Den 08.07.2020 14.49, skrev Paul Cercueil:
> Hi Daniel,
>
> Le mer. 8 juil. 2020 à 9:23, Daniel Vetter <daniel@xxxxxxxx> a écrit :
>> On Tue, Jul 07, 2020 at 04:32:25PM +0200, Noralf Trønnes wrote:
>>> (cc Dillon)
>>>
>>> Den 03.07.2020 19.26, skrev Sam Ravnborg:
>>> > Hi Noralf/Paul.
>>> >
>>> > Trying to stir up this discussion again.
>>> >
>>> > On Sun, Jun 14, 2020 at 06:36:22PM +0200, Noralf Trønnes wrote:
>>> >>
>>> >>
>>> >> Den 07.06.2020 15.38, skrev Paul Cercueil:
>>> >>> Hi,
>>> >>>
>>> >>> Here's a follow-up on the previous discussion about the current
>>> state of
>>> >>> DSI/DBI panel drivers, TinyDRM, and the need of a cleanup.
>>> >>>
>>> >>> This patchset introduces the following:
>>> >>> * It slightly tweaks the MIPI DSI code so that it supports MIPI
>>> DBI over
>>> >>> various buses. This patch has been tested with a non-upstream DRM
>>> >>> panel driver for a ILI9331 DBI/8080 panel, written with the DSI
>>> >>> framework (and doesn't include <drm/drm_mipi_dbi.h>), and
>>> non-upstream
>>> >>> DSI/DBI host driver for the Ingenic SoCs.
>>> >>>
>>> >>> * A SPI DBI host driver, using the current MIPI DSI framework.
>>> It allows
>>> >>> MIPI DSI/DBI drivers to be written with the DSI framework,
>>> even if
>>> >>> they are connected over SPI, instead of registering as SPI device
>>> >>> drivers. Since most of these panels can be connected over various
>>> >>> buses, it permits to reuse the same driver independently of
>>> the bus
>>> >>> used.
>>> >>>
>>> >>> * A TinyDRM driver for DSI/DBI panels, once again independent of
>>> the bus
>>> >>> used; the only dependency (currently) being that the panel must
>>> >>> understand DCS commands.
>>> >>>
>>> >>> * A DRM panel driver to test the stack. This driver controls Ilitek
>>> >>> ILI9341 based DBI panels, like the Adafruit YX240QV29-T
>>> 320x240 2.4"
>>> >>> TFT LCD panel. This panel was converted from
>>> >>> drivers/gpu/drm/tiny/ili9341.c.
>>> >>>
>>> >>> I would like to emphasize that while it has been compile-tested,
>>> I did
>>> >>> not test it with real hardware since I do not have any DBI panel
>>> >>> connected over SPI. I did runtime-test the code, just without
>>> any panel
>>> >>> connected.
>>> >>>
>>> >>> Another thing to note, is that it does not break Device Tree
>>> ABI. The
>>> >>> display node stays the same:
>>> >>>
>>> >>> display@0 {
>>> >>> compatible = "adafruit,yx240qv29", "ilitek,ili9341";
>>> >>> reg = <0>;
>>> >>> spi-max-frequency = <32000000>;
>>> >>> dc-gpios = <&gpio0 9 GPIO_ACTIVE_HIGH>;
>>> >>> reset-gpios = <&gpio0 8 GPIO_ACTIVE_HIGH>;
>>> >>> rotation = <270>;
>>> >>> backlight = <&backlight>;
>>> >>> };
>>> >>>
>>> >>> The reason it works, is that the "adafruit,yx240qv29" device is
>>> probed
>>> >>> on the SPI bus, so it will match with the SPI/DBI host driver.
>>> This will
>>> >>> in turn register the very same node with the DSI bus, and the
>>> ILI9341
>>> >>> DRM panel driver will probe. The driver will detect that no
>>> controller
>>> >>> is linked to the panel, and eventually register the DBI/DSI TinyDRM
>>> >>> driver.
>>> >>>
>>> >>> I can't stress it enough that this is a RFC, so it still has
>>> very rough
>>> >>> edges.
>>> >>>
>>> >>
>>> >> I don't know bridge and dsi drivers so I can't comment on that,
>>> but one
>>> >> thing I didn't like is that the DT compatible string has to be
>>> added to
>>> >> 2 different modules.
