Christoph Paasch

The reported interface-ID comes from ```sk->sk_bound_dev_if```, meaning the subflow did an SO_BINDTODEV before the connect(). Thus, for the initial subflow that entirely depends on whether or not the app used...

> thanks. I am fine with that, just a bit surprised, my reasoning being it's best not to submit the interface id if it's wrong. Yes - probably, when the...

@matttbe - what do you think? I agree with @teto that the interface is not consistent here.

wrt to deceiving interface id - the id you are getting is 0, right? It is confusing indeed, and I think that the nl-pm should not give an if-index when...

When ```so_bindtodev``` is 0, the outgoing interface purely depends on the routing. And, as you say - that can change. So, if the user-space daemon receives the src/dst ip-addresses without...

There is one place in ```mptcp_sub_close``` where we inc the refcnt and then schedule the work-queue. That could fail though, so ideally we should decrement again. However, that does not...

Thanks for the logs! It's yet unclear to me how this happens. I'm trying to trigger it with syzkaller. Let's see...

Hello! Thanks for this report! I think your patch makes sense. Just one thing: I believe you should rather check against ```meta_tp->window_clamp``` than the subflow's window-clamp. Because the subflow's window-clamp...

The question rather is: Why would ```meta_tp->window_clamp``` be so small that ```2 * rcv_window_now``` is bigger than it ? What values of ```window_clamp``` are you seeing?

window-clamp on meta-tp is 600 to 1100 ??? Wow... That's weird... Which branch exactly are you running on?