Re: [spi-devel-general] [RESEND][PATCH] SPI: xilinx_spi: Added platform driver and support for DS570

[David Brownell <david-b@pacbell.net>]

On Wednesday 24 June 2009, Richard Röjfors wrote:
> On 09-06-23 21.37, David Brownell wrote:
> > On Monday 15 June 2009, Richard Röjfors wrote:
> >> This patch splits xilinx_spi into three parts, an OF and a platform
> >> driver and generic part.
> >>
> >> The generic part now also works on X86 and also supports the Xilinx
> >> SPI IP DS570
> > 
> > Your Kconfig still mentions only DS464.
> 
> Correct, will update.

Just checking -- you didn't send an update yet, right?


> >> +	  See the "OPB Serial Peripheral Interface (SPI) (v1.00e)"
> >> +	  Product Specification document (DS464) for hardware details.
> >> +
> >> +config SPI_XILINX_PLTFM
> >> +	tristate "Xilinx SPI controller platform device"
> >> +	depends on SPI_XILINX
> > 
> > This should go first in the Kconfig, so that you don't goof up
> > the dependency display ... the OF bits depend on it.
> 
> The OF parts only depend on SPI_XILINX, not SPI_XILINX_PLTFM. But
> I can change the order anyway(?)

Maybe I misread something.  Make sure the Kconfig doesn't goof.


> >> +	/* devices to add to the bus when the host is up */
> >> +	struct spi_board_info *devices;
> >> +	u8 num_devices;
> > 
> > This is duplicating functionality in the SPI core code.
> > That's not really the way to fly.
> 
> I have discussed the same problem in the I2C list.
> 
> We have a PCIe device which is a MFD (multifunction device).
> 
> Several MFD devices could be connected to a standard PC ->
> we don't know the bus numbers in advance, and there could be
> several PCIe devices connected. Which mean the we can not use
> spi_register_board_info.

You could, actually, given metadata about those devices.
That's one of the things ACPI does -- in a way that many
of us find ugly! -- and the "device tree" some PPCs use.


> We can not call spi_add_device from the MFD because when it's
> probed the drivers for the SPI bus might not be available.

And you can't call spi_register_board_info() there since
the init section is gone.

 
> So the solution in the I2C case was to add this into the driver
> and if this tend to be a generic problem some generic solution might
> be implemented.
> 
> So I would like to have the same thing in here.

What I've done in such cases is make sure the bus adapter
driver (in this case, for the spi_master) can issue some
callback to code which makes spi_add_device(), or similar,
calls.

But if your initialization doesn't have that sort of
clean structure, it's probably kind of awkward to add
it just now...  :(

- Dave