiFi is set up deterministically.

What attribute are you trying to change?