Re: [ECOS] Two threads -two messages - one mbox

[Matthias Gorjup <gorjup at norik dot com>]

On Sunday 20 June 2004 19:32, Bart Veer wrote:
> >>>>> "Matthias" == Matthias Gorjup <gorjup@norik.com> writes:
>
>     Matthias> Hello,
>     Matthias> I am having problems with sending more than on
>     Matthias> mbox-meassage in one loop-iteration from one thread to
>     Matthias> another. Below is the code we are using:
>
>     <snip>
>
>     Matthias> If I only send one message and read one mesage, the code
>     Matthias> works fine. But if I send two messages in one iteration,
>     Matthias> the system reboots after the first iteration of the
>     Matthias> sending thread.
>
>     Matthias> I must say that I need to send two integer values from
>     Matthias> one thread to another during one iteration and the other
>     Matthias> thread needs to read both of them in one iteration as
>     Matthias> well.
>
>     Matthias> What could be wrong?
>     Matthias> Should I use a global structure with two integer
>     Matthias> elements and then send only a pointer to that structure
>     Matthias> (this would make it possible to send only one message in
>     Matthias> one iteration)? Should I use mutex-es or some other
>     Matthias> synchronizaiton mechanism?
>
> I can see two obvious problems with your code.
>
> 1) you are creating and starting the threads before the
>    synchronization data structures. In particular mbox_handle could
>    get used by thread1() or thread2() before main() has created it. In
>    your simple example code main() is likely to run at a higher
>    priority then the other threads so there is not actually a problem,
>    but that is likely to change as the code gets more complicated. It
>    is a good idea to initialize all synchronization data structures
>    before starting any threads.
>
> 2) you are not allocating any space for the mbox itself. Your sock_msb
>    argument to cyg_mbox_create() is an uninitialized pointer so it
>    will have a value of 0. Hence you are placing the mbox at location
>    0, where the interrupt vectors are likely to be depending on the
>    architecture you are using.
>
Thanks!! Now it works :)

Matthias Gorjup

>    I suspect that when you do the second mbox send you are overwriting
>    a critical interrupt vector, and the next time an interrupt
>    triggers the system will crash and burn.
>
>    Instead you need to allocate space for the mbox, something like:
>
>      cyg_handle_t  mbox_handle;
>      cyg_mbox	   mbox_data;
>
>      ...
>
>      int
>      main(int argc, char** argv)
>      {
>          cyg_mbox_create(&mbox_handle, &mbox_data);
>
>          /* now create and start the threads */
>      }
>
> Bart
>
> --
> Bart Veer                       eCos Configuration Architect
> http://www.ecoscentric.com/     The eCos and RedBoot experts

-- 
Matthias Gorjup
Norik Systems
www.norik.com
Tel: + 386 41 540 545
Tel: + 386 3 759 3077
Fax: + 386 3 759 3078


-- 
Before posting, please read the FAQ: http://ecos.sourceware.org/fom/ecos
and search the list archive: http://ecos.sourceware.org/ml/ecos-discuss