Zynq : Manual debugging of an RTEMS demo

[Chris Johns]

On 26/03/2016 2:43 AM, Badr El Hiouel wrote:
>
> I did the 'si' and more break points , and now I can see why it never stops
> .
>
> Breakpoint 7, arm_cp15_start_setup_translation_table (ttb=<optimized out>,
> client_domain=15, config_table=0x104348 <zynq_mmu_config_table>,
>      config_count=14) at
> ../../../../../.././xilinx_zynq_zedboard/lib/include/bsp/arm-cp15-start.h:133
> 133  for (i = 0; i < ARM_MMU_TRANSLATION_TABLE_ENTRY_COUNT; ++i) {
>
> Breakpoint 8, arm_cp15_start_setup_translation_table (ttb=<optimized out>,
> client_domain=15, config_table=0x104348 <zynq_mmu_config_table>,
>      config_count=14) at
> ../../../../../.././xilinx_zynq_zedboard/lib/include/bsp/arm-cp15-start.h:134
> 134    ttb [i] = 0;
>
> I checked the file arm-cp15-start.h and the debugger is stuck in this loop.
> When entering 'si' the gdb never stops on executing lines 133 then 134 then
> 133 again ...

Great work narrowing down the issue to here.

> I don't know why it never stops in this loop.

It might be something in the MMU table set up. If other users are using 
the BSP ok and not reporting issue then I wonder if XDM is interacting 
some how.

Note, running a debugger when the MMU table is set up can be tricky. You 
may need to chat to Xilinx and make sure XDM can handle the cache and 
MMU changes RTEMS does correctly. Any issue with XDM not detecting the 
change in the cache and MMU configuration and so not flushing the caches 
correctly can effect what happens.

Chris