Re: [PATCH DNI v3 04/15] arm64: renesas: r8a78000: Add FCPV[BI] nodes

[Geert Uytterhoeven]

	Hi Laurent,

On Fri, 30 Oct 2026 at 23:12:25 +0200, Laurent Pinchart wrote:
> On Thu, Oct 15, 2026 at 04:51:44PM +0900, Paul Elder wrote:
> > Add DT nodes for the FCPVB and FCPVI IP cores in the R-Car X5H SoC.
> > 
> > Signed-off-by: Laurent Pinchart <laurent.pinchart+renesas@xxxxxxxxxxxxxxxx>
> > Signed-off-by: Paul Elder <paul.elder+renesas@xxxxxxxxxxxxxxxx>
> > Reviewed-by: Laurent Pinchart <laurent.pinchart+renesas@xxxxxxxxxxxxxxxx>
> 
> I need to reconsider this...
> 
> > Tested-by: Laurent Pinchart <laurent.pinchart+renesas@xxxxxxxxxxxxxxxx>
> > ---
> >  arch/arm64/boot/dts/renesas/r8a78000.dtsi | 72 +++++++++++++++++++++++++++++++
> >  1 file changed, 72 insertions(+)
> > 
> > diff --git a/arch/arm64/boot/dts/renesas/r8a78000.dtsi b/arch/arm64/boot/dts/renesas/r8a78000.dtsi
> > index 9a07753b221541ed49a975e0c95f2f5d1fdc6ac2..2f8886f8b6545026397fd83969882f8c62afa9d3 100644
> > --- a/arch/arm64/boot/dts/renesas/r8a78000.dtsi
> > +++ b/arch/arm64/boot/dts/renesas/r8a78000.dtsi
> > @@ -1000,6 +1000,78 @@ mdlc_aon: system-controller@c1338000 {
> >  			bootph-all;
> >  		};
> >  
> > +		fcpvi0: fcp@c5030000 {
> > +			compatible = "renesas,fcpv";
> > +			reg = <0 0xc5030000 0 0x1000>;
> > +			clocks = <&cpg R8A78000_CPG_SGD4_VIO_DP_TX>;
> 
> ... as this may not be the correct clock.
> 
> The user manual lists the following SGD clocks for the media-related IP
> cores:
> 
> - CSI[0123]:		SGD1_VIO_CSI[0123]φ, SGD2_VIO_CSI[0123]φ,
> 			SGD4_VIO_CSI[0123]φ, SGD16_VIO_CSI[0123]φ
> 
> - VIN[012]:		SGD1_VIO_CSI0φ, SGD4_VIO_CSI0φ
> - VIN[345]:		SGD1_VIO_CSI1φ, SGD4_VIO_CSI1φ
> - VIN[678]:		SGD1_VIO_CSI2φ, SGD4_VIO_CSI2φ
> - VIN{9,10,11}:		SGD1_VIO_CSI3φ, SGD4_VIO_CSI3φ
> 
> - ISP CS[0123]:		SGD1_VIO_CSI[0123]φ, SGD2_VIO_CSI[0123]φ
> - ISP Core[0123]:	SGD1_VIO_ISP[0123]φ, SGD2_VIO_ISP[0123]φ
> 
> - FCPV[BI][0-9]:	SGD1_VIO_DP_TXφ, SGD4_VIO_DP_TXφ
> - FCPVD[ab][0-9]:	SGD1_VIO_DP_TXφ, SGD4_VIO_DP_TXφ
> - FCPVX[0123]:		SGD1_VIO_ISP[0123]φ, SGD4_VIO_ISP[0123]φ
> 
> - VSP[BI][0-9]:		SGD1_VIO_DP_OTHERφ, SGD4_VIO_DP_OTHERφ
> - VSPD[ab][0-9]:	SGD1_VIO_DP_TXφ, SGD4_VIO_DP_TXφ
> - VSPX[0123]:		SGD1_VIO_ISP[0123]φ, SGD4_VIO_ISP[0123]φ
> 
> - VCON[0123]:		SGD1_VIO_DP_OTHER0φ, SGD4_VIO_DP_OTHER0φ
> - VCON[4567]:		SGD1_VIO_DP_OTHER1φ, SGD4_VIO_DP_OTHER2φ
> - VCON[89]:		SGD1_VIO_DP_OTHER2φ, SGD4_VIO_DP_OTHER2φ
> 
> - DPTX[012]:		SGD2_VIO_DP_TXφ, SGD16_VIO_DP_TXφ
> 
> Those IP cores are connected in four types of pipelines:
> 
> - FCPV[BI] -> VSP[BI]
> - FCPVD -> VSPD -> VCON -> DPTX
> - FCPVX -> VSPX -> ISP Core
> - CSI -> VIN -> ISP CS
> 
> It would be quite logical for all IP cores in the same pipeline to use
> the same clock. This would be
> 
> - FCPV[BI] -> VSP[BI]:			SGD*_VIO_DP_OTHERφ
> - FCPVD -> VSPD -> VCON -> DPTX:	SGD*_VIO_DP_TXφ
> - FCPVX -> VSPX -> ISP Core:		SGD*_VIO_ISP[0123]φ
> - CSI -> VIN -> ISP CS:			SGD*_VIO_CSI[0123]φ
> 
> This mostly matches the clocks listed in the user manual, except for
> 
> - FCPV[BI] that is listed as using SGD*_VIO_DP_TXφ
> - VCON that is listed as using SGD*_VIO_DP_OTHER[012]φ
> 
> Now, note how in the clock diagram document the VIO "ISP" and "CSI"
> hierarchies both show 4 groups of SGD clocks (respectively named
> SGD*_VIO_ISP0 and SGD*_VIO_CSI0 in the first group), the VIO "DP"
> hierarchies show 3 groups of SGD clocks (named SGD*_VIO_DP0 in the first
> group) and the VIO "DP_OTHER" hierarchy shows a single group of SGD
> clocks (named SGD*_VIO_DP_OTHER).
> 
> This makes me believe that, wherever the user manual lists
> SGD*_VIO_DP_TXφ, it actually means SGD*_VIO_DP[012]φ. Similarly, where
> the user manual lists SGD*_VIO_DP_OTHER[012]φ for VCON, I believe it
> should be SGD*_VIO_DP[012]φ, both because there is a single
> SGD*_VIO_DP_OTHERφ, and because the rest of the VCON pipelines (FCPVD,
> VSPD and DPTX) all use SGD*_VIO_DP[012]φ.

