[Tinyos-devel] Bug in Resource.request interaction with ResourceRequested.requested

[Kevin Klues]

I agree this is in fact a bug.  I think the correct solution would be
to add one line inside of resource.release() for resId=NO_RES while in
the process of granting the resource.  Can you try this out and let me
know if it fixes the problem.  I'll check it in if you can verify that
it solves it.

inside Resource.release()

        if(call Queue.isEmpty() == FALSE) {
          reqResId = call Queue.dequeue();
          resId = NO_RES;
          state = RES_GRANTING;
          post grantedTask();
          call ResourceConfigure.unconfigure[id]();
        }

Kevin

On Mon, Aug 11, 2008 at 5:21 PM, Eric Decker <cire831 at gmail.com> wrote:
> I am making heavy use of arbitration in a T2 based mote.  I've discovered
> what looks
> like incorrect behaviour.  The code looks like:
> =>async command error_t Resource.request[uint8_t id]() {
>     signal ResourceRequested.requested[resId]();
>     atomic {
>       if(state == RES_CONTROLLED) {
>         state = RES_GRANTING;
>         reqResId = id;
>       }
>       else return call Queue.enqueue(id);
>     }
>     signal ResourceDefaultOwner.requested();
>     return SUCCESS;
>   }
>
> The problem is ResourceRequested.requested[resId]() gets signaled
> no matter what.  "resId" holds the value of the last holder of the resource
> it doesn't get modified when a resource owner does a release.
> I would assert that a ResourceRequested.requested shouldn't be signaled
> unless the resource is actively owned by resId.  ie.  signal only if state
> == RES_BUSY.
> So if a resource owner does a release it is possible (Resource.request is
> async) for a
> request to come in.  If ResourceRequested.requested is signaled, I assert
> that
> this is wrong.
> This can be protected inside of the ResourceRequested.requested
> event by checking for ownership but I think it would be cleaner to check for
> RES_BUSY
> inside of Resource.request.  I would assert ResourceRequested.requested
> should
> only be signaled when the resource is owned and a new request comes in.  It
> only should
> be signaled to the current owner.
> I observed the failure when executing the following code:
>         call UARTResource.release();
>         call GPSMsg.reset();
>         call GPSTimer.startOneShot(DT_GPS_MAX_REQUEST_TO);
>         call UARTResource.request();
> Even though we do the release,  when the request is executed, I observe that
> UARTResourceRequested.requested is getting signalled which seems odd.
> thoughts,
> eric
>
> --
> Eric B. Decker
> Senior (over 50 :-) Researcher
> Autonomous Systems Lab
> Jack Baskin School of Engineering
> UCSC
>
>



-- 
~Kevin