Message-ID: <20191120164548.GA168745@aptenodytes>
Date:   Wed, 20 Nov 2019 17:45:48 +0100
From:   Paul Kocialkowski <paul.kocialkowski@...tlin.com>
To:     Rob Herring <robh@...nel.org>
Cc:     dri-devel <dri-devel@...ts.freedesktop.org>,
        devicetree@...r.kernel.org,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        David Airlie <airlied@...ux.ie>,
        Daniel Vetter <daniel@...ll.ch>,
        Mark Rutland <mark.rutland@....com>,
        Maarten Lankhorst <maarten.lankhorst@...ux.intel.com>,
        Maxime Ripard <mripard@...nel.org>, Sean Paul <sean@...rly.run>
Subject: Re: [PATCH 1/2] dt-bindings: display: Add xylon logicvc bindings
 documentation

Hi,

On Wed 20 Nov 19, 10:02, Rob Herring wrote:
> On Wed, Nov 20, 2019 at 8:50 AM Paul Kocialkowski
> <paul.kocialkowski@...tlin.com> wrote:
> >
> > Hi,
> >
> > Circling back to this thread now, sorry for the delay.
> >
> > On Tue 24 Sep 19, 09:58, Rob Herring wrote:
> > > On Mon, Sep 23, 2019 at 10:33 AM Paul Kocialkowski
> > > <paul.kocialkowski@...tlin.com> wrote:
> > > >
> > > > Hi,
> > > >
> > > > On Fri 13 Sep 19, 20:16, Rob Herring wrote:
> > > > > On Fri, Sep 13, 2019 at 4:58 PM Paul Kocialkowski
> > > > > <paul.kocialkowski@...tlin.com> wrote:
> > > > > >
> > > > > > Hi Rob and thanks for the review!
> > > > > >
> > > > > > On Fri 13 Sep 19, 15:35, Rob Herring wrote:
> > > > > > > On Tue, Sep 10, 2019 at 05:34:08PM +0200, Paul Kocialkowski wrote:
> > > > > > > > The Xylon LogiCVC is a display controller implemented as programmable
> > > > > > > > logic in Xilinx FPGAs.
> > > > > > > >
> > > > > > > > Signed-off-by: Paul Kocialkowski <paul.kocialkowski@...tlin.com>
> > > > > > > > ---
> > > > > > > >  .../bindings/display/xylon,logicvc.txt        | 188 ++++++++++++++++++
> > > > > > > >  1 file changed, 188 insertions(+)
> > > > > > > >  create mode 100644 Documentation/devicetree/bindings/display/xylon,logicvc.txt
> > > > > > >
> > > > > > > Consider converting this to DT schema format. See
> > > > > > > Documentation/devicetree/writing-schema.rst (.md in 5.3).
> > > > > >
> > > > > > Oh right, that would certainly be much more future-proof!
> > > > > >
> > > > > > > > diff --git a/Documentation/devicetree/bindings/display/xylon,logicvc.txt b/Documentation/devicetree/bindings/display/xylon,logicvc.txt
> > > > > > > > new file mode 100644
> > > > > > > > index 000000000000..eb4b1553888a
> > > > > > > > --- /dev/null
> > > > > > > > +++ b/Documentation/devicetree/bindings/display/xylon,logicvc.txt
> > > > > > > > @@ -0,0 +1,188 @@
> > > > > > > > +Xylon LogiCVC display controller
> > > > > > > > +
> > > > > > > > +The Xylon LogiCVC is a display controller that supports multiple layers.
> > > > > > > > +It is usually implemented as programmable logic and was optimized for use
> > > > > > > > +with Xilinx Zynq-7000 SoCs and Xilinx FPGAs.
> > > > > > > > +
> > > > > > > > +Because the controller is intended for use in a FPGA, most of the configuration
> > > > > > > > +of the controller takes place at logic configuration bitstream synthesis time.
> > > > > > > > +As a result, many of the device-tree bindings are meant to reflect the
> > > > > > > > +synthesis configuration. These do not allow configuring the controller
> > > > > > > > +differently than synthesis configuration.
> > > > > > > > +
> > > > > > > > +Layers are declared in the "layers" sub-node and have dedicated configuration.
> > > > > > > > +In version 3 of the controller, each layer has fixed memory offset and address
> > > > > > > > +starting from the video memory base address for its framebuffer. With version 4,
> > > > > > > > +framebuffers are configured with a direct memory address instead.
> > > > > > > > +
> > > > > > > > +Matching synthesis parameters are provided when applicable.
> > > > > > > > +
> > > > > > > > +Required properties:
> > > > > > > > +- compatible: Should be one of:
> > > > > > > > +  "xylon,logicvc-3.02.a-display"
> > > > > > > > +  "xylon,logicvc-4.01.a-display"
> > > > > > > > +- reg: Physical base address and size for the controller registers.
> > > > > > > > +- clocks: List of phandle and clock-specifier pairs, one for each entry
> > > > > > > > +  in 'clock-names'
> > > > > > > > +- clock-names: List of clock names that should at least contain:
> > > > > > > > +  - "vclk": The VCLK video clock input.
> > > > > > > > +- interrupts: The interrupt to use for VBLANK signaling.
> > > > > > > > +- xylon,display-interface: Display interface in use, should be one of:
> > > > > > > > +  - "lvds-4bits": 4-bit LVDS interface (C_DISPLAY_INTERFACE == 4).
> > > > > > > > +- xylon,display-colorspace: Display output colorspace in use, should be one of:
> > > > > > > > +  - "rgb": RGB colorspace (C_DISPLAY_COLOR_SPACE == 0).
> > > > > > > > +- xylon,display-depth: Display output depth in use (C_PIXEL_DATA_WIDTH).
> > > > > > > > +- xylon,row-stride: Fixed number of pixels in a framebuffer row (C_ROW_STRIDE).
> > > > > > > > +- xylon,layers-count: The number of available layers (C_NUM_OF_LAYERS).
> > > > > > >
> > > > > > > Presumably some of this is determined by the display attached. Isn't it
> > > > > > > safe to assume the IP was configured correctly for the intended display
> > > > > > > and you can just get this from the panel?
> > > > > >
> > > > > > Layers are what corresponds to DRM planes, which are not actually indicated
> > > > > > by the panel but are a charasteristic of the display controller. In our case,
> > > > > > this is directly selected at bitstream synthesis time for the controller.
> > > > > >
> > > > > > So I'm afraid there is no way we can auto-detect this from the driver.
> > > > >
> > > > > Sorry, I referring to the set of properties above. In particular,
> > > > > xylon,display-interface and xylon,display-colorspace, though I don't
> > > > > know if the latter is talking in memory format or on the wire format.
> > > >
> > > > Both of these are about the wire format, which is also "hardcoded" at synthesis
> > > > time with no way to be detected afterwards, as far as I know. Memory format is
> > > > described in the layer sub-nodes.
> > >
> > > You have to attach the controller to something at the other end of the
> > > wire. A panel is only going to support 1 or a few wire formats, so you
> > > do likely know because the panel knows. In the case that a panel
> > > supports multiple wire formats, we do have some standard properties
> > > there. See the LVDS panel binding.
> >
> > Looking at the LVDS panel binding, I see that the LVDS types that I have
> > described as lvds-4bits and lvds-3bits are called jeida-24 and jeida-18.
> >
> > Either way, the controller cannot be dynamically configured to use one or
> > another: it is configured to support one at synthesis time and this doesn't
> > change.
> 
> Understood, but I was assuming you need to know how it was configured
> for some reason?

