AM2432: Add hardware watchpoint in CCS 12.6 fails for address 0x40

Hi Johannes,

Which version of SDK are you using?

Please tell us which example are you loading.

Regards,

Tushar

Hi Tushar,

The problem appears with our rather big C++ project utilizing both R5F cores of the AM2432 on our custom board. We use motor_control_sdk_am243x_09_02_00_09. I was not able to reproduce the issue with uart_echo_low_latency_polling_am243x-evm_r5fss0-0_nortos_ti-arm-clang example project. I haven't tried other examples.

A colleague of mine was able to reproduce the issue with our project and our custom board on his PC.

Any ideas what I could try to trace things down?

Best regards

Johannes

Hi Johannes,

Can you please check the MPU settings in the project for this address? Is this address 0x40 accessible by the core?

Regards,

Tushar

Hi Tushar,

As far as I can see, the settings are the same as the ones for TCMA in the example projects:

I can access the location in memory browser (read/write).

Best regards

Johannes

Hi Johannes,

Are you not able to put breakpoint at 0x40 from any r5f core? Or are you facing this issue with core-1 only?

Is this error happens every time or does it happens randomly?

Also which example are you trying to load? Please provide steps to reproduce the issue.

Regards,

Tushar

Hi Tushar,

I observe the problem reliably on core 2.

I haven't tried core 1 yet.

As I wrote above, I am unable to reproduce the problem with an example project but it occurs always when I load our software into our custom board. I usually load both cores with a script like this:
print("Open debug session")
dsCPU["r5fss0-0"] = debugServer.openSession( ".*MAIN_Cortex_R5_0_0" );
dsCPU["r5fss1-0"] = debugServer.openSession( ".*MAIN_Cortex_R5_1_0" );

print("Issue system reset");
dsCPU["r5fss0-0"].target.getResetType(1).issueReset();

disconnectCpu("r5fss0-0")
disconnectCpu("r5fss1-0")
connectHaltResetCpu("r5fss0-0")
connectHaltResetCpu("r5fss1-0")

loadCpu("r5fss0-0", full_path_to_core_1_outfile);
loadCpu("r5fss1-0", full_path_to_core_2_outfile);

Hi Johannes,

For me it looks like issue with the memory configuration settings done in the application for core2. Please compare the memory configuration settings of core2 with core1 and check are there any differences in them?

Regards,

Tushar

I have investigated the problem a bit more.

I can load our dual core project into the AM2434 based EVM (using the same script that I ususally use) and I am able to place a watchpoint at 0x40 on r5fss1-0.

The exact same procedure reproducibly does not work on our AM2432 based custom boad. When I try to set the watchpoint after loading the out files I get the error message that I have attached to my initial post.

I get the same error message when I try to add the watchpoint after loading one of the example projects (e.g., uart_echo_am243x-evm_r5fss0-0_freertos_ti-arm-clang) into r5fss1-0 of our custom board.

Best regards

Johannes

Hi Johannes,

Please share the example.syscfg file for analysis.

Regards,

Tushar

example - Kopie.zip

Hi Johannes,

I have checked the syscfg file and the MPU setting looks okay.

Meanwhile, I have checked the screenshot of error, the error message is coming for both cores. R5F0-0 and R5F1-0.

Can you please confirm, have you added watchpoint of each core separately? Two watchpoints for two cores.

Also how are you setting this watchpoints?

Regards,

Tushar