Experimental Physics and
Industrial Control System

At Diamond we run EPICS on a number of embedded processors (both ARM and PowerPC) running our own busybox based build of Linux.  I’ve developed a makefile based framework for assembling a customised rootfs, I’ve published my work on this here: http://controls.diamond.ac.uk/downloads/other/rootfs .

For ARM we’ve been able to stay very close to the standard kernels at http://kernel.org (our two targets are XCEP and PXA270 Colibri), but for PowerPC (both FPGA embedded and MVME5500) we’re currently stuck with some very old and large patch sets on old kernels (2.6.20 for MVME, 2.6.26 for Xilinx FPGA embedded PPC core) and I am hoping to remedy this.  If anybody else has experience with these targets, please let me know!

Regarding the caRepeater problem for a single standalone IOC I can offer a different solution that was recommended to me ages ago by Andrew Johnson (Dec 2006 it seems), which is simply to start a thread running caRepeaterThread().  For some reason this function is not published in any EPICS header file, so something like the following code will serve:

extern void caRepeaterThread(void *);

int main(int argc, char **argv)

{

    ...

    epicsThreadCreate(

        “CAC-repeater”, epicsThreadPriorityLow,

        epicsThreadStackSize(epicsThreadStackMedium),

        caRepeaterThread, 0);

    ...

}

I create the thread pretty early during IOC startup, seems to work quite well.

ifeq ($(T_A),linuxRT-x86)

 PROD_DEFAULT += caRepeater catime acctst caConnTest casw caEventRate

endif

Then, I did a build like so,

make CROSS_COMPILER_TARGET_ARCHS=linuxRT-x86

Thanks in advance for your comments and help. 

Cheers,

Ernest