ATF - Black screen when Secure Boot enabled (SB_EN=1) fixed by setting NXP_ROM_RSVD to 0x8000

[daviddg]

Hi,

I've an issue with an ls1043a soc. When the secure boot is enabled (SB_EN=1 in rcw) with the fuse blown with the keys (but ITS=0), the console stay black...

When debugging with a Jtag device, it comes out that the i/esbc is stuck in a function that looks like a panic: "0x1d0: b 0x1d0". The Scratch status register seems not initialised (random or 0 value).

And after few days of debugging and tries, I finally fixed the issue by patching the file "plat/nxp/soc-ls1043a/soc.def", changing the value of "NXP_ROM_RSVD" to"0x8000" instead of "0x5900". Actually the same change as the one done to the ls1046a in the commit hash 41556d6b428d2ae4b265bb8a87125dfb3f5bd307 (on the qoriq-atf repo)...

So, my guess is that the ls1043a soc had his ROM code updated "recently" with one similar to the ls1046a and give an issue when booting with SB_EN=1 only (no problems if the secure boot is not enabled thought).... The soc we have was bought 3 weeks ago, so it's a recent one: LS1043AXE8QQB-SBRG2206-1N96N.

As I've also an ls1046a chip, if I revert that commit for the ls1046a def, the same behaviour as the ls1043a happens: Black screen and stuck at (almost) the same location with the same assembly code (panic).

So, could someone at NXP could just confirm me my guess is right and the issue is really fixed ?

Thanks.

Regards,

David.

[Oswalag · Accepted solution]

Hello, 

The issue isn't related to the chip date(neither recent nor old) , there is a there is a known issue for LS1046a/43a since LSDK21.08. the patch you applied is the correct one as workaround for the mentioned issue.  

Regards.

[daviddg]

Perfect then. Thanks for the investigation and confirmation

Is there an official patch somewhere ?

[Oswalag]

Hello,

I'm investigating your issue, I'll update here.