Re: [PATCH] drm/mediatek: mtk_dsi: enable hs clock during pre-enable

[AngeloGioacchino Del Regno <angelogioacchino.delregno@collabora.com>]

On 7/8/26 14:13, Gary Bisson wrote:
> Hi Adam,
> 
> On Tue, Jul 07, 2026 at 08:25:41PM -0500, Adam Thiede wrote:
>> On 7/7/26 05:38, AngeloGioacchino Del Regno wrote:
>>> On 7/7/26 10:51, Thorsten Leemhuis wrote:
>>>> On 7/7/26 04:20, CK Hu (胡俊光) wrote:
>>>>> On Mon, 2026-07-06 at 12:16 +0200, Thorsten Leemhuis wrote:
>>>>>> On 6/22/26 15:23, Adam Thiede wrote:
>>>>>>> On 6/22/26 06:22, Gary Bisson wrote:
>>>>>>>> On Thu, Jun 18, 2026 at 04:06:28PM -0500, Adam Thiede wrote:
>>>>>>>>> On 1/20/26 05:36, Gary Bisson wrote:
>>>>>>>>>> Some bridges, such as the TI SN65DSI83, require the HS clock to be
>>>>>>>>>> running in order to lock its PLL during its
>>>>>>>>>> own pre-enable function.
>>>>>>>>>>
>>>>>>>>>> Without this change, the bridge gives the following error:
>>>>>>>>>> sn65dsi83 14-002c: failed to lock PLL, ret=-110
>>>>>>>>>> sn65dsi83 14-002c: Unexpected link status 0x01
>>>>>>>>>> sn65dsi83 14-002c: reset the pipe
>>>>>>>>>> [...]
>>>>>>>>>
>>>>>>>>> This commit was part of 7.1 and caused a problem for me.
>>>>>>>>> I'm running postmarketOS (basically Alpine Linux) on a Lenovo C330
>>>>>>>>> chromebook with a Mediatek MT8173 processor.
>>>>>>>>> The problem: when the display on my laptop
>>>>>>>>> powers off (via suspend or
>>>>>>>>> idle,
>>>>>>>>> like xset dpms off) the picture does not come back when the display
>>>>>>>>> powers
>>>>>>>>> back on (from resume). The display backlight
>>>>>>>>> comes on and brightness is
>>>>>>>>> adjustable but there is no picture. The only fix is to reboot.
>>>>>>>>>
>>>>>>>>> Reverting this commit and applying it as a patch on top of 7.1
>>>>>>>>> addresses the
>>>>>>>>> issue for me.
>>>>>>>>>
>>>>>>>>> You can view the config I'm using here:
>>>>>>>>> https://urldefense.com/v3/__https://gitlab.postmarketos.org/
>>>>>>>>> postmarketOS/pmaports/-/__;!!CTRNKA9wMg0ARbw! jrsPDtSEUdaINzLlq92Li8gmsEBkTOxZ6WUzNHjvIN6CyOJjHiHkNSOhIRPXFTLPlaYlxU2uvryVkwjUAN9bmax5$
>>>>>>>>> merge_requests/8819
>>>>>>>>>
>>>>>>>>> Is there any sort of testing or other debugging
>>>>>>>>> info I can provide to
>>>>>>>>> help
>>>>>>>>> address this issue?
>>>>>>>>
>>>>>>>> Thanks for reporting the issue, could you share some
>>>>>>>> logs? Is the driver
>>>>>>>> saying anything during resume? Also, what type of
>>>>>>>> panel is used on that
>>>>>>>> chromebook?
>>>>>>>
>>>>>>> The curious thing is that there are no real logs in dmesg or /var/log/
>>>>>>> messages about this. This picture just fails to come
>>>>>>> back. If there are
>>>>>>> some kernel params I can set to get deeper logging, that
>>>>>>> would help, but
>>>>>>> I'm not aware of any.
>>>>>>>
>>>>>>> I think the panel is a "BOE NV116WHM-T00" - I used this command to get
>>>>>>> info: cat /sys/class/drm/card0-eDP-1/edid | edid-decode
>>>>>>>
>>>>>>> Output:
>>>>>>> https://urldefense.com/v3/__https://termbin.com/8nbd__;!!
>>>>>>> CTRNKA9wMg0ARbw! jrsPDtSEUdaINzLlq92Li8gmsEBkTOxZ6WUzNHjvIN6CyOJjHiHkNSOhIRPXFTLPlaYlxU2uvryVkwjUAJooSyqL$
>>>>>>
>>>>>> This looked stalled. If I'm mistaken here, please let me known; but if
>>>>>> no solution is in sight, should we maybe just revert the change until a
>>>>>> proper was found?
>>>>>
>>>>> It's welcome anyone to provide a revert patch,
>>>>> but I would still wait for the fixup patch until 7.2-rc4.
>>>>> If no fixup patch exist, then apply the revert patch.
>>>>   From my understanding of things the position in the devel cycle doesn't
>>>> matter much in a case like this. To quote Linus statements from
>>>> https://www.kernel.org/doc/html/latest/process/handling-
>>>> regressions.html#on-how-quickly-regressions-should-be-fixed
>>>>
>>>> """
>>>>   From 2026-01-22:
>>>>
>>>>    But a user complaining should basically result in an immediate fix -
>>>>    possibly a "revert and rethink".
>>>>
>>>> With a later clarification on 2026-01-28:
>>>>
>>>>    It's also worth noting that "immediate" obviously doesn't mean "right
>>>>    this *second* when the problem has been reported".
>>>>
>>>>    But if it's a regression with a known commit that caused it, I think
>>>>    the rule of thumb should generally be "within a week", preferably
>>>>    before the next rc.
>>>> """
>>>>
>>>> Adam reported the problem about three weeks ago, so we are way past the
>>>> "rule of thumb" timeframe Linus set.
>>>>
>>>> Ciao, Thorsten
>>>
>>> This is a kind of odd situation here. The fix from Adam is actually
>>> correct, as in,
>>> the SN65DSI83 bridge gets broken without...
>>>
>>> ....but then, there's some more oddness going on: I tried to reproduce
>>> this on my
>>> MT8173 Elm device, but there I can resume the system just fine, and the
>>> display is
>>> up and running like normal?
>>>
>>> I'm not sure what to advice here at this point - just adding some info.
>>>
>>> Cheers,
>>> Angelo
>> Angelo, would you mind sharing your config, and the details of what you're
>> running? It might be a configuration difference.
> 
> As a FYI here is the latest branch I tested that patch on:
> https://github.com/gibsson/linux-next/commits/master-mtk/
> 
> As far as I remember suspend/resume was ok. It was tested with a
> Tungsten510 SMARC SOM (MT8370) + sn65dsi83 + tm070jdhg30 display.
> Both the bridge and the display are mainline, you can see that the main
> change is in the defconfig where a bunch of options are changed to be
> built-in (for NFS booting).
> 
> I'll try to repro against latest kernel this week and report back.
> 
> In your case, is the panel upstream as well?
> 
> Regards,
> Gary

Gary, thanks for chiming in.

The machine he's referring to is using a Parade PS8640 DSI to eDP bridge: check
mt8173-elm.dtsi for (slightly) more information.

Cheers,
Angelo