FlashAPI works on RAM-Version, but fails when running from Flash

Hmm, no one can help?

Was something wrong in my post, or is it the wrong forum?

Please comment,

best regards,

Stefan

Hmm, no one can help?

Was something wrong in my post, or is it the wrong forum?

Please comment,

 

Your post seems to have got missed by the team. Now that you've re-highlighted it, I hope the respective expert helps you out. I guess the TMS320F2806x Piccolo Flash API doc didn't help you?

Can be found here: C:\ti\controlSUITE\libs\utilities\flash_api\2806x\v100\doc

Regards,

Gautam

It helped me a lot when I started, and I thought that I included all necessary steps according to the do´s and dont´s List. But as I said, everything worked at once in RAM-version, but fails when running out of FLASH.

regards,

Stefan

Hello,

still no one available?

In the meantime I worked on other open issues, but this point is still important and very essential for me, so please help...

best regards,

Stefan

Hey Stefan,

sorry it took a while to get to this in my list.

Now you have loaded an application to flash, you main is in flash - it then relocates some of the functions to RAM and one of these functions in RAM calls into erase a flash sector.

1. Is the problem with any sector or when you try to erase few sectors it doesn't work?

2. We need to have a function call stack and see where these functions are located (identify if any of them are located in the flash sector you are trying to erase currently)

3. does the flash API in ROM calls back into any of the flash functions in application? where these call back functions are located - are they erased? - this can be checked by looking at you application MAP file, look for any functions with fl28x or other flash api identifiable string in the function name.

4. Did your application reserve the RAM needed by BootROM, hope you are not loading any thing into the RAM reserved for ROM. - I guess not since your RAM version works just fine 

I'll check with the flash API experts if there is a call back from the flash API in ROM to any of the user provided functions while doing erase operations.

Last time I helped one of the customer with exactly similar thing as yours, but that was on Concerto series, the problem turned out to be 2 and 3. Also he was able to erase few sectors but wasn't able to erase few others so we were able to localize the issue.

Best Regards

Santosh

Hi Santosh,

1. any sector!  When I examined today it didn´t even come to erase a single sector. As soon as the function Flash_Erase is called it now crashes and reboots.

2. According to the Linker-Cmd file the stack is located here:  .stack   : > RAMM0, PAGE = 1

which is confirmed in the map-file:

.stack     1    00000050    000003b0     UNINITIALIZED

                  00000050    000003b0     --HOLE--

3. Not sure if I fully understand what you mean here, this is an excerpt from the .map-file:

003ffebb   _Flash2806x_Program

003ffebd   _Flash2806x_Erase

003ffedf   _Flash2806x_Verify

003ffee7   _Flash2806x_ToggleTest

003ffef3   _Flash2806x_APIVersionHex

I think the addresses are in the range of BootROM, which should be correct...?

The Callback functions are put somewhere in RAML7, where my 'ramfuncs' reside:

000100b1 _MyCallbackFunction

00010411   _MyFlashAPI_callbackFunction

4. I think this line in the Linker-cmd file does what you mean:

PAGE 1: ....

BOOT_RSVD   : origin = 0x000000, length = 0x000050

Hope you have an idea....

best regards,

Stefan

Stefan Krammes said:

As soon as the function Flash_Erase is called it now crashes and reboots.

Can you make sure that there is no interrupt occurs when it enters "erase flash" APIs?

Best regards,

Maria

I tried to realise your suggestion by putting the DINT macro one line earlier, so it now looks like

DINT;

Status = Flash_Erase( SECTOR_B, &FlashStatus );

But the result is the same...

I tried to use another API-function like

if( Flash_APIVersionHex() != 0x0100 )
  asm("    ESTOP0");         // Unexpected API version

and this does not crash, so I think that even in my buggy FLASH-Version the BootROM is at least accessible and libs are included...

best regards

Stefan

For more details I attached the Map-files from both the RAM-Version and the FLASH-Version.

Hope this helps you to identify differences.

