Re: [PATCH 3/4] arm64: dts: qcom: sc8280xp: Add initial support for Ntmer TW220

[Konrad Dybcio <konrad.dybcio@oss.qualcomm.com>]

On 7/30/25 5:29 PM, Pengyu Luo wrote:
> On Wed, Jul 30, 2025 at 8:53 PM Konrad Dybcio
> <konrad.dybcio@oss.qualcomm.com> wrote:
>>
>> On 7/6/25 5:43 AM, Pengyu Luo wrote:
>>> On Sat, Jun 28, 2025 at 3:48 AM Konrad Dybcio <konrad.dybcio@oss.qualcomm.com> wrote:
>>>> On 6/17/25 11:29 AM, Pengyu Luo wrote:
>>>>> The Ntmer TW220 is a WOS tablet based on the Qualcomm SC8280XP platform,
>>>>> also known as the Robo&Kala 2-in-1 Laptop. Thanks to Hong for providing
>>>>> the unlocked device and early development work. This patch adds an
>>>>> initial device tree to enable basic functionality.
>>>>>
>>>>> Currently supported components include:
>>>>> - Bluetooth & Wi-Fi (board file regeneration required)
>>>>> - Battery charging (up to 15V/3A fixed PDO) and reporting via pmic-glink
>>>>> - Flash LEDs (front and rear)
>>>>> - Hall sensor (lid detection)
>>>>> - Keyboard (via Bluetooth or USB)
>>>>> - NVMe SSD
>>>>> - Power and volume keys
>>>>> - Simple-framebuffer
>>>>> - Sound (playback and capture; top-left DMIC only, top-right works only
>>>>>   on Windows)
>>>>> - Touchscreen and stylus (requires GPI DMA support [1] and stylus support [2])
>>>>> - USB Type-C ports
>>>>>
>>>>> The following components are currently non-functional:
>>>>> - Cameras (GalaxyCore GC5035; only sensor ID is detectable, no frames in libcamera;
>>>>>   partial driver can be found on LKML archives)
>>>>> - DSI display (blank screen with `dsi_err_worker: status=4`; primary DSI register
>>>>>   dump included below)
>>>>> - Stylus wireless charger (CPS4035)
>>>>> - UCSI over GLINK
>>>>>
>>>>> [1]: https://lore.kernel.org/linux-arm-msm/20250617090032.1487382-3-mitltlatltl@gmail.com
>>>>> [2]: https://lore.kernel.org/linux-input/20250605054855.403487-2-mitltlatltl@gmail.com
>>>>>
>>>>> Note: This series does **not** include any confidential material. Those
>>>>> who wish to run Linux on this device should contact Ntmer, as the
>>>>> bootloader is locked via secure boot.
>>>>>
>>>>> Co-developed-by: Hong Zhu <vanyang@smail.nju.edu.cn>
>>>>> Signed-off-by: Hong Zhu <vanyang@smail.nju.edu.cn>
>>>>> Signed-off-by: Pengyu Luo <mitltlatltl@gmail.com>
>>>>>
>>>>> dsi_ctrl, reg = <0 0x0ae94000 0 0x400>;
>>>>> 0xae94000 20050001 000001f3 0000000b dddd1011
>>>>
>>>> This is not something we want in the commit log
>>>>
>>>
>>> I will remove it. I need help, then I attached it, two of my sc8280xp
>>> devices require dsi to work. Reversing and guessing wasted a lot of
>>> time. I will appreciate it if qcom could support it.
>>
>> There should be nothing interesting compared to what's on SM8350 when it
>> comes to DSI on 8280.
>>
>> I would assume something's wrong with how your panel is being set up.
> 
> Just to mention it. According to some public sources, the Adreno 1095 DPU
> is included in the sm8350, while the Adreno 1199 DPU is included in the
> sc8280xp, they are slightly different. Here is a quick comparison
> 
>                    dpu_version     dsi_version     dsi_phy_version
> Adreno 1095 DPU:     7.0            2.5.0           4.2
> Adreno 1199 DPU:     8.0            2.5.1           4.2

Patchlevel/step versions are usually not supposed to break any software
interfaces, so I'd be *really* surprised if this is the case

DPU is rather different (the one on 8280 is way larger than 8350's), but
ultimately they're built of the same bricks

> 
> The panel setup has been tweaked and tested countless times, I really don't
> think I screwed it up. I had already brought up a lcd video panel with dual
> dsi and dual dsc on the sm8650. On the sc8280xp, I’ve worked with two
> different panels:
> 
> - One LCD video panel with dual DSI and dual DSC
> - One OLED video panel with dual DSI only
> 
> Both fail to display anything and complain about DSI FIFO errors
> (dsi_err_worker: status=4). Someone also encountered a similar issue on
> the SC8180X([1]), another WoA platform.
> 
> Anyway, I had quit. This doesn’t bother me anymore. Just mentioning the
> info for reference.

Sad to see you go.

If you still want to, I saw a fix for a sneaky bug earlier today on
LKML that may actually be related to your issue, given you seem to
have a really high res panel..

https://lore.kernel.org/linux-arm-msm/20250730123938.1038640-1-quic_amakhija@quicinc.com/

Konrad