tcp: stop all timers in tcp_usr_detach()

[rscheff]

Summary

When apruptly detaching a tcp session, stop all timers prior of discarding the tcpcb.

Reported-by: [e-mail]
Reported-by: [e-mail]

[tuexen]

Please also add:

Reported-by: [e-mail]

[rscheff]

• split the MPASS checks

[rscheff]

• stop timers instead

[glebius]

This should fix the panic observed of course. Question is: do we want every TCP state go through tcp_close() in its life cycle? Cause no we got a case that TIME_WAIT will just disappear. There are three options here: 1) use the patch as is 2) add tcp_state_change() here 3) add full tcp_close().

We may want to quickly check in the patch as is and then think better.

[rscheff]

Better revert to the prior state (with tcp_timer_stop() in tcp_discardcb()) - D43855