Re: [PATCH v4] usb: dwc2: dwc2_vbus_supply_init: fix error check

[Minas Harutyunyan]

Hi Heiko,

On 4/10/2018 7:37 PM, Heiko Stübner wrote:
> Am Dienstag, 10. April 2018, 15:52:25 CEST schrieb Minas Harutyunyan:
>> Hi Heiko,
>>
>> On 4/10/2018 4:28 PM, Heiko Stuebner wrote:
>>> Am Montag, 26. März 2018, 11:00:01 CEST schrieb Tomeu Vizoso:
>>>> devm_regulator_get_optional returns -ENODEV if the regulator isn't
>>>> there, so if that's the case we have to make sure not to leave -ENODEV
>>>> in the regulator pointer.
>>>>
>>>> Also, make sure we return 0 in that case, but correctly propagate any
>>>> other errors. Also propagate the error from _dwc2_hcd_start.
>>>>
>>>> Fixes: 531ef5ebea96 ("usb: dwc2: add support for host mode external vbus
>>>> supply") Cc: Amelie Delaunay <amelie.delaunay@xxxxxx>
>>>> Signed-off-by: Tomeu Vizoso <tomeu.vizoso@xxxxxxxxxxxxx>
>>>
>>> The patch that gets fixed here also breaks display-output on dwc2-based
>>> Rockchip devices (likely even more), probably due to making the regulator
>>> framework hickup.
>>
>> Could you please elaborate what mean "breaks display-output".
>> On which Kernel version you apply this patch?
> 
> I think I may have written that poorly. _Without_ this patch I get
> display breakage on the most recent torvalds/master (merge-window)
> where "usb: dwc2: add support for host mode external vbus supply" is
> applied and this patch fixes the issue.
> 
> "breaks display output" means both hdmi + edp output are missing
> also including the backlight staying off.
> 
> The patch we're fixing here, causes a null-pointer dereference in the
> regulator framework, which seems to also cause issues when other
> regulators are enabled, which I think is what I'm seeing here.
> 
> 
Thank you for clarification. Earlier you added "reviewed" tag, this 
feedback can assumed as "tested" tag.

Thanks,
Minas

> Heiko
> 
>>
>> Thanks,
>> Minas
>>
>>> With this patch applied, apart from not seeing the NULL-ptr, I also get
>>> display output on my rk3288-veycron Chromebook again, so
>>>
>>> Tested-by: Heiko Stuebner <heiko@xxxxxxxxx>
>>>
>>>> v2: Only overwrite the error in the pointer after checking it (Heiko
>>>>
>>>>       Stübner <heiko@xxxxxxxxx>)
>>>>
>>>> v3: Remove hunks that shouldn't be in this patch
>>>> v4: Don't overwrite the error code before returning it (kbuild test
>>>>
>>>>       robot <lkp@xxxxxxxxx>)
>>>>
>>>> ---
>>>>
>>>>    drivers/usb/dwc2/hcd.c | 13 ++++++++-----
>>>>    1 file changed, 8 insertions(+), 5 deletions(-)
>>>>
>>>> diff --git a/drivers/usb/dwc2/hcd.c b/drivers/usb/dwc2/hcd.c
>>>> index 190f95964000..c51b73b3e048 100644
>>>> --- a/drivers/usb/dwc2/hcd.c
>>>> +++ b/drivers/usb/dwc2/hcd.c
>>>> @@ -358,9 +358,14 @@ static void dwc2_gusbcfg_init(struct dwc2_hsotg
>>>> *hsotg)>>
>>>>    static int dwc2_vbus_supply_init(struct dwc2_hsotg *hsotg)
>>>>    {
>>>>
>>>> +	int ret;
>>>> +
>>>>
>>>>    	hsotg->vbus_supply = devm_regulator_get_optional(hsotg->dev, "vbus");
>>>>
>>>> -	if (IS_ERR(hsotg->vbus_supply))
>>>> -		return 0;
>>>> +	if (IS_ERR(hsotg->vbus_supply)) {
>>>> +		ret = PTR_ERR(hsotg->vbus_supply);
>>>> +		hsotg->vbus_supply = NULL;
>>>> +		return ret == -ENODEV ? 0 : ret;
>>>> +	}
>>>>
>>>>    	return regulator_enable(hsotg->vbus_supply);
>>>>    
>>>>    }
>>>>
>>>> @@ -4342,9 +4347,7 @@ static int _dwc2_hcd_start(struct usb_hcd *hcd)
>>>>
>>>>    	spin_unlock_irqrestore(&hsotg->lock, flags);
>>>>
>>>> -	dwc2_vbus_supply_init(hsotg);
>>>> -
>>>> -	return 0;
>>>> +	return dwc2_vbus_supply_init(hsotg);
>>>>
>>>>    }
>>>>    
>>>>    /*
> 
> 
>