RE: [PATCH] [drivers] [SPI] SPI_GPIO: add support for controllers withmissing MISO pin
From: Marek Szyprowski
Date: Thu Jun 25 2009 - 08:17:37 EST
Hello,
On Tuesday, June 23, 2009 8:55 PM, David Brownell wrote:
> > There are some boards that do not strictly follow SPI standard and use only 3 wires
> > (SCLK, MOSI, SS) for connecting some simple auxiliary chips and controls them with GPIO
> > based 'spi controller'. In this configuration the MISO line is missing (it is not
> > required if the chip does not transfer any data back to host). The example of such
> > board is a NCP ARM S3C64XX based machine. This patch adds support for such non-standard
> > configuration in GPIO-based SPI controller.
> >
> > Reviewed-by: Kyungmin Park <kyungmin.park@xxxxxxxxxxx>
> > Signed-off-by: Marek Szyprowski <m.szyprowski@xxxxxxxxxxx>
>
> This is missing one integrity feature: it should reject all
> requests that provide a read buffer, since obviously it can't
> satisfy any such requests.
>
> Also, please add a comment in spi_gpio_request() explaining
> this point: that output-only configs exist, and you're
> supporting them by letting MISO be invalid.
>
> Address those two points and this will be almost OK.
Thanks for your review. I will post an updated patch soon.
> Thing is, this raises two related issues: (a) there's the
> analagous input-only case where MOSI isn't used, e.g. for
> some kinds of sensor; and (b) there's also "real 3-wire SPI"
> (spi->mode & SPI_3WIRE) where interactions are limited to
> half duplex and one pin switches roles between MOSI and MISO.
>
> Clearly this "output-only" case is a subset of SPI_3WIRE (the
> MOMI/SISO pin can't switch direction) so one more change I
> want to see is requiring that spi->mode flag be set in all
> SPI devices registered when this mode is used.
>
> If you have time, it would be good to generaize this patch
> to cover all of those 3-wire modes ... accept all half
> duplex calls, use gpio_direction_*() to switch direction.
>
> If not -- e.g. nothing to test that with! -- then I can
> understand, but please make your changes with that model
> in mind, and leave an appropriate FIXME in place.
I'm sorry, but I have only a hw board with missing MISO pin so I won't
be able to properly check/debug any other configurations. However I
will add a support for missing MOSI pin as well (simple code
reusage).
> p.s. I'll CC you on one more related patch. Briefly,
> there's a new spi_master->flags field that should
> set SPI_MASTER_HALF_DUPLEX. Arguably, this specific
> case should also advertise something like NO_RX.
Thanks. I will add SPI_MASTER_NO_TX/NO_RX flags there.
Best regards
--
Marek Szyprowski
Samsung Poland R&D Center
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/