>>> >>
>>> >> As an example, a MI0283QT panel (ILI9341) supports these
>>> interface options:
>>> >>
>>> >> 1. SPI
>>> >> Panel setup/control and framebuffer upload over SPI
>>> >>
>>> >> 2. SPI + DPI
>>> >> Panel setup/control over SPI, framebuffer scanout over DPI
>>> >>
>>> >> 3. Parallel bus
>>> >> Panel setup/control and framebuffer upload over parallel bus
>>> >
>>> > To continue the configurations we should support:
>>> > - Panels where the chip can be configured to SPI, SPI+DPI,
>>> Parallel bus
>>> > (as detailed by Noralf above)
>>> > - Panels that supports only 6800 or 8080 - connected via GPIO pins or
>>> > memory mapped (maybe behind some special IP to support this)
>>> > Command set is often special.
>>> >
>>> > We will see a number of chips with many different types of displays.
>>> > So the drivers should be chip specific with configuration
>>> depending on
>>> > the connected display.
>>> >
>>> > What I hope we can find a solution for is a single file/driver
>>> that can
>>> > support all the relevant interface types for a chip.
>>> > So we would end up with a single file that included the necessary
>>> > support for ili9341 in all interface configurations with the
>>> necessary
>>> > support for the relevant displays.
>>> >
>>> > I do not know how far we are from this as I have not dived into the
>>> > details of any of the proposals.
>>>
>>> In an ideal world I would have liked to see the MIPI DBI parallel
>>> interface implemented using a new Linux parallel bus type. It could
>>> have
>>> drivers for gpio bitbanging and mmio in addition to other hw specific
>>> drivers. Now we could have a drm_mipi_dbi DRM driver that registers
>>> as a
>>> SPI client driver and a Parallel bus client driver. Or it can be a
>>> component driver for the existing DRM driver on the SoC.
I have realised that there is a problem here. A drm_mipi_drm driver
would require a device and this device would be virtual since the actual
device is the panel itself. IIRC we can't have virtual devices in the
Device Tree. So this won't work I think.
The same problem applies to my mipi-dbi-spi case below, it is also a
virtual device.
>>>
>>> I had plans to do this and made a prototype, but dropped it since it
>>> would probably require a lot of work getting in a new Linux bus type.
>>
>> Channelling my inner Greg KH:
>>
>> Please just create a new bus, it should be quite easy and boilerplate is
>> manageable.
I'm not sure I agree with that. Wouldn't this be something akin to SPI
or I2C? There's quite some code in those subsystems.
>
> The bus is already here, it's "mipi-dsi". DBI and DSI are basically the
> same thing, just that one is parallel and the other is serial.
Can the parallel interface on your hw be used outside the display domain
as say connected to an ADC? If not then a new Linux bus type won't help
in your case.
My concern is tying a parallel interface to the DRM subsystem if it can
be used as a generic interface. The Raspberry Pi has a SMI (Secondary
Memory Interface) parallel interface that can be used for DBI and I
believe the Beaglebone Black can have a hw bus through the realtime
unit. But I have no experience with either of these.
Noralf.
>
> -Paul
>
>> Greg, did I get this right? Maybe any recommendations for a simple
>> parallel bus with perhaps different register access paths depending upon
>> how it's all wired up exactly?
>> -Daniel
>>
>>> However if we're going to treat this parallel bus only as a MIPI DBI
>>> display interface but support gpio bitbanging and mmio as well, then we
>>> could add DRM drivers for each MIPI DBI bus (that don't have special
>>> parallel bus hw):
>>> - mipi-dbi-spi
>>> - mipi-dbi-gpio
>>> - mipi-dbi-mmio
>>>
>>> These drivers will register as a mipi_dsi_host adapted like Paul
>>> suggested.
>>>
>>> The panel drivers will be mipi_dsi_drivers. Now the panels should work
>>> regardless of bus type. They probably need to know about the bus type,
>>> at least whether the parallell bus is 8-bit or 16-bit wide.
>>>
>>> The current MIPI DBI SPI drivers (drm/tiny) will need to be treated
>>> specially to keep working with old Device Trees when moved over to
>>> drm/panel.
>>>
>>> Noralf.
>>>
>>>
>>> >>
>>> >> My suggestion is to have one panel driver module that can support
>>> all of
>>> >> these like this:
>>> > So I think we agree here.
>>> >
>>> >>
>>> >> For 1. and 2. a SPI driver is registered and if I understand your
>>> >> example correctly of_graph_get_port_by_id() can be used during
>>> probe to
>>> >> distinguish between the 2 options and register a full DRM driver
>>> for 1.
>>> >> and add a DRM panel for 2.
>>> >>
>>> >> For 3. a DSI driver is registered (adapted for DBI use like you're
>>> >> suggesting).
>>> >>
>>> >> Note that the interface part of the controller initialization will
>>> >> differ between these, the panel side init will be the same I assume.
>>> >
>>> > Sam
>>> >
>>
>> --
>> Daniel Vetter
>> Software Engineer, Intel Corporation
>> http://blog.ffwll.ch
>
>
>
_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/dri-devel
