Re: [PATCH] hid:Add missing error checking in mt_probe for creating sysfs_create_group

Re: [PATCH] hid:Add missing error checking in mt_probe for creating sysfs_create_group

[Benjamin Tissoires]

On Feb 02 2016 or thereabouts, Nicholas Krause wrote:
> This adds the missing error check in the function mt_probe for
> the call to the function sysfs_create_group for creating the
> group,mt_attrribute as based on the based input device's device
> kobject's member as otherwise we assume the sysfs directory is
> created successfully no matter the return value of
> sysfs_create_group.
> 
> Signed-off-by: Nicholas Krause <xerofoify@gmail.com>
> ---
>  drivers/hid/hid-multitouch.c | 2 ++
>  1 file changed, 2 insertions(+)
> 
> diff --git a/drivers/hid/hid-multitouch.c b/drivers/hid/hid-multitouch.c
> index 296d499..46ebf06 100644
> --- a/drivers/hid/hid-multitouch.c
> +++ b/drivers/hid/hid-multitouch.c
> @@ -1133,6 +1133,8 @@ static int mt_probe(struct hid_device *hdev, const struct hid_device_id *id)
>  		return ret;
>  
>  	ret = sysfs_create_group(&hdev->dev.kobj, &mt_attribute_group);
> +	if (ret)
> +		return ret;

I am not particularly happy with a hard failure here in case the sysfs
group doesn't bind. The group is mainly used for debug, and I am not
sure anybody but me uses it nowadays.

Could you amend the patch to put a warning in case of a failure but
still continue?

Cheers,
Benjamin

>  
>  	mt_set_maxcontacts(hdev);
>  	mt_set_input_mode(hdev);
> -- 
> 2.5.0
> 

Re: [PATCH] hid:Add missing error checking in mt_probe for creating sysfs_create_group

[Jiri Kosina]

On Wed, 3 Feb 2016, Benjamin Tissoires wrote:

> > diff --git a/drivers/hid/hid-multitouch.c b/drivers/hid/hid-multitouch.c
> > index 296d499..46ebf06 100644
> > --- a/drivers/hid/hid-multitouch.c
> > +++ b/drivers/hid/hid-multitouch.c
> > @@ -1133,6 +1133,8 @@ static int mt_probe(struct hid_device *hdev, const struct hid_device_id *id)
> >  		return ret;
> >  
> >  	ret = sysfs_create_group(&hdev->dev.kobj, &mt_attribute_group);
> > +	if (ret)
> > +		return ret;
> 
> I am not particularly happy with a hard failure here in case the sysfs
> group doesn't bind. The group is mainly used for debug, and I am not
> sure anybody but me uses it nowadays.
> 
> Could you amend the patch to put a warning in case of a failure but
> still continue?

Yeah. And also in case of hard failure, some cleanup would have to be done 
anyway (hid_hw_start() has already happened, etc).

-- 
Jiri Kosina
SUSE Labs