Well, the information that is really useful for the driver is whether it's
configured as LVDS or something else. The detail of 3bit vs 4bit doesn't really
matter for Linux/DRM but I kept it to that precision to stick close to the
hardware description. But yeah, having a single "lvds" choice in the
display-interface property for these two cases would work too and I can do that
if you prefer.

> > I'm not sure exactly what you implied here. Even if we can retreive the
> > wire format from the lvds-panel's data-mapping property, I don't think it shall
> > describe what the display controller was configured to. This information could
> > be used to make sure that both are compatible (in the driver), but that's about
> > it as far as I can see.
> 
> It's not the kernel's job to validate the DT is correct. Someone could
> just as easily define a panel that doesn't match with the configured
> format as they could having lvds-?bits set incorrectly.

So this is a situation that we are supposed to allow?
I assumed it would be in everyone's best interest to detect that there is a
mismatch, but I don't have any strong opinion about that.

> So get the wire format from the panel driver (either implied or by DT
> property) and assume that matches the configuration of the controller.

To be honest, it feels a bit weird to depend on a panel being connected to know
what display interface the driver should register. It might work in practice
since these media bus formats are specific to LVDS, so it can be correctly
deduced, but I don't understand the advantage of doing that.

It will probably also require parsing the panel remote endpoint's dt props as
I don't think we can get the bus format information from DRM before having
registered an encoder/connector on the controller side (so that's a
chicken-and-egg problem).

More generally, I don't see the issue behind describing this in the controller's
bindings: is it because of unwanted redundency or such?

> Though, I guess if the model is each end of the wire should advertise
> what it supports and the core picks the best format, then that only
> works if you advertise both formats. Or we could allow jeida-{24,18}
> property at both ends of the graph.

I don't think there is such negotiation currently implemented in DRM
(and many driver actually have a strong assumption that only one bus format
is provided).

If we want the DRM core to deal with that, we need drivers to report their bus
format abilities for each concerned encoder/connector. I'm not sure this could
be done in a generic way with a generic dt property since the of_graph to
encoder/connector association seems to be quite driver-specific anyway.
This would just be deporting a description of capabilities of the hardware from
the driver to dt, with no particular gain as far as I can see.

But maybe I'm missing the bigger picture here and describing hardware
capabilities in device-tree is something we want to go towards. I must admit
that I find the boundary quite blurry.

Cheers,

Paul

-- 
Paul Kocialkowski, Bootlin
Embedded Linux and kernel engineering
https://bootlin.com

Download attachment "signature.asc" of type "application/pgp-signature" (489 bytes)

Powered by blists - more mailing lists