[PATCH v2 1/4] arm64: dts: rk3399-u-boot: Delete vop assigned-clocks/rates

[Jagan Teki]

Hi Kever,

On Thu, Apr 2, 2020 at 2:48 PM Kever Yang <kever.yang at rock-chips.com> wrote:
>
> Hi Jagan,
>
> On 2020/3/31 下午1:59, Jagan Teki wrote:
> > On Tue, Mar 31, 2020 at 1:06 AM Mark Kettenis <mark.kettenis at xs4all.nl> wrote:
> >>> From: Jagan Teki <jagan at amarulasolutions.com>
> >>> Cc: sunil at amarulasolutions.com, u-boot at lists.denx.de,
> >>>          linux-rockchip at lists.infradead.org, linux-amarula at amarulasolutions.com,
> >>>          Jagan Teki <jagan at amarulasolutions.com>
> >>> Date: Mon, 30 Mar 2020 23:46:10 +0530
> >>> Content-Type: text/plain; charset=UTF-8
> >>>
> >>> Linux supporting assigned-clocks for VOP on rk3399 by assuming
> >>> U-Boot not initializing it on this linux commit:
> >>>
> >>> commit <617f4472bdd3> ("arm64: dts: rockchip: init rk3399 vop clock rates")
> >>>
> >>> There is no specific need to initialize these assigned clock
> >>> in U-Boot as video drivers still work with default aclk and
> >>> hclk values. So, these clocks are simply not supported by rk3399
> >>> clock driver.
> >>>
> >>> But, during stdio probe of vidconsole, the device probe
> >>> will try to check whether the assigned clocks on that video
> >>> console node is initialized or not? and return error if not.
> >>>
> >>> So, delete these property via -u-boot dtsi as there is
> >>> no specific need in U-Boot.
> >> Deleting these properties isn't very helpful as it means the U-Boot
> >> device tree can no longer be used by the kernel.  Isn't it a better
> >> idea to implement these clocks as stubs in the u-boot clock driver?
> > I did try this before sorting out these changes, seems like it
> > requires a bit more tweaking the clock wrt display code. I really
> > didn't see any use case as of now for just to print u-boot log on
> > display out, and more over this support has been broken since from
> > releases. so bypassing these nodes can be a solutions for now.
>
>
> I agree with Mark for not touch the dts first. I don't know the detail
> of display driver but:
>
> - The rk3399 driver use to work without touch dts from kernel;
>
> - the clock driver have a rk3399_vop_set_clk() which does not depends on
> dts.

The existing video drivers are written based on the puma dts and those
are not inline to Linux dts files, i.e. the reason the code is pushed
I think. The rest of rk3399 dtsi files are now inline to Linux as and
display out on these are broken from last 2 releases. so my idea is to
resolve the things one-after-another like
1. Make existing video stuff work with all rk3399 (this series along
with this patch)
2. Drop this patch change and make video drivers working w/o any
explicit changes in dts like this patch does.

Since step 2, would take time, and require close testing of all boards
I would like to pick the existing stuff for the release. Mark my words
to fix the things for the next release.

Jagan.