6835.Mapfiles.7z

best regards,

Stefan

Stefan,

thanks for the MAP files. Before we get into the debug, are you able to run the Example_2806xFlashProgramming project as it is on your set up (the example is under C:\ti\controlSUITE\device_support\f2806x\v136\F2806x_examples_ccsv5\flash_programming)?

the example erases all the sectors except sector A.

Best Regards

Santosh

Before I started integrating API routines into my own project I ran the example named Example_Flash2806x_API.c and it worked well, as usual.

best regards,

Stefan

Stefan,

since the program goes into weeds on calling the flash API in ROM, either the stack is getting overwritten or any of the variables needed by flash API ( FlashCallbackVar, FlashScalingVar) in ROM are corrupted or are not initialized or one of the call back function pointers are getting corrupted.

Before stepping into the ERASE function can you check if above are still good in RAM.

Also could you compare the MAP files and linker command files from the TI example and the map file from your project and make sure what needs to be in RAM is intact?

what is your time zone, if possible getting on a webex and call might be helpful, but lets try to debug it a bit more.

Best Regards

Santosh

Santosh,

to initialise the Flash-Variables I use the Lines

EALLOW;

Flash_CPUScaleFactor = SCALE_FACTOR;

Flash_CallbackPtr = NULL;

EDIS;

The TI-example also uses some pragma`s like

#pragma DATA_SECTION(Flash_CPUScaleFactor, "FlashScalingVar");

 Uint32 Flash_CPUScaleFactor;

#pragma DATA_SECTION(Flash_CallbackPtr, "FlashCallbackVar");

void (*Flash_CallbackPtr) (void);

which I don´t use.

Instead, I use 'F2806x_GlobalVariableDefs.c', which already includes the above lines. In the debugger I see that they contain some values that are not changing after being initialised.

I just ran the TI-example with my FLASH-Linker file, and it worked. The only thing that I had to comment out were some lines belonging to CLA (scratchpad...). Do you think this makes the difference? For reference I attached the used Linker-file.

Concerning the Map-files, what functions/libs should I precisely look for? As far as I can see, my own functions are located where they should be.

best regards,

Stefan

Stefan,

could you try below initialization in your main ()

   EALLOW;

   Flash_CallbackPtr = &MyCallbackFunction;

   EDIS;

and define below in your application, like the Flash API example does? 

/*------------------------------------------------------------------

  Callback function - must be executed from outside flash/OTP

-----------------------------------------------------------------*/

#pragma CODE_SECTION(MyCallbackFunction,"ramfuncs");

void MyCallbackFunction(void)

{      

// Toggle pin, service external watchdog etc

    MyCallbackCounter++;  

asm("    NOP");

}

if above doesn't help would you be able to send a zipped version of your application that is failing, I can try to reproduce and debug it.

Best Regards

Santosh

Santosh,

bad news, it didn´t help. It makes no difference if CallbackFunction() is used or not.

I tried to prepare my project so you can run it independantly from the specific hardware that I use (like I2C-Port-expander, SPI-DataFLASH, HyperTerminal-Communication...), so I had to remove al lot of Code to make someone else able to run it on a different hardware. Unfortunately this solved the problem, it works now!

Of course, this is not an option, so I restored the functions one by one until the failure was back again, but the result still makes no sense to me:

1) To make it work I need to turn off all PWM clocks:  SysCtrlRegs.PCLKCR1.all = 0;

Since I don´t use interrupts related to ePWMx I have no idea how this can affect an erase-function running from BootROM. All used ePWM-interrupts only start SOC´s which then call CLA-tasks. Anyway, turning off the ePWM clocks before erasinf/programming is fine for me.

2) To make it work I must not call a certain function related to HyperTerminal communication.

Here I´m still inverstigating, but no clue so far. I checked the RAM memory, but could not find significant changes like large blocks being altered, so I don´t think that my ramfuncs get corrupted. What I observed when stepping through the BootROM was that MyCallbackFunction() gets called quite often, before the Programm finally crashes. Maybe there´s a chance to put some more information into MyCallbackFunction() then just an up-counter...

best regards,

Stefan