Re: [PATCH v3 13/14] drm/mediatek: Support DRM plane alpha in OVL

From: Adam Thiede
Date: Thu Oct 03 2024 - 11:30:28 EST


On 10/3/24 00:17, Jason-JH Lin (林睿祥) wrote:
Jason:
That is a lot of information, and quite above my head! Thank you
though.

I should note that the log items I sent you are from the "good"
kernel - 6.11 with the commit reverted. Here is a much longer set of logs: https://termbin.com/co6v

I've rebuild 6.11 with the log statement enabled and the "bad"
behavior.
Here is a dmesg from that: https://termbin.com/xiev

Hi Adam,

I think something wrong with your dmesg links, both logs look the same.
We should see this log in the "bad" one:
fmt:0x34325258, has_alpha:0x0, alpha:0xffff, con:0x2000

Apologies, I did indeed upload the same file twice. Here is the "good" one:
https://termbin.com/tb33

And the "bad" one:
https://termbin.com/yhxx

I think the fact that we're not seeing that line in the "bad" one is part of the problem?

But anyway, I think the reason for the downgrade is clear enough to me.
So let's try to figure out the solution.

These logs are both from `dmesg`.

I'm fairly certain I've built with the patch you referenced enabled.
The kernels I run are just release kernels, not RCs or git branches or anything. The mainline v6.11 kernel is the one that has this problem.
If that patch has been merged into 6.11 (which, looks like it has) then it's in the kernel I'm building.

Got it.
Then OVL_CONST_BLEND might be the unsupported configuration in MT8173,
I think we should remove the XRGB8888 format for MT8173.

Could you please try this modification and see if it'll change to use
others supported format to show the text?

--- a/drivers/gpu/drm/mediatek/mtk_disp_ovl.c
+++ b/drivers/gpu/drm/mediatek/mtk_disp_ovl.c
@@ -102,12 +102,9 @@ static inline bool is_10bit_rgb(u32 fmt)
}

static const u32 mt8173_formats[] = {
- DRM_FORMAT_XRGB8888,
DRM_FORMAT_ARGB8888,
- DRM_FORMAT_BGRX8888,
DRM_FORMAT_BGRA8888,
DRM_FORMAT_ABGR8888,
- DRM_FORMAT_XBGR8888,
DRM_FORMAT_RGB888,
DRM_FORMAT_BGR888,
DRM_FORMAT_RGB565,

I've not been able to get the kernel to build with that patch; it keeps segfaulting at the end. I will keep attempting though.


Regards,
Jason-JH.Lin


- Adam Thiede