Re: [RFC PATCH 1/2] RDMA/rxe: drive UDP tunnel socket lifetime from the GID table

> I'm skeptical this AI conclusion is right? Or at least a leaking sock
> crashing the kernel sounds like a netdev bug, not something to avoid here.

rxe_ns_exit is unnecessary.  

I put pr_info in rxe_ns_exit(): in 40 iterations of a netns whose
netdev is moved away and which is then deleted, rxe_ns_exit() was
called 82 times and found no entry every time.

That is ordering, not luck. With a 3s msleep injected into rxe_del_gid()
to force the window open (timestamps recorded with kprobes):

  rxe_del_gid_in             65.844231
  rxe_del_gid_out            68.677344
  rxe_del_gid_in             68.679284
  ib_cache_cleanup_one_in    69.011549
  rxe_del_gid_out            71.535930
  gid_table_cleanup_one_in   71.536279
  gid_table_cleanup_one_out  71.536313
  ib_cache_cleanup_one_out   71.536321
  rxe_ns_exit_in             71.582488 


So rxe_ns_exit is dead code. I will remove it in v2.

thanks,
serhat