[v2] usb: dwc2: Fix build in periphal-only mode

The bus_suspended member of struct dwc2_hsotg is only present in builds
that support host-mode.

Fixes: 24d209dba5a3 ("usb: dwc2: Fix hibernation between host and device modes.")
Signed-off-by: Phil Elwell <phil@raspberrypi.com>
---
 drivers/usb/dwc2/core_intr.c | 4 ++++
 1 file changed, 4 insertions(+)

v2: Correct commit hash used in the Fixes line.

On Fri, May 28, 2021 at 10:13:50AM +0100, Phil Elwell wrote:
> The bus_suspended member of struct dwc2_hsotg is only present in builds
> that support host-mode.
> 
> Fixes: 24d209dba5a3 ("usb: dwc2: Fix hibernation between host and device modes.")
> Signed-off-by: Phil Elwell <phil@raspberrypi.com>
> ---
>  drivers/usb/dwc2/core_intr.c | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> v2: Correct commit hash used in the Fixes line.
> 
> diff --git a/drivers/usb/dwc2/core_intr.c b/drivers/usb/dwc2/core_intr.c
> index a5ab03808da6..03d0c034cf57 100644
> --- a/drivers/usb/dwc2/core_intr.c
> +++ b/drivers/usb/dwc2/core_intr.c
> @@ -725,7 +725,11 @@ static inline void dwc_handle_gpwrdn_disc_det(struct dwc2_hsotg *hsotg,
>  	dwc2_writel(hsotg, gpwrdn_tmp, GPWRDN);
>  
>  	hsotg->hibernated = 0;
> +
> +#if IS_ENABLED(CONFIG_USB_DWC2_HOST) ||	\
> +	IS_ENABLED(CONFIG_USB_DWC2_DUAL_ROLE)
>  	hsotg->bus_suspended = 0;
> +#endif
>  
>  	if (gpwrdn & GPWRDN_IDSTS) {
>  		hsotg->op_state = OTG_STATE_B_PERIPHERAL;
> -- 
> 2.25.1
> 

I do not understand, the field in the structure is present for all, why
is this crazy #if needed here?

I see that the commit you reference here did add the new line to set
bus_suspended, which seemed to be the point here.  Why will the #if
values matter here?

thanks,

greg k-h

Hi Greg,

On 28/05/2021 10:23, Greg Kroah-Hartman wrote:
> On Fri, May 28, 2021 at 10:13:50AM +0100, Phil Elwell wrote:
>> The bus_suspended member of struct dwc2_hsotg is only present in builds
>> that support host-mode.
>>
>> Fixes: 24d209dba5a3 ("usb: dwc2: Fix hibernation between host and device modes.")
>> Signed-off-by: Phil Elwell <phil@raspberrypi.com>
>> ---
>>   drivers/usb/dwc2/core_intr.c | 4 ++++
>>   1 file changed, 4 insertions(+)
>>
>> v2: Correct commit hash used in the Fixes line.
>>
>> diff --git a/drivers/usb/dwc2/core_intr.c b/drivers/usb/dwc2/core_intr.c
>> index a5ab03808da6..03d0c034cf57 100644
>> --- a/drivers/usb/dwc2/core_intr.c
>> +++ b/drivers/usb/dwc2/core_intr.c
>> @@ -725,7 +725,11 @@ static inline void dwc_handle_gpwrdn_disc_det(struct dwc2_hsotg *hsotg,
>>   	dwc2_writel(hsotg, gpwrdn_tmp, GPWRDN);
>>   
>>   	hsotg->hibernated = 0;
>> +
>> +#if IS_ENABLED(CONFIG_USB_DWC2_HOST) ||	\
>> +	IS_ENABLED(CONFIG_USB_DWC2_DUAL_ROLE)
>>   	hsotg->bus_suspended = 0;
>> +#endif
>>   
>>   	if (gpwrdn & GPWRDN_IDSTS) {
>>   		hsotg->op_state = OTG_STATE_B_PERIPHERAL;
>> -- 
>> 2.25.1
>>
> 
> I do not understand, the field in the structure is present for all, why
> is this crazy #if needed here?
> 
> I see that the commit you reference here did add the new line to set
> bus_suspended, which seemed to be the point here.  Why will the #if
> values matter here?

