[U-Boot] [PATCH] mtd: nand: denali_dt: Add reset support

[Masahiro Yamada]

Hi Marek,


2018-09-06 17:02 GMT+09:00 Marek Vasut <marek.vasut at gmail.com>:
> On 09/06/2018 03:46 AM, Masahiro Yamada wrote:
>> 2018-09-05 17:31 GMT+09:00 Marek Vasut <marek.vasut at gmail.com>:
>>> On 08/24/2018 03:10 PM, Marek Vasut wrote:
>>>> On 08/24/2018 03:05 PM, Masahiro Yamada wrote:
>>>>> Hi Marek,
>>>>
>>>> Hi,
>>>>
>>>>> 2018-08-21 22:57 GMT+09:00 Marek Vasut <marex at denx.de>:
>>>>>> Add optional reset support into the Denali NAND driver. In case there is
>>>>>> a valid reset entry in the DT, the reset gets deasserted before the NAND
>>>>>> controller gets used.
>>>>>>
>>>>>> Signed-off-by: Marek Vasut <marex at denx.de>
>>>>>> Cc: Masahiro Yamada <yamada.masahiro at socionext.com>
>>>>>> ---
>>>>>>  drivers/mtd/nand/denali_dt.c | 9 +++++++++
>>>>>>  1 file changed, 9 insertions(+)
>>>>>>
>>>>>> diff --git a/drivers/mtd/nand/denali_dt.c b/drivers/mtd/nand/denali_dt.c
>>>>>> index 65a7797f0f..6fcd7d3843 100644
>>>>>> --- a/drivers/mtd/nand/denali_dt.c
>>>>>> +++ b/drivers/mtd/nand/denali_dt.c
>>>>>> @@ -9,6 +9,7 @@
>>>>>>  #include <linux/io.h>
>>>>>>  #include <linux/ioport.h>
>>>>>>  #include <linux/printk.h>
>>>>>> +#include <reset.h>
>>>>>>
>>>>>>  #include "denali.h"
>>>>>>
>>>>>> @@ -64,6 +65,7 @@ static int denali_dt_probe(struct udevice *dev)
>>>>>>         const struct denali_dt_data *data;
>>>>>>         struct clk clk;
>>>>>>         struct resource res;
>>>>>> +       struct reset_ctl reset_ctl;
>>>>>>         int ret;
>>>>>>
>>>>>>         data = (void *)dev_get_driver_data(dev);
>>>>>> @@ -97,6 +99,13 @@ static int denali_dt_probe(struct udevice *dev)
>>>>>>
>>>>>>         denali->clk_x_rate = clk_get_rate(&clk);
>>>>>>
>>>>>> +       ret = reset_get_by_index(dev, 0, &reset_ctl);
>>>>>> +       if (!ret) {
>>>>>> +               reset_assert(&reset_ctl);
>>>>>> +               udelay(2);
>>>>>> +               reset_deassert(&reset_ctl);
>>>>>> +       }
>>>>>> +
>>>>>>         return denali_init(denali);
>>>>>>  }
>>>>>>
>>>>>
>>>>>
>>>>> I was testing this patch today
>>>>> because my SoC also has a reset line for NAND.
>>>>>
>>>>>
>>>>> This patch looks trivial enough,
>>>>> but if the NAND is reset here,
>>>>> my board fails to boot with
>>>>> "nand_base: timeout while waiting for chip to become ready"
>>>>>
>>>>>
>>>>> I do not know why.
>>>>> Please give me more time to figure out what is going on in my SoC.
>>>>
>>>> Sure. Maybe you need to un-reset more stuff or the reset polarity is wrong ?
>>
>>
>> I do not think so.
>>
>> The reset sequence of Denali is not so simple as you may think
>> because the HW sequencer runs after the reset de-assertion.
>> It includes some SoC-dependent parts.
>>
>>
>>
>>>
>>> Did you figure something out ?
>>>
>>
>>
>> Just testing one of my boards.
>>
>> It booted by adding mdelay() after reset_deassert().
>>
>>
>>
>>         reset_assert(&reset_ctl);
>>         udelay(2);
>>         reset_deassert(&reset_ctl);
>>         mdelay(1);           /* Why is this long delay required? */
>
> It works right away after ungating the reset on my board.



With my further tests, it typically takes 200 usec
for the chip identification sequence after the reset de-assertion.

The number is for my UniPhier SoC family.

This part is SoC-dependent,
so the situation is probably different for SOCFPGA.


I'd like to suggest the following sequence.

         reset_assert(&reset_ctl);
         udelay(1);
         reset_deassert(&reset_ctl);
         udelay(200);



Is this OK with you?


The IP datasheet says that the reset lines should be
kept 10 clock cycles.  I think udelay(1) is good enough for reset assertion.



BTW, are you applying this patch to your tree?

Or, do you want me to pick this up?



>> I will read HW docs more,
>> and talk to HW engineers
>> to figure out the right solution.
>
> OK


-- 
Best Regards
Masahiro Yamada