Re: Help with freescale p1010 based espi implementation.

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

On Wed, Oct 19, 2011 at 12:46:22AM -0500, Robin Holt wrote:
> I am a newbie to SPI and Broadcom Robo switch parts, so any correction
> to my direction early on is _VERY_ welcome.
> 
> 
> The Freescale P1010 communications SOC includes an ESPI controller
> which works with the spi-fsl-espi.c driver.  In that driver, the chip
> selects are limited to 4 by the SPCOM_CS field of the command register.
> Unfortunately, in the documentation, the only supported value for
> that field is '00b'.  In our board design, we decided to use GPIO5
> to effectively drive the select line of a 1-of-2 MUX chip (TI Part
> SN74CBTLV3257PWR pin 1).
> 
> My immediate need is a quick-fix to the espi driver to allow driving
> GPIO5 in addition to the two bits currently being used by the chip select
> portion of the command register.  By quick fix, I am looking for something
> to maintain locally until a more general method for abstracting the
> board specific CS enhancements from the ESPI driver.

The solution here is to extend the espi driver to directly manipulate
the gpio line (using gpiolib) before and after transfers.  Other gpio
controllers already support this mode of operation and you should be
able to find examples in the drivers/spi directory.

> I have modified the boot loader to have GPIO5 set as an output and setting
> it high prior to booting the kernel.  That reflects the strapping on
> the board and the inactive high state needed by the CS signal.
> 
> I believe that I want my flattened device tree to have
> fsl,espi-num-chipselects set to 8 (or at least 5), and then assign the
> two devices chipselect 4 (the SPI Flash part which is addressed with
> GPIO5=1 and command.CS field=0) and chipselect 0 (Broadcom Robo switch
> which is addressed with GPIO5=0 and command.CS field=0).

Yes, that sounds appropriate.

> 
> On the subject of the GPIOs, I see those are being defined by the
> mpc8xxx_gpio.c driver.  I do see messages during boot, and can use the
> sysfs interface to toggle GPIO5 (which is really mapped to GPIO 229, but
> that is probably my problem).  How would I locate that GPIO "correctly"
> from within the spi-fsl-espi.c driver?

Linux gpio numbers are entirely arbitrary, and you don't want to be
using the sysfs gpio interface for solving the problem.  You need to
use a "gpios" property to specify the controller and hwgpio number +
flags.  The driver will then be able to decode that property to get
the Linux gpio number for the CS line.

> Another, mostly unrelated question is, can I get the kernel to define the
> robo switch device without any kernel driver matching its FDT "compatible"
> entry?  What I believe the SDK needs is merely a SPI driver defined and
> then I think I can get the remainder of the Broadcom SDK working.

Do you want to access that device from userspace?  I believe that the
compatible value "linux,spidev" will make it available to userspace
(even if it is a bit of a hack; it would be better to be able to claim
unbound device directly from userspace).

g.


------------------------------------------------------------------------------
The demand for IT networking professionals continues to grow, and the
demand for specialized networking skills is growing even more rapidly.
Take a complimentary Learning@Ciosco Self-Assessment and learn 
about Cisco certifications, training, and career opportunities. 
http://p.sf.net/sfu/cisco-dev2dev