[U-Boot] [PATCH] pci: mx6: Implement reset callback

[Eric Nelson]

Hi Marek,

On 02/03/2014 11:17 AM, Marek Vasut wrote:
> On Monday, February 03, 2014 at 12:56:04 PM, Stefano Babic wrote:
>> Hi Marek,
>>
>> sorry for late answer.
>>
>> On 28/01/2014 20:32, Marek Vasut wrote:
>>> On Tuesday, January 28, 2014 at 04:06:12 PM, Stefano Babic wrote:
>>>> Hi Marek,
>>>>
>>>> On 24/01/2014 16:25, Marek Vasut wrote:
>>>>> Add a callback so that a board can implement it's own specific routine
>>>>> to toggle the port's nRESET line.
>>>>>
>>>>> Signed-off-by: Marek Vasut <marex at denx.de>
>>>>> Cc: Stefano Babic <sbabic at denx.de>
>>>>> ---
>>>>>
>>>>>   drivers/pci/pcie_imx.c | 12 +++++++++---
>>>>>   1 file changed, 9 insertions(+), 3 deletions(-)
>>>>>
>>>>> diff --git a/drivers/pci/pcie_imx.c b/drivers/pci/pcie_imx.c
>>>>> index 0a74867..b554075 100644
>>>>> --- a/drivers/pci/pcie_imx.c
>>>>> +++ b/drivers/pci/pcie_imx.c
>>>>> @@ -450,6 +450,13 @@ static int imx6_pcie_init_phy(void)
>>>>>
>>>>>   	return 0;
>>>>>
>>>>>   }
>>>>>
>>>>> +__weak int imx6_pcie_toggle_reset(void)
>>>>> +{
>>>>> +	/* This function ought to be overridden ! */
>>>>> +	puts("WARNING: Make sure the PCIe nRESET line is connected!\n");
>>>>> +	return 0;
>>>>> +}
>>>>> +
>>>>
>>>> Just to know: I assume that the nRESET is implemented with a GPIO.
>>>
>>> Yes, that's how it is on all designs I saw thus far (but see below).
>>>
>>>> I am
>>>> expecting then in the board files a diffusion of imx6_pcie_toggle_reset,
>>>> where the oinly difference is the number of GPIO.
>>>
>>> The problem is, there are boards with no nRESET connected to the slot.
>>
>> Any reference to the Sabrelite is, of course, purely coincidental. But
>> this is a hardware bug on a specific board and we should not adjust all
>> boards according to the broken one.
>
> Well ... SL and N6X both. For all I care, we can have #define
> MX6_PCIE_RESET_GPIO and if that's not defined, puke out this warning. And
> ultimatelly let this function be overriden anyway in case people used some GPIO
> expander or whatnot. So the change to this would be:
>
> __weak int imx6_pcie_toggle_reset(void)
> {
> #ifdef CONFIG_MX6_PCIE_RESET_GPIO
>   gpio_set...
>   mdelay();
>   gpio_set...
>   mdelay();
> #else
>   puts("Oh yeah, broken design :-(\n");

That's pretty harsh!

We have lots of stuff working without a GPIO...