Hi Laurent,
On Tue, Aug 20, 2019 at 7:41 PM Laurent Pinchart laurent.pinchart@ideasonboard.com wrote:
On Tue, Aug 20, 2019 at 09:53:44AM +0200, Geert Uytterhoeven wrote:
On Tue, Aug 20, 2019 at 9:47 AM Jacopo Mondi jacopo@jmondi.org wrote:
On Mon, Aug 19, 2019 at 03:45:54PM +0200, Geert Uytterhoeven wrote:
On Mon, Jul 8, 2019 at 9:58 AM Geert Uytterhoeven geert@linux-m68k.org wrote:
On Sat, Jul 6, 2019 at 4:07 PM Jacopo Mondi jacopo+renesas@jmondi.org wrote:
Add device tree bindings documentation for the Renesas R-Car Display Unit Color Management Module.
CMM is the image enhancement module available on each R-Car DU video channel on R-Car Gen2 and Gen3 SoCs (V3H and V3M excluded).
Signed-off-by: Jacopo Mondi jacopo+renesas@jmondi.org Reviewed-by: Laurent Pinchart laurent.pinchart@ideasonboard.com
Thanks for your patch!
--- /dev/null +++ b/Documentation/devicetree/bindings/display/renesas,cmm.txt @@ -0,0 +1,25 @@ +* Renesas R-Car Color Management Module (CMM)
+Renesas R-Car image enhancement module connected to R-Car DU video channels.
+Required properties:
- compatible: shall be one of:
- "renesas,rcar-gen3-cmm"
- "renesas,rcar-gen2-cmm"
Why do you think you do not need SoC-specific compatible values? What if you discover a different across the R-Car Gen3 line tomorrow? Does the IP block have a version register?
Do you have an answer to these questions?
It does not seem to me that CMM has any version register, nor there are differences between the different Gen3 SoCs..
However, even if we now define a single compatible property for gen3/gen2 and we later find out one of the SoC needs a soc-specific property we can safely add it and keep the generic gen3/gen2 one as fallback.. Does it work for you?
Unfortunately that won't work, as the existing DTBs won't have the soc-specific compatible value. You could still resort to soc_device_match(), but it is better to avoid that.
We've had the same discussion over and over for quite a long time :-) I wonder, now that we have implemented SoC-specific compatible values for many IP cores, how many of them have actually benefited from it ? I'm not considering IP cores where we knew from the start that each SoC was different (such as pinctrl or clocks for instance), but IP cores where we though all SoCs would be handled in the same way. I also wouldn't count ES-specific differences, as those are handled by soc_device_match() anyway.
Thank you for making me dive into this ;-)
For R-Car Gen3 only:
DRIF? The driver still matches against "renesas,rcar-gen3-drif", but we found a difference on M3-N (didn't check if that was documented initially or not).
PCIe? IIRC, there is still a special PHY register on one of the V3x SoCs that the driver doesn't handle yet, and wasn't documented initially.
RPC-IF? Lots of small differences (you can claim they were documented, but they were unexpected), and non-documented different magic values in the (not yet upstreamed) driver.
SOUND? R-Car E3 special handling was bolted on later.
Thermal? M3-W turned out to have a different Tj than the other SoCs using the same thermal module (yeah, thermal is a bad example, as some Gen3 SoCs use the Gen2 thermal module, so we needed to differentiate anyway).
USBHS? Initially we thought Gen3 was identical to Gen2. Later it turned out that (1) wasn't true, and (2) E3/D3 used a different PLL than the other Gen3 SoCs.
VIN? IIRC, we initialize thought a family-specific compatible value would work for Gen3, like it did for Gen2, but had to reconsider.
I may have missed some...
Convinced?
Gr{oetje,eeting}s,
Geert