On Monday 16 December 2013, Sergei Ianovich wrote:
> PXA27x memory bus can have up to 10 devices: up to 6 slower
> flash/SRAM/variable-latency-IO selected by nCS<0> to <5>, and up to 4
> partions of SDRAM selected by nSDCS<0> to <3>.
>
> It appears that the FPGA is directly connected to the memory bus and is
> selected by nCS<5>. According to MSC2 configuration (already in the
> mainstream U-Boot), the FPGA is configured as a synchronous SRAM with
> access cycle of 30x memory bus cycles. So I made it a top-level bus like
> pxabus.
>
> Ethernet devices are also connected to the memory bus via some kind of
> gate array. This one is a bit faster -- 15x memory cycles, which is
> still a lot. It explains why network transfers are never faster than 15
> Mbit/s.
Ok, I see. This sounds like some of the other platforms we have with
external memory buses. If there is a chance that Linux ever has to
program the per-CS settings into the bus controller, I would suggest
to represent that as well as a separate node, like this
extbus {
compatible = "simple-bus";
#address-cells = <2>; /* first cell is nCS, second is address */
#size-cells = <1>;
ranges = <0 0 0 0x04000000>
<1 0 0x04000000 0x04000000>
<2 0 0x08000000 0x04000000>
<3 0 0x0c000000 0x04000000>
<4 0 0x10000000 0x04000000>
<5 0 0x14000000 0x04000000>;
timings = ...;
flash@0 {
reg = <0 0x0 0x02000000>;
...
};
flash@1 {
reg = <1 0x0 0x02000000>;
...
};
fpga@5 {
#address-cells = <1>;
#size-cells = <1>;
ranges = <0 5 0x03000000 0x10000>;
...
};
};
In case there is a driver for the extbus node, I would not make it
"simple-bus" but instead add a separate compatible string to match
the driver, and let that driver call of_platform_populate
to probe the children after the bus is set up.
Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majord...@vger.kernel.org
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
