[linux-sunxi] [PATCH v2 3/7] phy: usb: sunxi: Introduce Allwinner A31 USB PHY support

[Hans de Goede]

Hi,

On 05/12/2014 02:24 PM, Kishon Vijay Abraham I wrote:
> Hi,
> 
> On Monday 12 May 2014 05:29 PM, Hans de Goede wrote:
>> Hi,
>>
>> On 05/12/2014 11:14 AM, Chen-Yu Tsai wrote:
>>> Hi,
>>>
>>> On Sat, May 10, 2014 at 8:56 PM, Maxime Ripard
>>> <maxime.ripard at free-electrons.com> wrote:
>>>> The USB phy controller in the A31 differs mostly from the older controllers
>>>> because it has a clock dedicated for each phy, while the older ones were having
>>>> a single clock for all the phys.
>>>>
>>>> Signed-off-by: Maxime Ripard <maxime.ripard at free-electrons.com>
>>>> Reviewed-by: Hans de Goede <hdegoede at redhat.com>
>>>> ---
>>>>  drivers/phy/phy-sun4i-usb.c | 35 ++++++++++++++++++++++++++---------
>>>>  1 file changed, 26 insertions(+), 9 deletions(-)
>>>>
>>>> diff --git a/drivers/phy/phy-sun4i-usb.c b/drivers/phy/phy-sun4i-usb.c
>>>> index e6e6c4ba7145..8bd89430d945 100644
>>>> --- a/drivers/phy/phy-sun4i-usb.c
>>>> +++ b/drivers/phy/phy-sun4i-usb.c
>>>> @@ -61,7 +61,6 @@
>>>>  #define MAX_PHYS                       3
>>>>
>>>>  struct sun4i_usb_phy_data {
>>>> -       struct clk *clk;
>>>>         void __iomem *base;
>>>>         struct mutex mutex;
>>>>         int num_phys;
>>>> @@ -71,6 +70,7 @@ struct sun4i_usb_phy_data {
>>>>                 void __iomem *pmu;
>>>>                 struct regulator *vbus;
>>>>                 struct reset_control *reset;
>>>> +               struct clk *clk;
>>>>                 int index;
>>>>         } phys[MAX_PHYS];
>>>>  };
>>>> @@ -146,13 +146,13 @@ static int sun4i_usb_phy_init(struct phy *_phy)
>>>>         struct sun4i_usb_phy_data *data = to_sun4i_usb_phy_data(phy);
>>>>         int ret;
>>>>
>>>> -       ret = clk_prepare_enable(data->clk);
>>>> +       ret = clk_prepare_enable(phy->clk);
>>>>         if (ret)
>>>>                 return ret;
>>>>
>>>>         ret = reset_control_deassert(phy->reset);
>>>>         if (ret) {
>>>> -               clk_disable_unprepare(data->clk);
>>>> +               clk_disable_unprepare(phy->clk);
>>>>                 return ret;
>>>>         }
>>>>
>>>> @@ -170,11 +170,10 @@ static int sun4i_usb_phy_init(struct phy *_phy)
>>>>  static int sun4i_usb_phy_exit(struct phy *_phy)
>>>>  {
>>>>         struct sun4i_usb_phy *phy = phy_get_drvdata(_phy);
>>>> -       struct sun4i_usb_phy_data *data = to_sun4i_usb_phy_data(phy);
>>>>
>>>>         sun4i_usb_phy_passby(phy, 0);
>>>>         reset_control_assert(phy->reset);
>>>> -       clk_disable_unprepare(data->clk);
>>>> +       clk_disable_unprepare(phy->clk);
>>>>
>>>>         return 0;
>>>>  }
>>>> @@ -228,8 +227,10 @@ static int sun4i_usb_phy_probe(struct platform_device *pdev)
>>>>         struct phy_provider *phy_provider;
>>>>         struct reset_control *reset;
>>>>         struct regulator *vbus;
>>>> +       bool dedicated_clocks;
>>>
>>> No default?
>>>
>>>>         struct resource *res;
>>>>         struct phy *phy;
>>>> +       struct clk *clk;
>>>>         char name[16];
>>>>         int i;
>>>>
>>>> @@ -249,15 +250,20 @@ static int sun4i_usb_phy_probe(struct platform_device *pdev)
>>>>         else
>>>>                 data->disc_thresh = 2;
>>>>
>>>> +       if (of_device_is_compatible(np, "allwinner,sun6i-a31-usb-phy"))
>>>> +               dedicated_clocks = true;
>>>> +
>>>
>>> dedicated_clocks is more than likely to be true from leftover data
>>> on the stack. This results in the usb phy driver probe failing, and
>>> the usb host drivers left in perpetual probe deferral on my Cubietruck.
>>>
>>> Adding an else section fixes this.
>>
>> Good catch, fixed this in the sunxi-devel branch for now (until Maxime sends a new version).
> 
> so I wont be taking this patch in my tree. Anyway FWIW
> Acked-by: Kishon Vijay Abraham I <kishon at ti.com>

To be clear, the sunxi-devel branch is a testing branch where I collect all the
pending sunxi related patches, it is not a way for patches to go upstream.

I don't know what the exact upstreaming plan for this patch-set is,
but logically this patch (once a fixed version is resend) should be going
upstream through you.

Regards,

Hans