[PATCH 0/8] Reconfigurable FPGAs over PCIe.

[Grant Likely]

On Fri, Jul 16, 2010 at 12:13 PM, Stephen Neuendorffer
<stephen.neuendorffer at xilinx.com> wrote:
> This sequence of patches has the goal of enabling an FPGA connected
> over PCIe to have multiple internal devices specified in a device
> tree..  Of course, such a device should work in any machine, in
> addition to those architectures which use device trees to describe the
> basic platform functionality: In particular, x86.  This requires some
> basic support on x86 in order to make device trees work, and small
> amount of driver code to light everything up.
>
> This has been tested on a Virtex5 board plugged into the PCIe slot of
> an x86 machine.  The device tree used looks like:
>
> /dts-v1/;
> / {
>        #address-cells = <1>;
>        #size-cells = <1>;
>   compatible = "simple-bus";
>        mb_plb: plb at 0 {
>                #address-cells = <1>;
>                #size-cells = <1>;
>                compatible = "xlnx,plb-v46-1.03.a", "xlnx,plb-v46-1.00.a", "simple-bus";
>                xps_hwicap_0: xps-hwicap at 80030000 {
>                        compatible = "xlnx,xps-hwicap-1.00.a";
>                        reg = < 0x80030000 0x10000 >;
>                        xlnx,family = "virtex5";
>                        xlnx,simulation = "false";
>                } ;
>   } ;
> }  ;
>
> Note that the bulk of the code is work in progress.  In particular,
> exactly how the device tree support should get extended to X86 is
> still up for grabs.  In addition, there are certain aspects of the
> device tree used which are suboptimal: The PCIe bridge translation
> range is currently hardcoded into the driver, and there is no
> provision for connecting interrupt lines into the PCIe bridge.
> Comments greatly appreciated...
>
> Grant: I've thrown in two hwicap patches, which should be final, if
> you want to pick them up that would be great.

Okay, I'll take a look.

g.