[RFC 00/13] arm64: allwinner: Add A64 DE2 pipeline support

[Jernej Škrabec]

Hi,

Dne sreda, 25. april 2018 ob 12:34:09 CEST je Jagan Teki napisal(a):
> On Tue, Apr 24, 2018 at 9:02 PM, Jernej Škrabec <jernej.skrabec at siol.net> 
wrote:
> > Hi,
> > 
> > Dne torek, 24. april 2018 ob 15:34:12 CEST je Jagan Teki napisal(a):
> >> Allwinner A64 has display engine pipeline like other Allwinner SOC's
> >> A83T/H3/H5.
> >> 
> >> A64 DE2 behaviour similar to Allwinner A83T where mixer0, connected to
> >> tcon0 with RGB, LVDS MIPI-DSI and mixer1, connected to tcon1 with HDMI.
> >> This series merely concentrated on HDMI pipeline and rest will add
> >> eventually.
> >> 
> >> patch 1: dt-bindings for a64 DE2 CCU
> >> 
> >> patch 2: a64 DE2 CCU node addition
> >> 
> >> patch 3: dt-bindings for a64 DE2 pipeline
> >> 
> >> patch 4 - 5: dt-bindings for a64 mixer0 and tcon-lcd
> >> 
> >> patch 6: a64 DE2 pipeline node addition
> >> 
> >> patch 7 - 8: dt-bindings for a64 HDMI and HDMI PHY
> >> 
> >> patch 9: a64 HDMI nodes addition
> >> 
> >> patch 10 - 11: dt-bindings for a64 mixer1 and tcon-tv
> >> 
> >> patch 12: a64 HDMI pipeline
> >> 
> >> patch 13: enable HDMI out on bananpi-m64
> >> 
> >> Tested HDMI on bananapi-m64 (along with DE2 SRAM C changes from [1]
> >> thread), able to detect the HDMI but, no penguins on screen.
> >> 
> >> Request for any suggestions.
> > 
> > You are mising sunxi-ng clock patches. PLL_VIDEO0 and PLL_VIDEO1 need
> > fixes by setting minimum stable frequency. Please note that datasheet may
> > have wrong information. That was obvious in H3 case and I had to check
> > minimum stable PLL_VIDEO frequency in BSP driver.
> 
> Can you point me the clock patches? 

Here is my A64 HDMI wip repo, which includes my clock changes:
https://github.com/jernejsk/linux-1/tree/a64_hdmi_wip

At some point HDMI output worked ok, I'm not sure in what state I left the 
code.

> I've phadled only CLK_PLL_VIDEO0
> on hdmi_phy So we need CLK_PLL_VIDEO1 as fourth clock?

I'm not sure what is the best way. I guess some research is needed. There is 
even the possibility that one bit (SUN8I_HDMI_PHY_PLL_CFG1_CKIN_SEL) in hdmi 
phy needs to be toggled depending on which clock is selected as HDMI source. I 
never finished my research since I'm waiting to SRAM C claiming patches.

At the end, HDMI driver should be tested by both, PLL_VIDEO0 and PLL_VIDEO1 as 
a HDMI clock source, otherwise there is no guarantee that we got binding 
right. A83T, H3 and H5 has only one possible HDMI parent clock, so it was much 
easier in this regard.

Unfortunately, BSP driver is not much of a help, since it was always 
distributed as a blob. Disassembly looks much like H3 driver, with few bits 
added. However, BSP clocks are pretty much hardcoded so they don't tell all 
the story.

Best regards,
Jernej