diff mbox

[media] cec: add missing inline stubs

Message ID 6419506.V8JtZUoqT3@wuerfel (mailing list archive)
State New, archived
Headers show

Commit Message

Arnd Bergmann July 1, 2016, 3:13 p.m. UTC 
On Friday, July 1, 2016 4:35:09 PM CEST Hans Verkuil wrote:
> On 07/01/2016 01:19 PM, Arnd Bergmann wrote:
> > The linux/cec.h header file contains empty inline definitions for
> > a subset of the API for the case in which CEC is not enabled,
> > however we have driver that call other functions as well:
> > 
> > drivers/media/i2c/adv7604.c: In function 'adv76xx_cec_tx_raw_status':
> > drivers/media/i2c/adv7604.c:1956:3: error: implicit declaration of function 'cec_transmit_done' [-Werror=implicit-function-declaration]
> > drivers/media/i2c/adv7604.c: In function 'adv76xx_cec_isr':
> > drivers/media/i2c/adv7604.c:2012:4: error: implicit declaration of function 'cec_received_msg' [-Werror=implicit-function-declaration]
> > drivers/media/i2c/adv7604.c: In function 'adv76xx_probe':
> > drivers/media/i2c/adv7604.c:3482:20: error: implicit declaration of function 'cec_allocate_adapter' [-Werror=implicit-function-declaration]
> 
> I don't understand this. These calls are under #if IS_ENABLED(CONFIG_VIDEO_ADV7842_CEC),
> and that should be 0 if MEDIA_CEC is not selected.
> 
> Am I missing some weird config combination?

This must have been a build error I ran into before your patch, when I
had this one applied locally instead:


which is still not overly nice, but it manages to avoid the
IS_REACHABLE() check and it lets MEDIA_CEC be a module.

	Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Comments

Hans Verkuil July 1, 2016, 3:22 p.m. UTC | #1 
On 07/01/2016 05:13 PM, Arnd Bergmann wrote:
> On Friday, July 1, 2016 4:35:09 PM CEST Hans Verkuil wrote:
>> On 07/01/2016 01:19 PM, Arnd Bergmann wrote:
>>> The linux/cec.h header file contains empty inline definitions for
>>> a subset of the API for the case in which CEC is not enabled,
>>> however we have driver that call other functions as well:
>>>
>>> drivers/media/i2c/adv7604.c: In function 'adv76xx_cec_tx_raw_status':
>>> drivers/media/i2c/adv7604.c:1956:3: error: implicit declaration of function 'cec_transmit_done' [-Werror=implicit-function-declaration]
>>> drivers/media/i2c/adv7604.c: In function 'adv76xx_cec_isr':
>>> drivers/media/i2c/adv7604.c:2012:4: error: implicit declaration of function 'cec_received_msg' [-Werror=implicit-function-declaration]
>>> drivers/media/i2c/adv7604.c: In function 'adv76xx_probe':
>>> drivers/media/i2c/adv7604.c:3482:20: error: implicit declaration of function 'cec_allocate_adapter' [-Werror=implicit-function-declaration]
>>
>> I don't understand this. These calls are under #if IS_ENABLED(CONFIG_VIDEO_ADV7842_CEC),
>> and that should be 0 if MEDIA_CEC is not selected.
>>
>> Am I missing some weird config combination?
> 
> This must have been a build error I ran into before your patch, when I
> had this one applied locally instead:
> 
> diff --git a/include/media/cec.h b/include/media/cec.h
> index c462f9b44074..564a6a06bed7 100644
> --- a/include/media/cec.h
> +++ b/include/media/cec.h
> @@ -187,7 +187,7 @@ static inline bool cec_is_sink(const struct cec_adapter *adap)
>  	return adap->phys_addr == 0;
>  }
>  
> -#if IS_ENABLED(CONFIG_MEDIA_CEC)
> +#if IS_REACHABLE(CONFIG_MEDIA_CEC)
>  struct cec_adapter *cec_allocate_adapter(const struct cec_adap_ops *ops,
>  		void *priv, const char *name, u32 caps, u8 available_las,
>  		struct device *parent);
> 
> because that was hiding the declarations when the code could not
> reach it. With your newer patch that is not possible any more.
> 
> I also wasn't aware that each of these already had their own Kconfig
> symbols. Could we just do this instead of your patch then:
> 
> diff --git a/drivers/media/i2c/Kconfig b/drivers/media/i2c/Kconfig
> index ce9006e10a30..73e047220905 100644
> --- a/drivers/media/i2c/Kconfig
> +++ b/drivers/media/i2c/Kconfig
> @@ -222,6 +222,7 @@ config VIDEO_ADV7604
>  config VIDEO_ADV7604_CEC
>  	bool "Enable Analog Devices ADV7604 CEC support"
>  	depends on VIDEO_ADV7604 && MEDIA_CEC
> +	depends on VIDEO_ADV7604=m || MEDIA_CEC=y
>  	---help---
>  	  When selected the adv7604 will support the optional
>  	  HDMI CEC feature.
> @@ -243,6 +244,7 @@ config VIDEO_ADV7842
>  config VIDEO_ADV7842_CEC
>  	bool "Enable Analog Devices ADV7842 CEC support"
>  	depends on VIDEO_ADV7842 && MEDIA_CEC
> +	depends on VIDEO_ADV7842=m || MEDIA_CEC=y
>  	---help---
>  	  When selected the adv7842 will support the optional
>  	  HDMI CEC feature.
> @@ -475,6 +477,7 @@ config VIDEO_ADV7511
>  config VIDEO_ADV7511_CEC
>  	bool "Enable Analog Devices ADV7511 CEC support"
>  	depends on VIDEO_ADV7511 && MEDIA_CEC
> +	depends on VIDEO_ADV7511=m || MEDIA_CEC=y
>  	---help---
>  	  When selected the adv7511 will support the optional
>  	  HDMI CEC feature.
> diff --git a/drivers/media/platform/vivid/Kconfig b/drivers/media/platform/vivid/Kconfig
> index 8e6918c5c87c..8e31146d079a 100644
> --- a/drivers/media/platform/vivid/Kconfig
> +++ b/drivers/media/platform/vivid/Kconfig
> @@ -26,6 +26,7 @@ config VIDEO_VIVID
>  config VIDEO_VIVID_CEC
>  	bool "Enable CEC emulation support"
>  	depends on VIDEO_VIVID && MEDIA_CEC
> +	depends on VIDEO_VIVID=m || MEDIA_CEC=y
>  	---help---
>  	  When selected the vivid module will emulate the optional
>  	  HDMI CEC feature.
> 
> which is still not overly nice, but it manages to avoid the
> IS_REACHABLE() check and it lets MEDIA_CEC be a module.

