[PATCH next RFC 0/5] rockchip: panic() when DRAM init fails

[Quentin Schulz]

Hi Kever,

On 11/7/24 4:35 AM, Kever Yang wrote:
> Hi Quentin,
> 
> On 2024/11/6 01:21, Quentin Schulz wrote:
>> I am the unfortunate owner of an RK3399 Puma that is failing DRAM init
>> every now and then with the following cryptic error message:
>> """
>> pctl_start: Failed to init pctl channel 0
>> """
>>
>> When it fails, the current logic is implemented in such a way that the
>> board enters an infinite while loop. This is not ideal for embedded
>> systems that are not easily accessible and also an issue when having
>> devices in a CI lab for example.
> 
> It's fine for this patch to make the choice in panic to hang or to reset 
> the system.
> 
> But I don't understand how this help you, does it may success to init 
> the dram after
> 
> the reset in your case? I think in most case the board will keep reset 
> because this is
> 
> more likely hardware issue and not able to became success after reset.
> 

Exactly. After a reset (manual by pressing the reset button, or by 
panic()) the board works again. I don't have exact numbers for the 
probability for this to happen, but I had a board doing boot-into-Linux 
tests in a loop and it failed multiple times in the same day. Gut 
feeling is around ~10-20 times for ~2200 reboots.

I am NOT saying this is not related to a HW issue though, but I've been 
using that RK3399 Puma for a while, and sometimes I have this issue. The 
rest of the time it seems to "work fine" (haven't run memtest or 
extensive testing on that unit though).

I am NOT sure BUT I think applying 
https://github.com/rockchip-linux/u-boot/commit/e1652d3918b182e107addd5e6f451655f239efbc 
on upstream U-Boot helped lower the probability (it still happened 
though!). Could be totally unrelated though and just flukes.

Cheers,
Quentin