TMDS243EVM: OSPI_DIAG example fails

Daisuke Maeda

Guru 20205 points

Part Number: TMDS243EVM
Other Parts Discussed in Thread: TMDS64EVM

Our customer sees OSPI_DIAG example fails on TMDS243EVM.
I have confirmed that it reproduces on TMDS64EVM as well.

[OSPI Flash Diagnostic Test] Starting ...
[OSPI Flash Diagnostic Test] Flash Manufacturer ID : 0x34
[OSPI Flash Diagnostic Test] Flash Device ID : 0x5B1A
[OSPI Flash Diagnostic Test] Executing Flash Erase on first block...
[OSPI Flash Diagnostic Test] Done !!!
[OSPI Flash Diagnostic Test] Performing Write-Read Test...
ERROR: ospi_flash_diag_test_compare_buffers:185: OSPI read data mismatch !!!
Some tests have failed!!

This issue seems to be caused by insufficient wait time after Flash has been erased because the example works fine when more than a suitable wait time is added after Flash is erased in the code.

  DebugP_log("[OSPI Flash Diagnostic Test] Executing Flash Erase on first block...\r\n");
  status = OSPI_norFlashErase(ospiHandle, offset);
  if(SystemP_SUCCESS == status)
  {
      DebugP_log("[OSPI Flash Diagnostic Test] Done !!!\r\n");
  }
  else
  {
      DebugP_log("[OSPI Flash Diagnostic Test] Erase Failed !!!\r\n");
  }
#if 1 // Changed codes
  int32_t loop_count;
  for (loop_count = 0; loop_count < 40000000; loop_count++); // for wait
#endif
  DebugP_log("[OSPI Flash Diagnostic Test] Performing Write-Read Test...\r\n");
  status = OSPI_norFlashWrite(ospiHandle, offset, gOspiTxBuf, APP_OSPI_DATA_SIZE);

Could you tell me the correct fix for the example to work fine?

Best regards,

Daisuke

over 2 years ago

0 Aakash Kedia over 2 years ago

TI__Mastermind 26145 points

Hi Daisuke Maeda,

If the issue is reproducible on am243x-evm, can you please attach the project in which the issue is replicable ?

We would like to root-cause the issue and then analyze the correct fix for the same.

Best Regards,
Aakash

0 Daisuke Maeda over 2 years ago in reply to Aakash Kedia

Guru 20205 points

Hi Aakash-san,

The example in MCU+ SDK 08.05.00.24 should be able to reproduce the issue. Could you please try it?

C:\ti\mcu_plus_sdk_am243x_08_05_00_24\examples\drivers\ospi\ospi_flash_diag\am243x-evm

Best regards,

Daisuke

0 Aakash Kedia over 2 years ago in reply to Daisuke Maeda

TI__Mastermind 26145 points

Hi Daisuke Maeda,

We have replicated the issue on our end. I am not sure how did it miss from our end. Apologies for the inconvenience.
We have not root-caused the issue so we are not sure, what exactly went wrong here. We will rectify the issue and send an patch by 2nd Feb to unblock you on the same.

Best Regards,
Aakash

0 Aakash Kedia over 2 years ago in reply to Aakash Kedia

TI__Mastermind 26145 points

Hi Daisuke Maeda,

One observation on our end is that the example works fine with Release build but there seems to be some problems with Debug build.
Can you test the same on your end and confirm ?

Hope this unblocks you.

Best Regards,
Aakash

0 Daisuke Maeda over 2 years ago in reply to Aakash Kedia

Guru 20205 points

Hi Aakash-san,

Thank you for your reply. Sorry for the late reply.

The example worked without error with the Release build.

Logs on the CCS console are not output until the target core is suspended, is this behaviour correct?

Why does the example fail with the Debug build?

I am concerned that the method used to check the completion of Flash erase in the example may be incorrect.

Best regards,

Daisuke

+1 Aakash Kedia over 2 years ago in reply to Daisuke Maeda

TI__Mastermind 26145 points

Hi Daisuke Maeda,

Daisuke Maeda said:
Why does the example fail with the Debug build?

The issue was with the flash dummy cycles. We were not initializing the parameters for it so in Release mode, it was assuming it to be 0 whereas in Debug Mode it was initializing it with some dummy values.

I think this flash should be good enough for you. Can you try the same on your end ?

/cfs-file/__key/communityserver-discussions-components-files/908/flashDiagDiff.patch

Hope this helps.

Best Regards,
Aakash

0 Daisuke Maeda over 2 years ago in reply to Aakash Kedia

Guru 20205 points

Hi Aakash-san,

Thank you for your reply.

The patch file you attached seems to indicate that the setting to define dummy cycles (rdParams->dummyBits = 0) has been added.

Are different versions of the source code (ospi_v0.c) used in debug and release mode?
In other words, were no dummy cycles defined in debug mode and dummy cycles defined (rdParams->dummyBits = 0) in release mode?

The source code (ospi_v0.c) does not seem to be used directly in the project.

Does the lib file (drivers.am243x.r5f.ti-arm-clang.debug.lib) need to be rebuilt to apply the patch?

Best regards,

Daisuke

+1 Aakash Kedia over 2 years ago in reply to Daisuke Maeda

TI__Mastermind 26145 points

Hi Daisuke-san,

Daisuke Maeda said:
Are different versions of the source code (ospi_v0.c) used in debug and release mode?
In other words, were no dummy cycles defined in debug mode and dummy cycles defined (rdParams->dummyBits = 0) in release mode?

No they are the same files. It was not defined in either mode but the compiler was able to treat dummyBits as 0 for release mode but did not do the same for debug mode. This is how the stack handling was done in the compiler.

Daisuke Maeda said:
Does the lib file (drivers.am243x.r5f.ti-arm-clang.debug.lib) need to be rebuilt to apply the patch?

Yes, you need to rebuild the libraries for this change to take the effect.

Best Regards,
Aakash

0 Daisuke Maeda over 2 years ago in reply to Aakash Kedia

Guru 20205 points

Hi Aakash-san,

Thank you for your reply.

I understand that the initial value of the dummy cycles is coincidentally set to 0 in Release mode due to build optimisation etc, but must be set to 0 explicitly for correct operation.

Will the issue be fixed in the next release SDK?

Best regards,

Daisuke

+1 Aakash Kedia over 2 years ago in reply to Daisuke Maeda

TI__Mastermind 26145 points

Hi Daisuke-san,

We will fix the same in next revision of SDK.

Best Regards,
Aakash

0 Daisuke Maeda over 2 years ago in reply to Aakash Kedia

Guru 20205 points

Hi Aakash-san,

Thank you for your reply.

After I modified the source code (ospi_v0.c) as per the patch file and then rebuilt the \drivers\lib files, the OSPI_DIAG example worked fine in both Debug and Release builds.

Best regards,

Daisuke

Arm-based microcontrollers

Arm-based microcontrollers forum

TMDS243EVM: OSPI_DIAG example fails