The only IS_REACHABLE is for the RC_CORE check, and that should remain.

With my patch MEDIA_CEC can remain a module provided MEDIA_SUPPORT is also
a module. All drivers depending on MEDIA_CEC also depend on MEDIA_SUPPORT,
so that works.

As I mentioned in a previous post, this will change in the future but while
it is in staging I prefer to keep it like this.

Regards,

	Hans
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Arnd Bergmann July 1, 2016, 3:37 p.m. UTC | #2 
On Friday, July 1, 2016 5:22:32 PM CEST Hans Verkuil wrote:
> > diff --git a/drivers/media/platform/vivid/Kconfig b/drivers/media/platform/vivid/Kconfig
> > index 8e6918c5c87c..8e31146d079a 100644
> > --- a/drivers/media/platform/vivid/Kconfig
> > +++ b/drivers/media/platform/vivid/Kconfig
> > @@ -26,6 +26,7 @@ config VIDEO_VIVID
> >  config VIDEO_VIVID_CEC
> >       bool "Enable CEC emulation support"
> >       depends on VIDEO_VIVID && MEDIA_CEC
> > +     depends on VIDEO_VIVID=m || MEDIA_CEC=y
> >       ---help---
> >         When selected the vivid module will emulate the optional
> >         HDMI CEC feature.
> > 
> > which is still not overly nice, but it manages to avoid the
> > IS_REACHABLE() check and it lets MEDIA_CEC be a module.
> 
> The only IS_REACHABLE is for the RC_CORE check, and that should remain.

I believe that is already taken care of by my earlier "[media] cec: add
RC_CORE dependency" patch, https://patchwork.linuxtv.org/patch/34892/
which seems to handle the dependency more gracefully (preventing nonsense
configurations rather than just not using RC_CORE).

> With my patch MEDIA_CEC can remain a module provided MEDIA_SUPPORT is also
> a module. All drivers depending on MEDIA_CEC also depend on MEDIA_SUPPORT,
> so that works.

To clarify, the problem with the option above is that VIDEO_VIVID_CEC
is a 'bool' option, and Kconfig lets that be turned on if both
VIDEO_VIVID and MEDIA_CEC are enabled, including the case where MEDIA_CEC
is a module and VIDEO_VIVID is not.

Your patch avoids that problem by making MEDIA_CEC a 'bool', my patch
above is an alternative by ensuring that VIDEO_VIVID_CEC cannot be
enabled if MEDIA_CEC is a module and VIDEO_VIVID is not.

