Re: [PATCH 0/7] ARM: dts: qcom: msm8974: klte: Enable some hardware

[Iskren Chernev]

On 7/1/20 1:30 PM, Brian Masney wrote:
> Hi Iskren,
>
> On Tue, Jun 30, 2020 at 05:09:05PM +0300, Iskren Chernev wrote:
>> Enable support for various hw found on the Samsung Galaxy S5:
>> - touchkey (the two buttons around the home button)
>> - touchscreen
>> - notification led
>> - wifi
>> - external SD card
>>
>> Please note that for working wifi the correct firmware is needed. Check [1]
>> for links and locations.
>>
>> Also note, that to actually run a mainline kernel on the klte, you'd need
>> to apply this patch [2]. Any feedback on getting this to run on pure
>> mainline are welcome.
>>
>> [1] https://gitlab.com/postmarketOS/pmaports/-/blob/master/firmware/firmware-samsung-klte/APKBUILD
>> [2] https://gitlab.com/postmarketOS/linux-postmarketos/-/commit/765f55b248cd3b231af8431fe2f2aeca263b4e4b
>
> Good to see more msm8974 support upstream!
>
> Regarding the second patch, that should only be needed in order to
> use the GPU on these devices.

For the klte, without this delay rpm init patch, the device reboots soon
after being booted. I tested with pure v5.8 and next-20200701 (both have
basic klte support, including usb ethernet for ssh). With the patch
applied, it works like a charm.

Note that there is no GPU enabled in this setup (unless I'm missing some
piece from the SOC dtsi).

> I hope that patch won't be needed once
> IOMMU support is added, which is the last major missing piece in order
> to have the GPU working upstream. I posted a RFC patch [3] in January
> but didn't get any suggestions. I suspect some kind of memory corruption
> since the board gets unstable with that patch.

If the IOMMU is only used for graphics, then the problem remains for klte.

> If that hack patch is still needed to use the GPU once IOMMU support is
> in place, then I planned to troubleshoot it further by adding some log
> statements to various probe functions with and without that patch to see
> the probe order between the various subsystems. I suspect the issue is
> that a clock isn't ticking yet.
>
> [3] https://lore.kernel.org/lkml/20200109002606.35653-1-masneyb@xxxxxxxxxxxxx/T/#u

I'll be happy to help with that.

> Brian

Iskren