Sorry to waste your brain cycles on this. There is a problem, but it only exists 
in branches where the blamed commit (24d209dba5a3) has been back-ported as a 
Fix, because it depends on commit 012466fc8ccc which isn't a Fix and therefore
hasn't been back-ported. Sadly 012466fc8ccc doesn't back-port cleanly on its own 
- either more cherry-picks or a temporary patch like mine will be needed.

Phil

On 28/05/2021 10:51, Greg Kroah-Hartman wrote:

> On Fri, May 28, 2021 at 10:37:48AM +0100, Phil Elwell wrote:
>> Hi Greg,
>>
>> On 28/05/2021 10:23, Greg Kroah-Hartman wrote:
>>> On Fri, May 28, 2021 at 10:13:50AM +0100, Phil Elwell wrote:
>>>> The bus_suspended member of struct dwc2_hsotg is only present in builds
>>>> that support host-mode.
>>>>
>>>> Fixes: 24d209dba5a3 ("usb: dwc2: Fix hibernation between host and device modes.")
>>>> Signed-off-by: Phil Elwell <phil@raspberrypi.com>
>>>> ---
>>>>    drivers/usb/dwc2/core_intr.c | 4 ++++
>>>>    1 file changed, 4 insertions(+)
>>>>
>>>> v2: Correct commit hash used in the Fixes line.
>>>>
>>>> diff --git a/drivers/usb/dwc2/core_intr.c b/drivers/usb/dwc2/core_intr.c
>>>> index a5ab03808da6..03d0c034cf57 100644
>>>> --- a/drivers/usb/dwc2/core_intr.c
>>>> +++ b/drivers/usb/dwc2/core_intr.c
>>>> @@ -725,7 +725,11 @@ static inline void dwc_handle_gpwrdn_disc_det(struct dwc2_hsotg *hsotg,
>>>>    	dwc2_writel(hsotg, gpwrdn_tmp, GPWRDN);
>>>>    	hsotg->hibernated = 0;
>>>> +
>>>> +#if IS_ENABLED(CONFIG_USB_DWC2_HOST) ||	\
>>>> +	IS_ENABLED(CONFIG_USB_DWC2_DUAL_ROLE)
>>>>    	hsotg->bus_suspended = 0;
>>>> +#endif
>>>>    	if (gpwrdn & GPWRDN_IDSTS) {
>>>>    		hsotg->op_state = OTG_STATE_B_PERIPHERAL;
>>>> -- 
>>>> 2.25.1
>>>>
>>> I do not understand, the field in the structure is present for all, why
>>> is this crazy #if needed here?
>>>
>>> I see that the commit you reference here did add the new line to set
>>> bus_suspended, which seemed to be the point here.  Why will the #if
>>> values matter here?
>> Sorry to waste your brain cycles on this. There is a problem, but it only
>> exists in branches where the blamed commit (24d209dba5a3) has been
>> back-ported as a Fix, because it depends on commit 012466fc8ccc which isn't
>> a Fix and therefore
>> hasn't been back-ported. Sadly 012466fc8ccc doesn't back-port cleanly on its
>> own - either more cherry-picks or a temporary patch like mine will be
>> needed.
> So should we revert this commit from the stable releases where it showed
> up?  Which ones specifically?
>
> If so, please let me and stable@vger.kernel.org know and we can take
> care of it there.

Reverting back-ports of 24d209dba5a3 would be sufficient, although you are then 
left with the problem that 24d209dba5a3 was intended to address.

I'll email the stable list.

Phil