[PATCH v2] drm/bridge: analogix_dp: Grab runtime PM reference for DP-AUX

Brian Norris briannorris at chromium.org
Tue Feb 15 16:40:49 PST 2022


On Tue, Feb 15, 2022 at 3:46 PM Doug Anderson <dianders at chromium.org> wrote:
> On Tue, Feb 15, 2022 at 2:52 PM Brian Norris <briannorris at chromium.org> wrote:
> > It still makes me wonder what the point
> > of the /dev/drm_dp_aux<N> interface is though, because it seems like
> > you're pretty much destined to not have reliable operation through
> > that means.
>
> I can't say I have tons of history for those files. I seem to recall
> maybe someone using them to have userspace tweak the embedded
> backlight on some external DP connected panels? I think we also might
> use it in Chrome OS to update the firmware of panels (dunno if
> internal or external) in some cases too? I suspect that it works OK
> for certain situations but it's really not going to work in all
> cases...

Yes, I believe I'm only submitting patches like this because fwupd
apparently likes to indiscriminately whack at dpaux devices:
https://github.com/fwupd/fwupd/tree/main/plugins/synaptics-mst#kernel-dp-aux-interface

That seems like a bad idea.

(We've already disabled that plugin on these systems, but it seems
wise not to leave the stumbling block here for the next time.)

> I suppose this just further proves the point that this is really not a
> great interface to rely on. It's fine for debugging during hardware
> bringup and I guess in limited situations it might be OK, but it's
> really not something we want userspace tweaking with anyway, right? In
> general I expect it's up to the kernel to be controlling peripherals
> on the DP AUX bus. The kernel should have a backlight driver and
> should do the AUX transfers needed. Having userspace in there mucking
> with things is just a bad idea. I mean, userspace also doesn't know
> when a panel has been power cycled and potentially lost any changes
> that they might have written, right?
>
> I sorta suspect that most of the uses of these files are there because
> there wasn't a kernel driver and someone thought that doing it in
> userspace was the way to go?

*shrug* beats me.

Brian



More information about the Linux-rockchip mailing list