[U-Boot] [PATCH 03/18] sf: fix sf probe

[Nikita Kiryanov]

On 04/08/14 17:02, Tom Rini wrote:
> On Mon, Aug 04, 2014 at 04:45:57PM +0300, Nikita Kiryanov wrote:
>>
>>
>> On 04/08/14 16:10, Marek Vasut wrote:
>>> On Monday, August 04, 2014 at 02:48:54 PM, Nikita Kiryanov wrote:
>>>> Hi Marek,
>>>>
>>>> On 03/08/14 16:46, Marek Vasut wrote:
>>>>> On Sunday, August 03, 2014 at 09:34:33 AM, Nikita Kiryanov wrote:
>>>>>> MXC SPI driver has a feature whereas a GPIO line can be used as a CS
>>>>>> signal. This is set up by joining the CS and GPIO values into a single
>>>>>> value using (cs | gpio << 8), and passing it off as a CS value. This
>>>>>> breaks the sf probe command, because it is no longer possible to invoke
>>>>>> it as sf probe <cs>. Instead, the user must use sf probe <cs | gpio <<
>>>>>> 8>.
>>>>>>
>>>>>> Fix this by introducing a new board function: board_spi_cs_gpio().
>>>>>> When called, board_spi_cs_gpio() will return the gpio number for the
>>>>>> cs value it is given.
>>>>>>
>>>>>> Cc: Jagannadha Sutradharudu Teki <jagannadh.teki at gmail.com>
>>>>>> Cc: Eric Nelson <eric.nelson at boundarydevices.com>
>>>>>> Cc: Eric Benard <eric at eukrea.com>
>>>>>> Cc: Fabio Estevam <fabio.estevam at freescale.com>
>>>>>> Cc: Tim Harvey <tharvey at gateworks.com>
>>>>>> Cc: Stefano Babic <sbabic at denx.de>
>>>>>> Cc: Tom Rini <trini at ti.com>
>>>>>> Signed-off-by: Nikita Kiryanov <nikita at compulab.co.il>
>>>>>
>>>>> Just curious, but is this fixing generic SF code or MXC SPI driver ? I'd
>>>>> think the later, but it's not obvious from neither the description nor
>>>>> the subject. I don't quite understand the problem that you're trying to
>>>>> fix either, what happened, did the user command interface change ?
>>>>
>>>> The U-Boot shell command "sf probe" can accept a chip select value, but
>>>> if the SPI device on the other end requires an active chip-select over
>>>> multiple transactions (achieved in the MXC SPI driver using a GPIO),
>>>> simply typing something like "sf probe 0" will not work.
>>>
>>> Why not ?
>>>
>>>> This is because whatever the user passes as chip select is propagated
>>>> to the driver, and the driver expects this value to have GPIO
>>>> information. So for example, if IMX_GPIO_NR(2, 30) is used to force
>>>> active chip select 0, then instead of "sf probe 0" the user will have
>>>> to type "sf probe 15872".
>>>
>>> You mean sf probe 0:15872 , right ?
>>
>> It's the same thing:
>> sf probe [[bus:]cs] [hz] [mode]
>>
>> The point is that cs 0 has to be represented as "15872", instead of "0".
>
> Eeep.  That seems very likely to be gotten incorrect by users.
>
> Can we do something like:
> mxc_spi.c:
> __weak int board_map_spi_cs_value(int desired_cs) { return -EINVAL; }
>
> fooboard.c:
> board_map_spi_cs_value(int desired_cs) {
>    if (desired_cs == 0)
>      return IMX_GPIO_NR(2, 30);
>    else
>      return -EINVAL;
> }

That's pretty much what the patch does.

>
> I think it'll be very bad if the user has to type 'sf probe 0:15872' or
> 'sf probe 15872' since that's a programming detail rather than saying
> bank 2, gpio 30 (which I assume is what IMX_GPIO_NR means).
>

Agreed.