2014-07-03 12:37:36

by Jean-Michel Hautbois

Subject: Lattice ECP3 FPGA with i.MX6

Hi,

I have a board, with a Freescale i.MX6 chip and a ECP3-35 FPGA on SPI.
I tried to load the firmware using the lattice-ecp3-config driver, but
it fails with this error :

lattice-ecp3 spi32766.3: FPGA bitstream configuration driver registered
lattice-ecp3 spi32766.3: Error: No supported FPGA detected (JEDEC_ID=808004c2)!

In the driver, the id is :

#define ID_ECP3_35 0xc2048080

Obviously, there is a big/little endian issue... Do I need to instruct
the device-tree in a specific way in order to get the bus in the
correct order ? Or is this a known issue maybe ?

Thanks,
JM

2014-07-03 12:55:29

by Stefan Roese

Subject: Re: Lattice ECP3 FPGA with i.MX6

On 03.07.2014 14:37, Jean-Michel Hautbois wrote:
> I have a board, with a Freescale i.MX6 chip and a ECP3-35 FPGA on SPI.
> I tried to load the firmware using the lattice-ecp3-config driver, but
> it fails with this error :
>
> lattice-ecp3 spi32766.3: FPGA bitstream configuration driver registered
> lattice-ecp3 spi32766.3: Error: No supported FPGA detected (JEDEC_ID=808004c2)!
>
> In the driver, the id is :
>
> #define ID_ECP3_35 0xc2048080
>
> Obviously, there is a big/little endian issue... Do I need to instruct
> the device-tree in a specific way in order to get the bus in the
> correct order ? Or is this a known issue maybe ?

No. This driver was implemented and tested in a MPC5200 system. Most
likely I missed some endian issues as you already noticed. I suggest you
start with looking at this line:

jedec_id = *(u32 *)&rxbuf[4];

And add some endian functions here, e.g. be32_to_cpu(). This might help
with the detection. But other endian related issues might still be
present in other parts of the driver as well.

HTP.

Thanks,
Stefan

2014-07-03 15:06:04

by Jean-Michel Hautbois

Subject: Re: Lattice ECP3 FPGA with i.MX6

> No. This driver was implemented and tested in a MPC5200 system. Most likely
> I missed some endian issues as you already noticed. I suggest you start with
> looking at this line:
>
> jedec_id = *(u32 *)&rxbuf[4];
>
> And add some endian functions here, e.g. be32_to_cpu(). This might help with
> the detection. But other endian related issues might still be present in
> other parts of the driver as well.

Replacing this line with the following works better :
jedec_id = be32_to_cpu(*(u32 *)&rxbuf[4]);

At least, on detecting the FPGA...
I will send a patch, but there is only three places where there is a
be32_to_cpu() to add.