Jon Ward wrote:
-------------------------------
There are numerous other ways you could handle this. For example, you could create a FIFO and use a counting semaphore. But, I don't know enough about your requirements (like how many tasks call your protocol task, how fast it has to be, and so on).

First of all thanks a lot Jon for your help.
For clearing my case:
1. Three tasks are calling the send task simultaneously, and the data that each task sends to this task need to be unique.
2. The send task supposes to send message (compose from the data this task received) to other micro on the board. So, that task supposes to be pretty fast.

Is the memory pool with the struct solution is the best option here? I think that using memory management functions in tasks that the entire time active is not so good idea (get pool and free pool every few micro seconds). Am I right?

Thanks,
Roberto