RF2xx: Resource.requested called from SoftwareInit. Needs to be fixed.

[tinyos-issues]

Original author: cire...@gmail.com (April 06, 2012 06:14:36)

The following files call Resource.requested from SoftwareInit which is broken.

The Arbiter being used is Fcfs and requires the queue to be initialized. This occurs via a wiring in SoftwareInit. These wirings are unordered and thus the queue may or may not be initialized when Resource.requested is called from SoftwareInit.

These should probably be replaced by calls to Resource.immediateRequest. This will change the execution path (previous event driven, changed to non-event). So careful evaluation should be done.

However, the current implementation is incorrect and an error waiting to happen.

tos/chips/rf230/RF230DriverHwAckP.nc
tos/chips/rf230/RF230DriverLayerP.nc
tos/chips/rf212/RF212DriverLayerP.nc

Original issue: http://code.google.com/p/tinyos-main/issues/detail?id=131

[tinyos-issues]

From thomas.s...@gmail.com on June 27, 2012 20:39:52
cc2420x and cc2520 drivers do the same thing as they all come from the same ancestry.

[tinyos-issues]

From jmatts...@dius.com.au on July 02, 2012 01:14:16
The other option is to not use the pre-canned Fcfs in the driver, but wire one up "by hand" in a way that guarantees it to be initialized before use.

I'd consider using immediateRequest() to still be incorrect, as it's called on an uninitialized component. The code path turns out to be safe(?), but it's not the proper thing to do. Just my $.02 :)

[tinyos-issues]

From cire...@gmail.com on July 02, 2012 03:15:26
I agree that immediateRequest is a kludge (an thus somewhere on the incorrect side of things). It was suggested only as an immediate temporary workaround while a more correct fix is done and brought in to the code base.

I am surprised that this is taking so long to fix. The current code is clearly broken.

If it effected me directly or if it effects someone else who is using the TinyProd code base, I'd fix it there. But I'm not currently using any of the radio stacks effected. Sooo.... I am purposedly minimizing how much divergence between the main TinyOS development tree and TinyProd because then it is easier to track.

I'm hoping this will get fixed after 2.1.2 is released. Although, I would think this is something that should be fixed for 2.1.2.