[net2-wg] Collection

[Kyle Jamieson]

With Rodrigo's help, I've instrumented the FE with calls to send 
diagnostic packets to the uart in case of exceptional circumstances. 
Those could be:

- message pool empty
- send queue full
- no route from the router
- subsend is off
- subsend is busy
- senddone from subsend doesn't make sense
- queue entry pool empty

Now we can see more of what's going on in the FE.

Philip Levis wrote:
> I see four major places that might be the cause of the problem. There 
> are two simple ways to start the diagnosis. The first is packet 
> sniffing: use a base station to see who is sending what (if at all). The 
> second is to use the LEDs in three specific circumstances.
> 
> 1) Queueing/buffer management: ForwardingEngineP is corrupting packets, 
> dropping packets, or otherwise preventing traffic from getting through.
>   - Instrument ForwardingEngineP to turn on Led0 when its message pool 
> is empty or its send queue is full and send an error message to the UART.
> 
> 2) Route selection/link estimation: The link estimation table is 
> oscillating wildly, leading to an unstable tree or there appear to be no 
> routes.
>   - Instrument TreeRoutingEngineP to turn on led1 when it has no route 
> and send a periodic message containing the parent ID to the UART. Send a 
> UART message whenever the parent changes.
> 
> 3) Loops in the routing tree are causing the network to saturate with 
> packets, false negatives acks are causing the number of packets in the 
> tree to increase greatly.
>   - Instrument ForwardingEngineP to count the number of retransmissions 
> it makes; if it goes beyond a small number (e.g., 4), send an error 
> message to the UART
> 
> Does anyone have additional hypotheses/tests?
> 
> Phil
> _______________________________________________
> net2-wg mailing list
> net2-wg at millennium.berkeley.edu
> https://mail.millennium.berkeley.edu/cgi-bin/mailman/listinfo/net2-wg