[Tinyos-devel] RE: 2.0.1- T2 Radio Stack

[David Moss]

Hi Geoff -

There were some significant updates to the CC2420 radio stack from 2.0.1 to
2.0.2, and a good chunk of it was related to debugging and reliability.  I
don't recall seeing a bug like you're describing, except on earlier versions
of the CC2420 stack (not sure if "earlier" means 2.0.1, or before). There
was certainly a problem at one point where the MSP430 SPI bus would lock up
the node if anything tried to do an operation with a 0-byte length.  This
has been fixed in recent versions of TinyOS.  And the 2.0.2 CC2420 radio
stack is much more stable than the 2.0.1 radio stack because it's been unit
tested.  The unit testing alone fixed a large number of bugs.  Even after
2.0.2, the current CVS repository contains a few minor bug fixes to the
stack.  

I'm CC'ing -devel on this to see if anyone else has experienced lock-up
issues as you're describing.  At a minimum, I would recommend dropping in
the latest CC2420 radio stack to see if that makes a difference. The
interfaces provided by the stack should be the same, so dropping in a new
stack will be a matter of editing your .platform file to pull in the new
directories.

-David






-----Original Message-----
From: Geoffrey Werner-Allen [mailto:werner at eecs.harvard.edu] 
Sent: Wednesday, March 12, 2008 10:55 AM
To: David Moss
Subject: 2.0.1- T2 Radio Stack

Hey David:

Not sure if you're the right target for this, but it seems like you've been 
doing the active development in the T2 CC2420 radio stack so here goes.
Feel 
free to forward as appropriate.

I'm seeing a lock up on TMote Sky nodes that seems related to radio 
communication.  My setup is quite simple: I have two nodes, on acting as the

root of a CTP routing tree (and forwarding packets to a dedicated node
running 
BaseStation) and another as a leaf.  Initiating a large data transfer from
the 
lead node through the root hangs the root node almost without fail.  This is
a 
very hard crash in that a) the node seems completely hung and b) it doesn't
even 
get to the point where a watchdog timer can reboot the system (we have one 
active).  At various points I thought that this was a function of
interleaving 
Flash I/O or medium-frequency sampling (100Hz) with the radio communication,
but 
I can get it to happen between two nodes that are (essentially) doing little

else: running the TimeSync protocol and participating in Dissemination, but 
that's it.  The bug is highly-repeatable.

I can tickle this on T2.0.1 and T2.0.2.  Haven't tried the bleeding-edge CVS

yet, but I can if it's helpful.  I'm not going to be a terribly large amount
of 
help debugging this before the Sensys deadline, but I'm concerned about my 
ability to utilize some of the LPL features in the 2.0.1-on stacks, so I
thought 
I'd ask if this is something you've seen that fixed in checkins downstream
from 
2.0.2.

Hope all's well!

Best,

-gwa-
-- 
geoffrey werner-allen :: 617.694.7261 :: www.eecs.harvard.edu/~werner