	Arnd
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Hans Verkuil July 1, 2016, 5:30 p.m. UTC | #3 
On 07/01/2016 05:37 PM, Arnd Bergmann wrote:
> On Friday, July 1, 2016 5:22:32 PM CEST Hans Verkuil wrote:
>>> diff --git a/drivers/media/platform/vivid/Kconfig b/drivers/media/platform/vivid/Kconfig
>>> index 8e6918c5c87c..8e31146d079a 100644
>>> --- a/drivers/media/platform/vivid/Kconfig
>>> +++ b/drivers/media/platform/vivid/Kconfig
>>> @@ -26,6 +26,7 @@ config VIDEO_VIVID
>>>  config VIDEO_VIVID_CEC
>>>       bool "Enable CEC emulation support"
>>>       depends on VIDEO_VIVID && MEDIA_CEC
>>> +     depends on VIDEO_VIVID=m || MEDIA_CEC=y
>>>       ---help---
>>>         When selected the vivid module will emulate the optional
>>>         HDMI CEC feature.
>>>
>>> which is still not overly nice, but it manages to avoid the
>>> IS_REACHABLE() check and it lets MEDIA_CEC be a module.
>>
>> The only IS_REACHABLE is for the RC_CORE check, and that should remain.
> 
> I believe that is already taken care of by my earlier "[media] cec: add
> RC_CORE dependency" patch, https://patchwork.linuxtv.org/patch/34892/
> which seems to handle the dependency more gracefully (preventing nonsense
> configurations rather than just not using RC_CORE).
> 
>> With my patch MEDIA_CEC can remain a module provided MEDIA_SUPPORT is also
>> a module. All drivers depending on MEDIA_CEC also depend on MEDIA_SUPPORT,
>> so that works.
> 
> To clarify, the problem with the option above is that VIDEO_VIVID_CEC
> is a 'bool' option, and Kconfig lets that be turned on if both
> VIDEO_VIVID and MEDIA_CEC are enabled, including the case where MEDIA_CEC
> is a module and VIDEO_VIVID is not.
> 
> Your patch avoids that problem by making MEDIA_CEC a 'bool', my patch
> above is an alternative by ensuring that VIDEO_VIVID_CEC cannot be
> enabled if MEDIA_CEC is a module and VIDEO_VIVID is not.

The problem I have with both your suggestions is that options disappear just
because one dependency is a module. Very unexpected and hard to figure out.

I am not enthusiastic about the IS_REACHABLE(RC_CORE) solution either, but it
will do for now while it is in staging. I still think it is better then making
MEDIA_CEC depend on whether RC_CORE is a module or not. The RC support is an
optional extension and CEC still works fine without it.

Regards,

	Hans
--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/include/media/cec.h b/include/media/cec.h
index c462f9b44074..564a6a06bed7 100644
--- a/include/media/cec.h
+++ b/include/media/cec.h
@@ -187,7 +187,7 @@  static inline bool cec_is_sink(const struct cec_adapter *adap)
 	return adap->phys_addr == 0;
 }
 
-#if IS_ENABLED(CONFIG_MEDIA_CEC)
+#if IS_REACHABLE(CONFIG_MEDIA_CEC)
 struct cec_adapter *cec_allocate_adapter(const struct cec_adap_ops *ops,
 		void *priv, const char *name, u32 caps, u8 available_las,
 		struct device *parent);

because that was hiding the declarations when the code could not
reach it. With your newer patch that is not possible any more.

I also wasn't aware that each of these already had their own Kconfig
symbols. Could we just do this instead of your patch then:

diff --git a/drivers/media/i2c/Kconfig b/drivers/media/i2c/Kconfig
index ce9006e10a30..73e047220905 100644
--- a/drivers/media/i2c/Kconfig
+++ b/drivers/media/i2c/Kconfig
@@ -222,6 +222,7 @@  config VIDEO_ADV7604
 config VIDEO_ADV7604_CEC
 	bool "Enable Analog Devices ADV7604 CEC support"
 	depends on VIDEO_ADV7604 && MEDIA_CEC
+	depends on VIDEO_ADV7604=m || MEDIA_CEC=y
 	---help---
 	  When selected the adv7604 will support the optional
 	  HDMI CEC feature.
@@ -243,6 +244,7 @@  config VIDEO_ADV7842
 config VIDEO_ADV7842_CEC
 	bool "Enable Analog Devices ADV7842 CEC support"
 	depends on VIDEO_ADV7842 && MEDIA_CEC
+	depends on VIDEO_ADV7842=m || MEDIA_CEC=y
 	---help---
 	  When selected the adv7842 will support the optional
 	  HDMI CEC feature.
@@ -475,6 +477,7 @@  config VIDEO_ADV7511
 config VIDEO_ADV7511_CEC
 	bool "Enable Analog Devices ADV7511 CEC support"
 	depends on VIDEO_ADV7511 && MEDIA_CEC
+	depends on VIDEO_ADV7511=m || MEDIA_CEC=y
 	---help---
 	  When selected the adv7511 will support the optional
 	  HDMI CEC feature.
diff --git a/drivers/media/platform/vivid/Kconfig b/drivers/media/platform/vivid/Kconfig
index 8e6918c5c87c..8e31146d079a 100644
--- a/drivers/media/platform/vivid/Kconfig
+++ b/drivers/media/platform/vivid/Kconfig
@@ -26,6 +26,7 @@  config VIDEO_VIVID
 config VIDEO_VIVID_CEC
 	bool "Enable CEC emulation support"
 	depends on VIDEO_VIVID && MEDIA_CEC
+	depends on VIDEO_VIVID=m || MEDIA_CEC=y
 	---help---
 	  When selected the vivid module will emulate the optional
 	  HDMI CEC feature.