Use xilinx video drivers in PCIe device

Use xilinx video drivers in PCIe device

[Franck Jullien]

Hi,

I need to use a Xilinx video infrastructure on a PCIe board.
As far as I understand it, all Xilinx video drivers make use of the
device-tree for configuration.
However, my idea is to create a MFD device to bind video drivers. That
would require Xilinx video drivers to check platform_data and continue
with device tree configuration if it is null or use platform data if
available.

Do you think such a change in Xilinx drivers can be considered
upstream ? Is this the way to go ?

Thanks in advance,

Franck.

Re: Use xilinx video drivers in PCIe device

[Laurent Pinchart]

Hi Frank,

On Tuesday 02 February 2016 17:05:06 Franck Jullien wrote:
> Hi,
> 
> I need to use a Xilinx video infrastructure on a PCIe board.
> As far as I understand it, all Xilinx video drivers make use of the
> device-tree for configuration.

Correct. Those drivers target the Xilinx SoC FPGAs, no standalone FPGAs 
connected to an external CPU.

> However, my idea is to create a MFD device to bind video drivers. That
> would require Xilinx video drivers to check platform_data and continue
> with device tree configuration if it is null or use platform data if
> available.
> 
> Do you think such a change in Xilinx drivers can be considered
> upstream ? Is this the way to go ?

Your use case is certainly valid, so I'm certainly open to supporting it in 
the drivers.

I'm wondering whether your MFD decide driver could create a DT fragment to 
describe the IP cores topology. That way we could reuse the existing DT 
support in individual drivers.

-- 
Regards,

Laurent Pinchart

Re: Use xilinx video drivers in PCIe device

[Franck Jullien]

2016-02-04 10:45 GMT+01:00 Laurent Pinchart <laurent.pinchart@ideasonboard.com>:
> Hi Frank,
>
> On Tuesday 02 February 2016 17:05:06 Franck Jullien wrote:
>> Hi,
>>
>> I need to use a Xilinx video infrastructure on a PCIe board.
>> As far as I understand it, all Xilinx video drivers make use of the
>> device-tree for configuration.
>
> Correct. Those drivers target the Xilinx SoC FPGAs, no standalone FPGAs
> connected to an external CPU.
>
>> However, my idea is to create a MFD device to bind video drivers. That
>> would require Xilinx video drivers to check platform_data and continue
>> with device tree configuration if it is null or use platform data if
>> available.
>>
>> Do you think such a change in Xilinx drivers can be considered
>> upstream ? Is this the way to go ?
>
> Your use case is certainly valid, so I'm certainly open to supporting it in
> the drivers.
>
> I'm wondering whether your MFD decide driver could create a DT fragment to
> describe the IP cores topology. That way we could reuse the existing DT
> support in individual drivers.
>

I'm working on such a solution (DT framgent, or more precisely
devitree dynamic feature).
I'll keep you informed whenever I get something usable.

Franck.

Re: Use xilinx video drivers in PCIe device

[Laurent Pinchart]

Hi Frank,

On Monday 08 February 2016 08:40:11 Franck Jullien wrote:
> 2016-02-04 10:45 GMT+01:00 Laurent Pinchart:
> > On Tuesday 02 February 2016 17:05:06 Franck Jullien wrote:
> >> Hi,
> >> 
> >> I need to use a Xilinx video infrastructure on a PCIe board.
> >> As far as I understand it, all Xilinx video drivers make use of the
> >> device-tree for configuration.
> > 
> > Correct. Those drivers target the Xilinx SoC FPGAs, no standalone FPGAs
> > connected to an external CPU.
> > 
> >> However, my idea is to create a MFD device to bind video drivers. That
> >> would require Xilinx video drivers to check platform_data and continue
> >> with device tree configuration if it is null or use platform data if
> >> available.
> >> 
> >> Do you think such a change in Xilinx drivers can be considered
> >> upstream ? Is this the way to go ?
> > 
> > Your use case is certainly valid, so I'm certainly open to supporting it
> > in the drivers.
> > 
> > I'm wondering whether your MFD decide driver could create a DT fragment to
> > describe the IP cores topology. That way we could reuse the existing DT
> > support in individual drivers.
> 
> I'm working on such a solution (DT framgent, or more precisely
> devitree dynamic feature).
> I'll keep you informed whenever I get something usable.

Thank you. If you're faced with design choices please feel free to pick our 
brains at any time without waiting for a complete implementation.

-- 
Regards,

Laurent Pinchart