Gracefully shutting down of NetStackImpl

[bdbai]

I notice that in NetStackImpl::new there are a few tokio::spawn calls with a loop inside their async blocks. Some of the spawned tasks seems to run forever such as the timer.

My suggestion is to wrap these tasks with futures::future::abortable and manually abort them while dropping a NetStackImpl. Do you think it is necessary to do so?

[eycorsican]

We need a way to close all apps/connections/inbounds/outbounds gracefully, not only the netstack.

[bdbai]

Simply making the root Future abortable will work I guess.

[bdbai]

Aborting these root Futures makes block_on return. After that, we could manually call shutdown_background on the runtime to stop all spawned tasks.

[eycorsican]

I'm not sure how rt.shutdown_background() works, the doc says:

Note however, that because we do not wait for any blocking tasks to complete, this may result in a resource leak (in that any blocking tasks are still running until they return.

[bdbai]

Do we have tasks that may block indefinitely?