Re: [PATCH v3 00/14] Add DRM driver for Solomon SSD16xx e-paper display controllers
From: Thakkar, Devarsh
Date: Thu Nov 19 2026 - 12:25:28 EST
Hi Maxime, Javier,
On 11/10/2026 6:35 PM, Maxime Ripard wrote:
> On Tue, Nov 10, 2026 at 01:39:19PM +0100, Javier Martinez Canillas wrote:
>> Devarsh Thakkar <devarsht@xxxxxx> writes:
>>
>> Hello Devarsh,
[...]
>>
>> [...]
>>
>>> drm/solomon: ssd16xx: Expose refresh mode as plane property
>>> drm/solomon: ssd16xx: Expose color mode as plane property
>>> drm/solomon: ssd16xx: Expose session management as plane properties
>>
>> There is a fundmental issue with these 3 patches. They are adding custom
>> device-specific plane properties, but the KMS/DRM subsystem requires all
>> KMS properties to be generic and standardized:
>>
>> https://docs.kernel.org/gpu/drm-kms.html#requirements
>
> And with an established open-source project using it.
>
Thanks for the review, and agreed on the requirements. I will drop
the three property patches from v4 and use fixed per-panel defaults,
so v4 adds no new uAPI.
I would also like your guidance on how to make these properties generic,
in particular on the open-source user-space requirement.
To my knowledge, no display server or compositor targets e-paper today.
Users write standalone applications that drive the panel through KMS,
and ours are being built on LVGL [1], an MIT-licensed embedded GUI
library. I see that its upstream DRM backend already uses atomic commits
and sets plane and CRTC properties, so supporting a generic e-paper
property there should be feasible.
1. Would support in the upstream LVGL DRM backend be an acceptable
open-source user?
2. Would a generic e-paper refresh mode property, created by a
core helper and decoded by the core, be acceptable in principle?
I would start with the refresh mode alone.
3. Would a CRTC property be the right object? A refresh drives
the whole panel with one waveform after the planes are composed,
so with more than one plane a per-plane mode could conflict
within one commit? Though in general i have seen single plane
epaper panels only.
Some background: I should have mentioned in the commit messages that
we brought up this topic at this year's Display Next Hackfest too, where
I presented on e-paper refresh modes and how parts of them could be
generalized in a common helper.
Similar to drm_crtc_create_scaling_filter_property(), we can create
a helper the driver passes a default and the modes its panel supports,
drm_crtc_create_epaper_refresh_property(crtc,
DRM_EPAPER_REFRESH_PARTIAL, /* default */
BIT(DRM_EPAPER_REFRESH_PARTIAL) |
BIT(DRM_EPAPER_REFRESH_FULL) |
BIT(DRM_EPAPER_REFRESH_FAST));
and the core creates an "epaper_refresh" enum property limited to
those modes, and decodes it into the CRTC state, as it does for
the scaling filter property today.
I have since written up an update of those slides in [2]. It goes
through each v3 property and its use cases, and covers which ones we
think can be generic and which should stay driver internal at least for
now. To summarize, for now the generic candidates are as below:
- refresh mode per update (partial / full / fast); the driver
declares the subset its panel supports
- refresh mode for the first frame of a new DRM master
- clearing the panel when a new master opens, when it closes
(including a crash), and on CRTC disable. A bistable panel
otherwise keeps the last image.
These depend on the application rather than the panel. For example,
a boarding-pass kiosk must erase the passenger's details on exit,
while a conference badge should keep the name shown.
I see similar choices exist outside SSD16xx, so in future if new epaper
drivers are added they may benefit with these generic properties too.
For e.g., The Waveshare reference drivers have separate full / fast /
partial update calls [3], and NXP's downstream EPDC driver takes a
waveform and a partial / full mode with each update [4]. The out-of-tree
rockchip_ebc driver uses partial updates for each frame and a global
refresh otherwise, and the PineNote tree adds a private ioctl to request
one global refresh [5].
Could you please share your opinion on this ?
Once the direction is clear, I will work on the generic properties as
a separate RFC, so that they do not block the driver.
[1]
https://github.com/lvgl/lvgl/blob/2ee5654440bfc21a2cade4aa6e248f8d34b5e33e/src/drivers/display/drm/lv_linux_drm.c#L442-L493
[2]
https://github.com/devarsht/linux-next/blob/epaper_v4_cleanup/epaper-refresh-modes.md
[3]
https://github.com/waveshareteam/e-Paper/blob/d28b3c4cfc78eef1c6cd6e6e3013654e449d2864/RaspberryPi_JetsonNano/python/lib/waveshare_epd/epd2in13_V4.py#L268-L287
[4]
https://github.com/nxp-imx/linux-imx/blob/06213bb0115f587f55ec1f45fac03544855f97a7/include/uapi/linux/mxcfb.h#L131-L134
[5]
https://github.com/m-weigand/linux/blob/11a849851a8d31dd8f0eec003d45763dd3212c77/drivers/gpu/drm/rockchip/rockchip_ebc.c#L1418-L1437
Regards,
Devarsh
> Maxime