Sorry Florian, I had a couple issues with my email client yesterday and 
I'm not sure if you have received my reply.

Please see my reply inline.

On 5/31/2018 3:16 PM, Florian Fainelli wrote:

On 05/31/2018 10:24 AM, Ray Jui wrote:

quoted

On 5/31/2018 1:22 AM, Cl?ment P?ron wrote:

quoted

Hi Rai,

On Wed, 30 May 2018 at 19:25, Ray Jui [off-list ref] wrote:

quoted

Hi Cl?ment,

Correct me if I'm wrong, but I thought the trend is to move to use
earlycon that can be activated from kernel command line for early print
before the serial driver is loaded.

Have you tried earlcon?

No, only tested this method.

Thanks,
Clement

If I remember it correctly, I think the trend is to use earlycon. There
are obvious shortcomings by making this configuration compile time based.

This is true, though on ARM 32-bit kernels DEBUG_LL gets used by the
kernel self-decompressor and also before earlycon has a chance to run,
this is useful to debugging memory issues where your memory
configuration is incorrect typically.

This is true.

Either way is fine with me (accepting or dropping) the patch, though
there is probably minimal impact in just accepting such a change.

Will leave the decision to you then.

Thanks,

Ray