Re: exynos5422-odroid-xu3: MCPM: only 6 of 8 CPUs coming online

[Kevin Hilman <khilman@kernel.org>]

Abhilash Kesavan <kesavan.abhilash@gmail.com> writes:

> On Thu, Oct 23, 2014 at 9:26 PM, Kevin Hilman <khilman@kernel.org> wrote:
>> Abhilash Kesavan <a.kesavan@samsung.com> writes:
>>
>>> The change was based on my reading of the platsmp file of the
>>> Odroid-xu3 3.10 kernel and issues experienced previously. From what I
>>> understand, the platsmp code in 3.10 kernel is trying to reset the A7s
>>> and my change just bypasses the check (no magic here ). The change was
>>> only suggested as a workaround to get 8 cores up using MCPM and not an
>>> actual fix.  I had explained this in my mail to Kevin. So, if this
>>> were to be mainlined we would probably need to use something like
>>> exynos_core_restart in mcpm along with the workaround. But, this is
>>> all guesswork on my part and someone with knowledge of the 3.10
>>> kernel/bootloader needs to comment.
>>
>> Unfortunately, the harkernel tree git repo doesn't help understand where
>> the change came from because an entire v3.10.y based Samsung BSP is
>> checked into the tree on top of a v3.8.y kernel in a single commit !!!
>>
>> Abhilash, in the absence of the original author, could write up a patch
>> with your change?
>>
>> Or, if you can help me with a detailed changelog, I can work on the patch.
>> I don't have any exynos docs, so it would be good to have a description
>> of what reading/writing to that register are actually doing.
>
> I could post a patch with an explanation of how I think the workaround
> works, but there are a few issues:
> - I am not completely confident that this is the best/right change.

Maybe post as RFC.

> - We are adding code to the exynos mcpm back-end to work-around a
> bootloader setting for Odroid-xu3. We would have to make it specific
> to odroid somehow as we do not require it for 5420/5800 chromebooks.

Hmm, if you have an idea how/where to fix it in u-boot, I'd be glad to
give that a try.  I'm using the hardkernel u-boot repo:
https://github.com/hardkernel/u-boot.git, branch odroidxu3-v2012.07, and
would be happy to test any fixes there.

> - This might solve the 8 core boot-up issue but other scenarios might
> still fail like the one Joonyoung has already reported.

I replied yesterday to that report, and the one Joonyoung reported
exists with or without your proposed fix, so I don't think they're
directly related.

>>
>> Also, can you explain what you meant by using exynos_core_restart?  I
>> don't see that function in mainline.
>
> It is present in linux-next and kgene's for-next branch as part of
> this commit "ARM: EXYNOS: SWRESET is needed to boot secondary CPU on
> exynos3250".

OK, I see it now.

Kevin