The clock diagram has 3 VIO_DP[012] hierarchies and a single
VIO_DP_OTHER, while the clock list has 3 VIO_DP_OTHER[012] hierarchies
and a single VIO_DP_TX.
The clock diagram has VIO_MAIN (system bus) clocks, while the clock list
has VIO_OTHER and VIO_BUS clocks.

> For full consistency, I also suppose that the FCPV[BI][0-9] instances
> match the corresponding VSP[BI][0-9] and use the SGD*_VIO_DP_OTHERφ
> clocks, not the listed SGD*_VIO_DP_TXφ clocks.
> 
> This being said, in practice this probably makes no difference, as all
> those clocks derive from the same SGD1_VIO_MAIN clock and none of them
> are programmable or gateable.

Probably, but it would be good to have that confirmed.

> If the above explanation makes sense to you (and to Geert), we can ask
> Renesas to check if the issue I suspect are errors in the documentation.

Agreed.

Current SCMI firmware does not have the CSI3 group.  It does have
VIO_IM[012], which was replaced by DP_OTHER{012] according to the
"Differences" sheet in the clock list.

I guess we really need to know if e.g. all these SGD* clocks are the
same, and if they can be gated individually.  Same for the various DIV6
clocks, as there is only a single register to control them (which
probably controls the parent/singleton in reality).

Gr{oetje,eeting}s,

						Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
							    -- Linus Torvalds