[U-Boot] SPI and spi_cs_activate

[U-Boot] SPI and spi_cs_activate

[Wolfgang Wegner]

Hi list,

I am trying to get the spi_mmc driver by Robert Selberg and Hans Eklund
to run on my new Coldfire board. (The driver was posted as a patch some
time ago, clearly marked as not ready for mainline. I just did not figure
out the correct search term to give the message ID.)

Now I am running into problems because the driver uses spi_cs_[de]activate,
which is rather useless on coldfire which does automatic cs.

Looking through other drivers' code it seems to me that spi_cs_[de]activate
was never meant to be called from anybody else but some special SPI
drivers that can not aotumatically control cs lines or for systems
where SPI is done in software and so setting the appropriate GPIOs
is handed over to special board-dependent code.

Is this correct, or am I missing something?

Regards,
Wolfgang

[U-Boot] SPI and spi_cs_activate

[Mike Frysinger]

On Friday 29 January 2010 10:59:28 Wolfgang Wegner wrote:
> I am trying to get the spi_mmc driver by Robert Selberg and Hans Eklund
> to run on my new Coldfire board. (The driver was posted as a patch some
> time ago, clearly marked as not ready for mainline. I just did not figure
> out the correct search term to give the message ID.)
> 
> Now I am running into problems because the driver uses spi_cs_[de]activate,
> which is rather useless on coldfire which does automatic cs.
>
> Looking through other drivers' code it seems to me that spi_cs_[de]activate
> was never meant to be called from anybody else but some special SPI
> drivers that can not aotumatically control cs lines or for systems
> where SPI is done in software and so setting the appropriate GPIOs
> is handed over to special board-dependent code.

there are plenty of SPI controllers which allow the CS to be manually 
controlled.  this isnt a "special" driver.

the MMC/SPI driver was written on a Blackfin system.  it working on any other 
system is coincidental as a result of leveraging the common frameworks.  feel 
free to submit fixes.
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20100129/3f571dfb/attachment.pgp 

[U-Boot] SPI and spi_cs_activate

[Wolfgang Wegner]

Hi Mike,

On Fri, Jan 29, 2010 at 11:11:36AM -0500, Mike Frysinger wrote:
> On Friday 29 January 2010 10:59:28 Wolfgang Wegner wrote:
> > I am trying to get the spi_mmc driver by Robert Selberg and Hans Eklund
> > to run on my new Coldfire board. (The driver was posted as a patch some
> > time ago, clearly marked as not ready for mainline. I just did not figure
> > out the correct search term to give the message ID.)
> > 
> > Now I am running into problems because the driver uses spi_cs_[de]activate,
> > which is rather useless on coldfire which does automatic cs.
> >
> > Looking through other drivers' code it seems to me that spi_cs_[de]activate
> > was never meant to be called from anybody else but some special SPI
> > drivers that can not aotumatically control cs lines or for systems
> > where SPI is done in software and so setting the appropriate GPIOs
> > is handed over to special board-dependent code.
> 
> there are plenty of SPI controllers which allow the CS to be manually 
> controlled.  this isnt a "special" driver.

OK, of course controlling CS manually is possible on most systems (via
GPIO). What I meant is that most systems do not care to control it
automatically (mostly for complexity reasons or because of real
HW limitations, as far as I understand).

> the MMC/SPI driver was written on a Blackfin system.  it working on any other 
> system is coincidental as a result of leveraging the common frameworks.  feel 
> free to submit fixes.

This is my intention in case I understand in which direction I have
to fix. (And in case I then get it to work in such a way, of course.)

So, maybe a more clear question:

Are all SPI drivers supposed to supply functions to manually control
the CS lines? If so, what means shall be provided to disable automatic
control?

Regards,
Wolfgang

[U-Boot] SPI and spi_cs_activate

[Mike Frysinger]

On Friday 29 January 2010 11:18:10 Wolfgang Wegner wrote:
> On Fri, Jan 29, 2010 at 11:11:36AM -0500, Mike Frysinger wrote:
> > On Friday 29 January 2010 10:59:28 Wolfgang Wegner wrote:
> > > I am trying to get the spi_mmc driver by Robert Selberg and Hans Eklund
> > > to run on my new Coldfire board. (The driver was posted as a patch some
> > > time ago, clearly marked as not ready for mainline. I just did not
> > > figure out the correct search term to give the message ID.)
> > >
> > > Now I am running into problems because the driver uses
> > > spi_cs_[de]activate, which is rather useless on coldfire which does
> > > automatic cs.
> > >
> > > Looking through other drivers' code it seems to me that
> > > spi_cs_[de]activate was never meant to be called from anybody else but
> > > some special SPI drivers that can not aotumatically control cs lines or
> > > for systems where SPI is done in software and so setting the
> > > appropriate GPIOs is handed over to special board-dependent code.
> >
> > there are plenty of SPI controllers which allow the CS to be manually
> > controlled.  this isnt a "special" driver.
> 
> OK, of course controlling CS manually is possible on most systems (via
> GPIO). What I meant is that most systems do not care to control it
> automatically (mostly for complexity reasons or because of real
> HW limitations, as far as I understand).

i'm not talking about GPIOs.  i'm talking about dedicated CS pins that are 
connected to the controller.

> > the MMC/SPI driver was written on a Blackfin system.  it working on any
> > other system is coincidental as a result of leveraging the common
> > frameworks.  feel free to submit fixes.
> 
> This is my intention in case I understand in which direction I have
> to fix. (And in case I then get it to work in such a way, of course.)

see how the linux mmc/spi driver does it.  i dont think linux provides a 
dedicate cs function for SPI drivers to use, but the linux mmc/spi doesnt seem 
to have a problem.

> Are all SPI drivers supposed to supply functions to manually control
> the CS lines?

*shrug*

> If so, what means shall be provided to disable automatic control?

i dont understand the question.  implement the two functions that are part of 
the SPI API.
-mike
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part.
Url : http://lists.denx.de/pipermail/u-boot/attachments/20100129/850c0515/attachment.pgp