rk2918 support
Steven Newbury
steve at snewbury.org.uk
Sat Oct 4 04:34:36 PDT 2014
On Fri, 2014-10-03 at 23:37 +0200, Heiko Stübner wrote:
> Hi Steven,
>
> Am Freitag, 3. Oktober 2014, 19:12:01 schrieb Steven Newbury:
> > I'm about to make a start bringing up Mer/Wayland (with the new
> > Vivante DRM/Gallium3D driver) on an older tablet which uses the
> > Rockchip rk2918 (aka rk29?) SoC.
> >
> > I can't find any definitive information on whether there is support
> > for this chip in the mainline kernel, I see rk2928 in the
> > rockchip/Kconfig, but no mention of rk2918.
>
> rk2928 is a single-core Cortex-A9, while your rk2918 is a Cortex-A8.
> And the
> rk2928 support in the kernel also just consists of "should print at
> least a
> little serial startup output before dying".
>
Explains why grepping the kernel sources revealed very little!
>
> > If this isn't in mainline, is there any intention to get it in?
>
> Only if somebody sends in the necessary pieces ;-) . My personal cut-
> off is at
> the rk3066, simply because that is the "oldest" board I own.
None of these chips are all that old, really, are they? It's such a
shame all the work was done in an isolated silo, there's a lot of hw
out there that really could be quite useful if it wasn't stuck on old
versions of Android after the vendor moved onto their next model or
went out of business.
>
>
> > How
> > much work would it be to get it forward ported along the lines of
> > the
> > other Rockchip SoC code?
>
> After getting some output on the console, you would work your way up.
> Clocksource/clockeven is still the dw_apb_timer (already supported).
> The
> hardest part will probably be getting a real clock-tree from the
> clock.c in
> the second git repo you linked.
>
I don't know, opening up the device and finding the serial port header
and hooking up and matching ttl levels might be tricky... (I'm hoping
to avoid this)
I guess I'll have to take a look at the dts code. I've never looked
in there before!
As Naoki has now pointed out, I should be using:
https://github.com/linux-rockchip/kernel_rockchip
It seems to have all the code in there. I guess it should just be a
matter of importing/updating the drivers and converting the board
configs to device-tree entires. Hopefully the new drm driver can
support all Rockchip SoCs.
> Next step would the pinctrl support for it.
>
> Most peripherals (like the i2c controllers) also have evolved in
> some way
> through the generations, so you'll need to check what needs to be
> adapted.
I'm _less_ worried about this. It's not usually too hard to forward
port drivers where the API has been around for a while, usually it's
just a matter of applying similar changes in areas that need updating
to those that have been applied to drivers already in the tree...
Usually... :-)
>
> And of course there is currently no display support available at all
> [with
> some initial drm patches for the rk3288 currently pending].
Well, hopefully it's not too different to the rk2918 in that area...
I need to clarify some things:
The Rockchip SoC provides the basic framebuffer crtc, colourspace
conversion, overlays and output connections (common to all Rockchip
SoCs), separate from the GPU (Vivate/Mali), and again seperate from
the video decoder and de-interlacing hw. Correct? I'm trying to get
my head around this. So the GPU specific code, in my case, the
"etnaviv" Gallium3D driver is the only GPU specific part just needs to
hook up to the kernel interface.
The previously mentioned repo has an fbdev driver so hopefully that's
enough to get the screen to light up:
https://github.com/linux-rockchip/kernel_rockchip/blob/rockchip-3.0-stable/drivers/video/rockchip/rk_fb.c
At least if I can get something on the screen it will make it easier
to test the other drivers until the new drm/kms driver is ready.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 181 bytes
Desc: This is a digitally signed message part
URL: <http://lists.infradead.org/pipermail/linux-rockchip/attachments/20141004/8eb638db/attachment.sig>
More information about the Linux-rockchip
mailing list