Re: [cadence-spi] daisy chain

[Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>]

Hi Adrian,

CC Lukas for #daisy-chained-devices

On Fri, Jun 12, 2020 at 6:26 PM Adrian Fiergolski
<adrian.fiergolski@xxxxxxxxxxxxx> wrote:
> I have a daisy chain of three ltc2634 slaves (iio/dac/ltc2632.c)
> connected to a single chip select of the cadence-spi master. I have the
> impression such a configuration is supported by none of those two
> drivers. I could try to extend both, however, I haven't found any other
> SPI driver, where I could find implementation inspiration. Is it
> supported by kernel?

drivers/gpio/gpio-max3191x.c supports "#daisy-chained-devices".
drivers/gpio/gpio-74x164.c supports multiple shift registers through the
"registers-number" DT property.

So both drivers handle this in their SPI slave drivers.

Of course this does not handle the mixed case, i.e. daisy-chaining
different types of devices.

> The documentation mentions only about the common 'daisy-chained-devices'
> property (devicetree/bindings/common-properties.txt). However, in order
> to try to implement it in the master driver, IMHO, the spi subsystem
> would need to have a call 'no-operation' to other nodes on the
> daisy-chain, which are not addressed by the given SPI access. Is there
> any recommended approach to address this case?

Supporting this in a generic way would indeed be nice, as it would mean
individual SPI slave drivers no longer have to care about it.
However, that may be difficult, as the master needs to known which
dummy (no-op) data is safe to shift through the non-addresses SPI slaves.

Gr{oetje,eeting}s,

                        Geert

-- 
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds