Synchronous Message Passing

Paolo Mantegazza mantegazza at aero.polimi.it
Fri Oct 8 18:14:37 CEST 2004


Shravan Rayanchu wrote:
> Hi,
> 
> This is what the setup is: 
> 
> I am writing a new system call, say connect. That is,
> the user would make a call to connect. Now connect
> would run in the kernel space.
> 
> inside connect {
> 
> I form a REQUEST and I want to send it to an RT Task.
> 
> Then wait for the corresponding CONFIRM to come.
> 
> then return the result to user ( result is given to me
> by the RT Task in the form of a CONFIRM structure).
> 
> Problem :
> 1. What should be data structure to be used to pass
> the information
> 
> 2. What should be the WAIT mechanism?
> 
> There many be many connects issued. How should I
> handle the corresponding confirms? Can I solve this
> with a FIFO queue. Synchronous message passing
> facility (RPC) is provided but it requires that both
> the ends are RT Tasks. But for me : on one side there
> is an RT Task and on the other side the Linux kernel
> (system call)
> 

I do not understand the problem if you are writing a system call it 
means that it has to be used by tasks anyhow. Moreove you have an RTAI 
task in kernel space also. So whay not use RTAI rpc as it is, with the 
adavantage of having it in distributed mode also (NETRPC).

It is not needed to be in real time to use RTAI in user space.

Once more your solution has the name: LXRT (IMHO).

Paolo.






More information about the Rtai mailing list