[U-Boot] IO Error Propagation on Armada-385

[Stefan Roese]

Hi Chris,

(added Kosta to Cc)

On 21.11.19 08:43, Chris Packham wrote:
> Hi Stefan and Luka,
> 
> I was wondering if you could shed some light on this commit
> 
> https://gitlab.denx.de/u-boot/u-boot/commit/501c098a1f2cdaa930cb1a7166d3724467890a64
> 
> At $dayjob we're running into an issue where certain invalid PCI
> accesses when running Linux lead to the CPU hanging. Setting the IO
> Err Prop bit back to 1 seems to prevent the hang.
> 
> Can you recall why u-boot needs this bit cleared? We'll probably need
> to make the kernel restore the setting but I'd expect the kernel
> maintainers to point out that we're essentially restoring the hardware
> default so we'd need to explain that u-boot clears it because X but
> the kernel needs it set.

Frankly, I don't remember the details leading to this commit, sorry (too
long ago). I also searched some older mails for any reference here, but
failed so far.

Perhaps Kosta or someone else at Marvell remembers the details here?
Kosta, could you please take a look?

Thanks,
Stefan