Re: [PATCH V2 0/2] spi/bcm63xx: fix multi transfer messages

[Grant Likely <grant.likely-s3s/WqlpOiPyB63q8FvJNQ@public.gmane.org>]

On Sun,  3 Feb 2013 15:15:11 +0100, Jonas Gorski <jogo-p3rKhJxN3npAfugRpC6u6w@public.gmane.org> wrote:
> The bcm63xx SPI controller does not support keeping CS up after doing a
> transfer. Since this is problematic for most typical use cases, this
> patchset introduces a workaround by combining small enough messages
> to one transfer, rejecting anything that can't be fulfilled with the
> hardware.
> 
> Patch one properly rejects anything impossible to transfer with these
> limitations.
> Patch two introduces logic for combining transfers to one to be able to
> use it for typical use cases (register accesses and flash access).
> 
> Build and run tested on a BCM6368 with a SPI controlled switch attached
> requiring write-then-read with CS asserted.
> 
> Changes V1 -> V2:
>  * split into two patches
>  * fixed return type of bcm63xx_txrx_bufs()
>  * slightly reworked bcm63xx_txrx_bufs, obsoleting one local variable
>  * added a bit more comments in the code
>  * added error messages indicated why transfers were rejected

Another option is to allow longer transfers if a GPIO is used for the CS
line. A lot of SPI controllers need to do that.

That isn't a critique of this patch because it is essentially a bug fix,
but rather a way you can work around the limitation.

g.

> 
> Jonas Gorski (2):
>   spi/bcm63xx: reject transfers unable to transfer
>   spi/bcm63xx: work around inability to keep CS up
> 
>  drivers/spi/spi-bcm63xx.c |  179 +++++++++++++++++++++++++++++++--------------
>  1 file changed, 125 insertions(+), 54 deletions(-)
> 
> -- 
> 1.7.10.4
> 

-- 
Grant Likely, B.Sc, P.Eng.
Secret Lab Technologies, Ltd.

------------------------------------------------------------------------------
Free Next-Gen Firewall Hardware Offer
Buy your Sophos next-gen firewall before the end March 2013 
and get the hardware for free! Learn more.
http://p.sf.net/sfu/sophos-d2d-feb