Re: [PATCH 09/14] Blackfin SPI driver: Fix SPI driver to work with SPI flash ST25P16 on bf548

[Bryan Wu <bryan.wu@analog.com>]

On Wed, 2007-10-31 at 01:02 -0700, David Brownell wrote:
> On Wednesday 31 October 2007, Bryan Wu wrote:
> > IMO, the spi_transfer.speed_hz <= spi_board_info.max_speed_hz and if
> > spi_trasnfer.speed_hz is 0, we should use spi_board_info.max_speed_hz.
> > From the meaning of max_speed_hz, the spi_transfer.speed_hz should not
> > beyond max_speed_hz.
> 
> According to the interface spec that's not how it works; so I'm
> not sure what you base your opinion on.  It's not defined with
> such constraints.
> 
My opinion is based on the name of max_speed_hz.
Actually, I know it is for the controller to compute the baud rate from
system bus clock or something similar.

> That might be a reasonable policy to adopt in many cases, and
> nothing prevents any given protocol driver from choosing to follow
> it.  But likewise, if the driver chooses _not_ to follow it, it's
> wrong for a controller driver to add its own private rules.
> 
> > In your explanation, the max_speed_hz is just a default value. the
> > transfer speed_hz can beyond max_speed_hz.
> 
> Kerneldoc for board info says;
> 
>  * @max_speed_hz: Initializes spi_device.max_speed_hz; based on limits
>  *      from the chip datasheet and board-specific signal quality issues.
> 
> So yes -- initial value, it can be changed.  If a driver says
> 
> 	spi->max_speed_hz = X;
> 	spi_setup(spi);
> 
> that's how it requests a different clock rate ceiling.  It can lower
> the rate; it can raise it.
> 
> 
> The reason it's called max_speed_hz is that most systems compute the
> clock rate by dividing a base clock, and few can achieve that exact
> value.  The name indicates that it's to be treated as an upper limit,
> not a lower limit or exact value.  The controller driver should get
> as close to that rate as it can, without going over.
> 

Thanks for this clarification.
max_speed_hz acts as a default clock rate ceiling in the spi_setup(spi)
At the beginning of each real spi transfer, we should check the
spi_transfer.speed_hz like this:

---
		/* Speed setup (surely valid because already checked) */
		tmp = transfer->speed_hz;
		if (tmp == 0)
			tmp = chip->max_speed_hz;
		spi_set_baud_rate(tmp);
---

-Bryan