base: core: Remove WARN_ON from link dependencies check

Message ID 20180712080623.21203-1-benjamin.gaignard@st.com
State Superseded
Headers show
Series
  • base: core: Remove WARN_ON from link dependencies check
Related show

Commit Message

Benjamin Gaignard July 12, 2018, 8:06 a.m.
In some cases the link between between customer and supplier
already exist. Do not warn about already existing dependencies
because device_link_add() take care of this case.

Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@st.com>

---
 drivers/base/core.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

-- 
2.15.0

Comments

Greg KH July 12, 2018, 8:55 a.m. | #1
On Thu, Jul 12, 2018 at 10:06:23AM +0200, Benjamin Gaignard wrote:
> In some cases the link between between customer and supplier

> already exist. Do not warn about already existing dependencies

> because device_link_add() take care of this case.


Why would a link already exist that is asked to be created again?  What
code path causes this?

> 

> Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>

> Signed-off-by: Benjamin Gaignard <benjamin.gaignard@st.com>

> ---

>  drivers/base/core.c | 4 ++--

>  1 file changed, 2 insertions(+), 2 deletions(-)

> 

> diff --git a/drivers/base/core.c b/drivers/base/core.c

> index df3e1a44707a..fcdc17f0f349 100644

> --- a/drivers/base/core.c

> +++ b/drivers/base/core.c

> @@ -105,7 +105,7 @@ static int device_is_dependent(struct device *dev, void *target)

>  	struct device_link *link;

>  	int ret;

>  

> -	if (WARN_ON(dev == target))

> +	if (dev == target)

>  		return 1;

>  

>  	ret = device_for_each_child(dev, target, device_is_dependent);

> @@ -113,7 +113,7 @@ static int device_is_dependent(struct device *dev, void *target)

>  		return ret;

>  

>  	list_for_each_entry(link, &dev->links.consumers, s_node) {

> -		if (WARN_ON(link->consumer == target))

> +		if (link->consumer == target)

>  			return 1;


Both of these WARN_ON are for valid code?  That feels really odd to me,
I need more explanation here please.

thanks,

greg k-h
Benjamin Gaignard July 12, 2018, 9:18 a.m. | #2
2018-07-12 10:55 GMT+02:00 Greg KH <gregkh@linuxfoundation.org>:
> On Thu, Jul 12, 2018 at 10:06:23AM +0200, Benjamin Gaignard wrote:

>> In some cases the link between between customer and supplier

>> already exist. Do not warn about already existing dependencies

>> because device_link_add() take care of this case.

>

> Why would a link already exist that is asked to be created again?  What

> code path causes this?


It could happen that the link exist because a device use it parent as supplier.
That case has been describe by Marek in this thread (I forgot to add
it in the commit message, sorry):
https://lkml.org/lkml/2018/7/9/356

>

>>

>> Reported-by: Marek Szyprowski <m.szyprowski@samsung.com>

>> Signed-off-by: Benjamin Gaignard <benjamin.gaignard@st.com>

>> ---

>>  drivers/base/core.c | 4 ++--

>>  1 file changed, 2 insertions(+), 2 deletions(-)

>>

>> diff --git a/drivers/base/core.c b/drivers/base/core.c

>> index df3e1a44707a..fcdc17f0f349 100644

>> --- a/drivers/base/core.c

>> +++ b/drivers/base/core.c

>> @@ -105,7 +105,7 @@ static int device_is_dependent(struct device *dev, void *target)

>>       struct device_link *link;

>>       int ret;

>>

>> -     if (WARN_ON(dev == target))

>> +     if (dev == target)

>>               return 1;

>>

>>       ret = device_for_each_child(dev, target, device_is_dependent);

>> @@ -113,7 +113,7 @@ static int device_is_dependent(struct device *dev, void *target)

>>               return ret;

>>

>>       list_for_each_entry(link, &dev->links.consumers, s_node) {

>> -             if (WARN_ON(link->consumer == target))

>> +             if (link->consumer == target)

>>                       return 1;

>

> Both of these WARN_ON are for valid code?  That feels really odd to me,

> I need more explanation here please.


The documentation of the function is clear about return values:
"Check if @target depends on @dev or any device dependent on it (its
child or ts consumer etc).  Return 1 if that is the case or 0
otherwise."
so, for me, not need to warn user about something that is expected.

Benjamin
>

> thanks,

>

> greg k-h
Mark Brown July 12, 2018, 10:25 a.m. | #3
On Thu, Jul 12, 2018 at 11:18:26AM +0200, Benjamin Gaignard wrote:
> 2018-07-12 10:55 GMT+02:00 Greg KH <gregkh@linuxfoundation.org>:


> > Why would a link already exist that is asked to be created again?  What

> > code path causes this?


> It could happen that the link exist because a device use it parent as supplier.

> That case has been describe by Marek in this thread (I forgot to add

> it in the commit message, sorry):

> https://lkml.org/lkml/2018/7/9/356


If we don't remove the warning then we'd need to have some way for
generic code to check if a link that it wants to create exists already
since if more than one thing adds links there's always a chance that two
of them will come up with the same idea for a link.  We could export
device_is_dependent() and have them check that, or add a new flag that
skips the check for example.  This is however probably going to affect a
reasonable percentage of potential users.
Mark Brown July 13, 2018, 4:14 p.m. | #4
On Thu, Jul 12, 2018 at 10:06:23AM +0200, Benjamin Gaignard wrote:
> In some cases the link between between customer and supplier

> already exist. Do not warn about already existing dependencies

> because device_link_add() take care of this case.


Reviwed-by: Mark Brown <broonie@kernel.org>
Rafael J. Wysocki July 16, 2018, 9:45 a.m. | #5
On Thursday, July 12, 2018 11:18:26 AM CEST Benjamin Gaignard wrote:
> 2018-07-12 10:55 GMT+02:00 Greg KH <gregkh@linuxfoundation.org>:

> > On Thu, Jul 12, 2018 at 10:06:23AM +0200, Benjamin Gaignard wrote:

> >> In some cases the link between between customer and supplier

> >> already exist. Do not warn about already existing dependencies

> >> because device_link_add() take care of this case.

> >

> > Why would a link already exist that is asked to be created again?  What

> > code path causes this?

> 

> It could happen that the link exist because a device use it parent as supplier.

> That case has been describe by Marek in this thread (I forgot to add

> it in the commit message, sorry):

> https://lkml.org/lkml/2018/7/9/356


So please add this information to the patch changelog and resend.

Thanks,
Rafael

Patch

diff --git a/drivers/base/core.c b/drivers/base/core.c
index df3e1a44707a..fcdc17f0f349 100644
--- a/drivers/base/core.c
+++ b/drivers/base/core.c
@@ -105,7 +105,7 @@  static int device_is_dependent(struct device *dev, void *target)
 	struct device_link *link;
 	int ret;
 
-	if (WARN_ON(dev == target))
+	if (dev == target)
 		return 1;
 
 	ret = device_for_each_child(dev, target, device_is_dependent);
@@ -113,7 +113,7 @@  static int device_is_dependent(struct device *dev, void *target)
 		return ret;
 
 	list_for_each_entry(link, &dev->links.consumers, s_node) {
-		if (WARN_ON(link->consumer == target))
+		if (link->consumer == target)
 			return 1;
 
 		ret = device_is_dependent(link->consumer, target);