TDA4VEN-Q1: memtester issue in SBL boot mode

Hi,

This seems like maybe an issue related to moving from 8GB total DDR memory to 4GB total DDR memory. 

What are the full list of changes you made to account for the memory difference between the TI EVM and your custom board?

Regards,
Kevin

Hi Kevin,

In fact,using memtester 2G can test successfully in SPL boot. 

For the change from SPL boot to SBL boot, I only replace board_ddrReginit.h and rebuild sbl and replace the tiboot3.bin to SD card.

Is SBL boot to adapt to DDR modifications not just replacing board_ddrReginit.h?

Regards.

Hi Xie,

Is it the correct understanding that you only see this issue with SBL and not SPL. If so, this seems to confirm that the issue should not be related with the DDR register settings. 

Let me see if someone from our software team can help further.

xie jc said:

Is SBL boot to adapt to DDR modifications not just replacing board_ddrReginit.h?

For configuring the DDR sub-system from hardware perspective, that should be enough. I.e., as long as software / CPU is trying to access memory in a valid range then functionally it should be enough and work.

But from a system software perspective, there may be a need to update total memory size in other locations to prevent software from trying to access a memory location where no physical memory is present. For instance, if you were compiling code and linking the code to particular memory regions, build scripts might define the memory address range. Then code might get linked to an address where no physical memory is present. This is just an example though - ultimately the software team will need to confirm what assumptions are made about total DDR memory size and how that is determined.

This type of problem used to occur in u-boot code which would pass the total memory size to the Linux kernel. The output of the DDR tool was consumed by the DDR driver, which was initialized by the R5 core (tiboot3.bin). However, the total memory size was passed from the A72 binaries (tispl.bin / u-boot.img), where no DDR driver was present. Therefore, there was a need to manually update the total memory size in the A72 u-boot image so that the Linux kernel was aware of the correct amount of memory available (see section 3.1.2 in the following app note: https://www.ti.com/lit/pdf/spracu8). I have not been able to keep up with the newer SDK releases so I don't know what has changed or what might be unique for your particular SDK. 

Regards,
Kevin

Hi Kevin,

Thanks.I will make modifications via https://www.ti.com/lit/pdf/spracu8 and provide you with feedback.

But I am still confused that the boot mode from SPL to SBL, the Uboot, kernel, and fs are all the same file. Why does the problem of memtester failure occur when SBL boot？

Regards.

Hi Xie jc,

xie jc said:

But I am still confused that the boot mode from SPL to SBL, the Uboot, kernel, and fs are all the same file. Why does the problem of memtester failure occur when SBL boot？

Could you please tell me which boot flow you are using?

Regards,

Karthik

Hi Karthikeyan,

We are using SBL boot now.

Regards.

Hi All,

This issue can be reproduced in EVM when change the LPDDR4 from 8GB to 4GB as well, and the issue is only in SBL, so pls help priority this, this will be common issue for all customer.

BR,

Biao 

Hi Biao & Karthikeyan,

Spl startup replaces tiboot3.bin+tispl.bin, sbl startup replaces tiboot3.bin+app, and other files have not been modified or replaced. When querying the DRAM address size during sbl and spl startup, sbl actually detects 8g of memory size and spl detects 4g. Why is this?

Regards.

Hi JC，

Thanks for update.

BR,

Biao