[U-Boot] [PATCH 1/3] samsung: board: support eMMC reset using DT

[Simon Glass]

Hi,

On 11 January 2015 at 20:45, Joonyoung Shim <jy0922.shim at samsung.com> wrote:
> On 01/12/2015 10:45 AM, Jaehoon Chung wrote:
>> On 01/11/2015 02:46 AM, Simon Glass wrote:
>>> Hi,
>>>
>>>
>>> On 9 January 2015 at 23:53, Joonyoung Shim <jy0922.shim at samsung.com> wrote:
>>>> On 01/09/2015 08:13 PM, Jaehoon Chung wrote:
>>>>> On 01/09/2015 04:31 PM, Joonyoung Shim wrote:
>>>>>> +Cc Jaehoon,
>>>>>>
>>>>>> On 01/09/2015 03:18 AM, Simon Glass wrote:
>>>>>>> Hi Joonyoung,
>>>>>>>
>>>>>>> On 7 January 2015 at 22:44, Joonyoung Shim <jy0922.shim at samsung.com> wrote:
>>>>>>>> Some exynos boards require special handling of SD4_nRESET_OUT line for
>>>>>>>> eMMC memory to perform complete reboot e.g. Odroid X2/U3/XU3 boards.
>>>>>>>>
>>>>>>>> This will support eMMC reset using DT from reset_misc of samsung common
>>>>>>>> board file and each board files can support eMMC reset on non DT case.
>>>>>>>>
>>>>>>>> Signed-off-by: Joonyoung Shim <jy0922.shim at samsung.com>
>>>>>>>> ---
>>>>>>>>  board/samsung/common/board.c                   | 27 ++++++++++++++++++++++++++
>>>>>>>>  board/samsung/odroid/odroid.c                  |  2 +-
>>>>>>>>  doc/device-tree-bindings/exynos/emmc-reset.txt | 15 ++++++++++++++
>>>>>>>>  3 files changed, 43 insertions(+), 1 deletion(-)
>>>>>>>>  create mode 100644 doc/device-tree-bindings/exynos/emmc-reset.txt
>>>>>>>>
>>>>>>>> diff --git a/board/samsung/common/board.c b/board/samsung/common/board.c
>>>>>>>> index c04ab3c..3218463 100644
>>>>>>>> --- a/board/samsung/common/board.c
>>>>>>>> +++ b/board/samsung/common/board.c
>>>>>>>> @@ -40,6 +40,8 @@ __weak int exynos_power_init(void)
>>>>>>>>         return 0;
>>>>>>>>  }
>>>>>>>>
>>>>>>>> +__weak void exynos_reset_misc(void) {}
>>>>>>>> +
>>>>>>>>  #if defined CONFIG_EXYNOS_TMU
>>>>>>>>  /* Boot Time Thermal Analysis for SoC temperature threshold breach */
>>>>>>>>  static void boot_temp_check(void)
>>>>>>>> @@ -400,3 +402,28 @@ int misc_init_r(void)
>>>>>>>>         return 0;
>>>>>>>>  }
>>>>>>>>  #endif
>>>>>>>> +
>>>>>>>> +void reset_misc(void)
>>>>>>>> +{
>>>>>>>> +#ifdef CONFIG_OF_CONTROL
>>>>>>>
>>>>>>> I think all Samsung boards have this defined, so this #ifdef can be removed.
>>>>>>>
>>>>>>
>>>>>> OK, if it is true, we can remove this and i will not to support non DT case.
>>>>>>
>>>>>>>> +       struct gpio_desc gpio = {};
>>>>>>>> +       int node;
>>>>>>>> +
>>>>>>>> +       node = fdt_node_offset_by_compatible(gd->fdt_blob, 0,
>>>>>>>> +                       "samsung,emmc-reset");
>>>>>>>> +       if (node < 0)
>>>>>>>> +               return;
>>>>>>>> +
>>>>>>>> +       gpio_request_by_name_nodev(gd->fdt_blob, node, "reset-gpio", 0, &gpio,
>>>>>>>> +                                  GPIOD_IS_OUT);
>>>>>>>> +
>>>>>>>> +       if (dm_gpio_is_valid(&gpio)) {
>>>>>>>> +               /* Reset eMMC*/
>>>>>>>> +               dm_gpio_set_value(&gpio, 0);
>>>>>>>> +               mdelay(10);
>>>>>>>
>>>>>>> That's a very long reset! Is that what it needs?
>>>>>>>
>>>>>>
>>>>>> Jaehoon, is there any guide about wait time of reset at eMMC spec?
>>>>>
>>>>> Well, I'm not sure, but as Simon's comments, it's long time.
>>>>> (Minimum is 1us...Maximum isn't defined.)
>>>>>
>>>>
>>>> This just comes from reset_misc() of board/samsung/odroid/odroid.c file.
>>>> Even, hardkernel u-boot waits 50msec.
>>>
>>> The only info I have immediately to hand is the 'JEDEC Standard No.
>>> 84-A441' (eMMC) and it says a 1uS pulse is required. So can we change
>>> this to udelay(1)?
>>
>> It's minimum value, that's not problem that 10msec is set.
>> But if change this value, i think good that it changes to greater value than 1us.
>>
>
> Actually, i don't have any idea about proper wait time, so i will use
> just current 10msec with TODO comments until getting right guide about
> wait time. Is it ok?

Find with me, you are just moving code. I think you should also
mention that the eMMC spec says 1us and someone should test it.

Regards,
Simon