fix multithreading bug about default_io in iiod-reader

[catkira]

PR Description

fixes #1211

PR Type

• Bug fix (a change that fixes an issue)

PR Checklist

• I have conducted a self-review of my own code changes
• I have commented new code, particulary complex or unclear areas
• I have checked that I did not intoduced new warnings or errors (CI output)
• I have checked that components that use libiio did not get broken

[cristina-suteu]

Hello. Can you please do a rebase/ clean up of the commits? It is a little unclear what's going on

[catkira]

There is no definitve fix yet. You can follow our discussion in the issue.

[nunojsa]

There is no definitve fix yet. You can follow our discussion in the issue.

I think the fix will be in line with serializing accesses to the default IO. I did implemented a possible fix but then I was pulled to some other stuff. I just pushed it here:

https://github.com/analogdevicesinc/libiio/tree/staging/multiple-default-io-users

If you time to look at it and give it a test... But this is totally untested code and it's likely to fail so totally fair if you don't want to try it 😅 . It's based on Paul's approach but it goes a step beyond with respect to refcounting. I think we have a fundamental issue in the client side with iiod_responder_get_default_io(). That gives us a pointer to a refcounted object but we do not increase the count. Adding that we have async waiting, we're opening a door for use after free (theoretical but I think possible). I'm also using the refcount value to detect whether or not the IO has active waiters.