On Wed, Nov 28, 2012 at 12:19 AM, David Rodenas <[email protected]> wrote:
> Hi Eric > > Up to now, it doesn't work. In fact, when it is supposed the piece of code > executes, the mote resets, and I cannot understand it! I'm also debugging > the code but I'm missing something because I don't see anything strange. > If it was me doing this, I'd put a debugger on it. I also write code with explicit Panics and the code checks itself. I know that doesn't help your current situation. But standing on the outside of a black box trying to figure out why the beasty doesn't work doesn't in my experience work. You have to open the box and find out what is going on. > Lets see if I understood your doubts. In relation to who keeps track of > what buffers are available or not, I thing that functionality is part of > the QueueC component and yours. > This is too hard to do via words. The language barrier doesn't help but you are doing very well all things considered. > Yours because you are who mark when a packet has to be dequeued aimed at > being transmitted or discarded, > That sounds right. > and of the QueueC because it points to a "free" space of memory where you > can enqueue a new packet as well as the head of this queue. > I don't understand that. Not sure what you are trying to say. > On the other hand, when you need to enqueue a new packet, you have to > return a empty message_t to the MAC by making a call to Pool.get(). Then, > you enqueue the packet by calling Queue.enqueue(new_message_t) . > This is part of the confusion doing this with words. What MAC? When is the message/packet being returned to the MAC? Are we receiving or transmitting? Who are you enqueueing the packet to via Queue.enqueue. The connections between the different implementations (like Queue) is what the wiring specification defines. Both components, PoolC and QueueC, point to different spaces of memory but > they keep track of if you have space to store more packets, the piece of > memory where you enqueue/return a message_t... . And when you dequeue a > packet, after the event sendDone(msg) is signaled, you make a call to > Pool.put(msg) (I'm sure you're aware of this). This make sense for me, > but as I said, for now, it doesn't work! > I think you are figuring it out. But without getting in and seeing exactly what it is doing all you can do is say "it doesn't work". And I'm not going to be able to use my majik crystal ball to look inside your remote mote and see any better. Sorry it that sounds snide. Not meant to be. eric > > Regards > > David > > ------------------------------ > Date: Tue, 27 Nov 2012 16:31:15 -0800 > Subject: Re: [Tinyos-help] Drop packets in TinyOS > From: [email protected] > To: [email protected] > CC: [email protected] > > > > On Tue, Nov 27, 2012 at 12:16 AM, David Rodenas <[email protected]>wrote: > > Hi Eric > > In any case, I have a possible answer which is simpler than I thought. I > was addressing the problem wrong, thinking that a message that was > enqueued, it was really a peace of dynamic/external memory that came (was > transmitted) to the receiver node, so it increased the already occupied > memory (there is a reason of why a thought this: I also work with Network > Simulator 2 - NS2 - and there you use dynamic memory for messages). This is > totally wrong with TinyOS, when you receive a data message and enqueue it, > you are only copying the whole content to the corresponding memory location > (you defined Queue as static memory) and increasing a counter of enqueued > messages. > Then you can use PoolC to return an empty message_t to the system. Thus, > when you make a call to Queue.dequeue(), what this function returns is a > pointer to a static portion of memory (of size message_t) as well as > decreases the counter of messages enqueued. Therefore, when another > message_t is enqueued, this occupies the same portion of memory so the node > never run outs of memory because of what I what asking. > > Am I right? > > > The gist is certainly correct. The devil as always is in the details. > > I do hand debugging of code like this to make sure that a) the general > case works, b) is argueably correct, and then c) examine the corner cases > (where can it screw up, and are there assumptions made about how the code > works that can be violated). But I didn't write or test this code. > > so your mileage may vary (YMMV). > > But I do think you have the gist of how it behaves. > > The question is who keeps track of what buffers are available and what > denotes the buffer being busy. If you dequeue the packet from the > "queue", does it something else need to be told that this packet is > available? Should it get put back into the Pool? > > eric > > > Thanks again, regards > > David > > ------------------------------ > Date: Mon, 26 Nov 2012 19:51:48 -0800 > Subject: Re: [Tinyos-help] Drop packets in TinyOS > From: [email protected] > To: [email protected] > CC: [email protected] > > > Looking at this problem from the queueing mechanism is coming at the > problem from the inside out. > > So it is pretty difficult trying to answer the question you are asking. > > To do so really requires looking at the data flow (message_t) and buffer > flow at various points in packet transmission and reception. > > The answer to your question depends on what the original caller of the > enqueue wants done. And then there is the whole wiring issue and how to > pull the whole thing together. > > Simply dequeuing the message certainly isn't enough because that most > likely leads to lost messages. But the whole scheme needs to be worked out > in a systemic fashion. And it would be nice if it were documented. It > certianly is a hole. Eventually I'll get to it. > > I certainly understand what you are trying to accomplish. Been there done > that. Sorry I don't have a better answer for you. > > eric > > > On Mon, Nov 26, 2012 at 7:48 AM, David Rodenas <[email protected]>wrote: > > Hi all > > > I would like to know how to discard a message which is stored in a queue > (component QueueC of message_t). To better explain this, lets suppose we > have a maximum number of retries during we are trying to forward a data > message. This is stored in a queue waiting for being transmitted. A failed > transmission can occur due to collisions, interferences, etc. So when this > maximum number is reached, the data message should be discarded, making > this same procedure with the following message (head of the queue). Here > there is a simple code of what I'm asking: > > maximum_retries = 3; > num_retries = 0; > > if (num_retries > maximum_retries) > { > message_t * drop_msg = call Queue.dequeue(); > drop(drop_msg); > } > > The point is that I'm not sure if just dequeuing the message_t is enough. > This message occupies a piece of memory space and this has to be freed. Am > I right? Should I use the function free as in any C-based language when > dynamic memory allocation is used? > > Thanks > > David > > _______________________________________________ > Tinyos-help mailing list > [email protected] > https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help > > > > > -- > Eric B. Decker > Senior (over 50 :-) Researcher > > > > _______________________________________________ > Tinyos-help mailing list > [email protected] > https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help > > > > > -- > Eric B. Decker > Senior (over 50 :-) Researcher > > > -- Eric B. Decker Senior (over 50 :-) Researcher
_______________________________________________ Tinyos-help mailing list [email protected] https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
