Re: [Tipc-discussion] Re: hang while deleting ports

[Jon M. <jon...@er...>]

Well, we do have bottom halves (timer and signal tasklets) that
may interfere with message sending/reception if executed, 
even on UPs. I am sure there are cases  when the _bh() version 
is not really needed, but I wanted to keep it safe when I  developed it.

Something we can look into later, maybe.

/jon

Mark Haverkamp wrote:


On Wed, 2004-05-12 at 07:15, Jon Maloy wrote:

  

Mark,

I think you can easily verify your hypothesis by replacing 

spin_lock_bh() with a spin_trylock_bh() in tipc_recv_msg(), and

then just discard the packet and continue the loop if it fails.

(If this happens the packet will be retransmitted anyway.)

It is worth a try.

    



I'll give it a try.



Out of curiosity, why are you using the *_bh versions of spinlock?