No privileged mode in main()

Martin,

          The mode (user/prev) is set by the startup code which runs before you get the control to main().

          In StarterWare this is configured in init.s file (present in system config lib). But while creating ccs project seems the this file is not getting included and ccs takes its default configuration which might set the mode to user mode.

          I guess you have not linked system_config lib from StarterWare- can you check and confirm this ?

The following links might help you,

http://processors.wiki.ti.com/index.php/StarterWare_Project_Creation

http://processors.wiki.ti.com/index.php/AM335X_StarterWare_Environment_Setup#TMS470_from_CCS_project

Regards

Baskaran

Hi Baskaran,

thanks for your response.

Yes, it is linked.
The following path had been added to the linkers search path,
    "C:\Pr\TI\AM335X_StarterWare_02_00_01_01\binary\armv7a\cgt_ccs\am335x\system_config\Debug"
Besides that
    "${SW_WURZEL}\binary\armv7a\cgt_ccs\am335x\system_config\Debug\system.lib"
had been added to the "Include library file or command file as input" window.

In the system project (where the library is created) is a file 'init.asm' (not init.s).
It contains code to switch to privileged mode.

Trying to set a breakpoint at 'Entry' (in the final project) is refused as 'Entry' cannot be resolved. According to the .map file,  the .txt section  does not seem to contain code from init.asm.

Meanwhile I found an interesting thread at
    http://e2e.ti.com/support/arm/sitara_arm/f/791/p/207053/735133.aspx#735133.
But I can't see how it can help me.

Regards,
Martin

Martin,

           Thanks for confirming that you are linking system_config lib. Can you let me know which linker script file (.cmd) are you using ?

           You need to use the one that comes with StarterWare package  \\build\armv7a\cgt_ccs\am335x\<board>\<example app>\

           This ensures the Entry symbol is linked.

Regards

Baskaran

Hi Baskaran,

I had been using
C:\Pr\TI\CCS\ccsv5\ccs_base\arm\include\AM3359.cmd

Now that I am using gpioLEDBlink.cmd, A8 is in privileged mode.
Thank you very much for your fast response and the precise support information. That solved the problem.

One more remark:
Switching from privileged to user mode is ok.
But using CPUSwitchToPrivilegedMode() to switch back to privileged mode is not (I watched CPSR in the registers window) . However, the function at least
returns.

Regards,
Martin

Martin,

           Regarding CPUSwitchToPrivilegedMode(). Actually the code for this API is correct and it is working correctly.

           The debugger typically relies on svc_handler for its operations. And we also rely on this handler for switching back to privilege mode. So during debug process the debugger takes control and the expected behavior is not seen. you can confirm this by 2 methods,

• Run the code standalone (with out debugger connected)
• Before executing CPUSwitchToPrivilegedMode() from CCS, do free run (menu -> run -> free run). In this case CCS wont use svc_handler and you can see the expected mode change. I confirmed this by free run for below code. and when the control hits while(1), the system is in svc mode.

 CPUSwitchToUserMode();
 CPUSwitchToPrivilegedMode();
 while(1);

regards

baskaran

Hi Baskaran,

how would I proof that the CPU really is in priviledged mode when it reaches the while(1) loop?
Now with the new command file, CPUSwitchToPrivilegedMode() also returns in debug mode.

Thanks!

Regards,

Martin

The API will definitely return but the mode will be changed to privilege mode. To confirm you need to check the mode in cpsr register.

Hi Baskaran,

sorry, my question has not been precise.

I am aware that CPSR has to be checked. But how is that register accessed from C/C++?
Is it mapped to memory? I did not find any hint in the manual.

Thank you!

Martin

it is not memory mapped. You can use the below code to read the cpsr reg.

unsigned int CPUReadCpsr(void)
{
    asm("    mrs     r0, CPSR");
}

the return value from this function will have the contents of cpsr.

Regards

Baskaran

The function works fine, but the compiler complains about a missing return statement.

Is there a chance for suppressing this warning for just that one function?

Thank you!

Regards,

Martin