2020-08-01 09:45:06

by Ivan Safonov

[permalink] [raw]
Subject: [PATCH] staging: r8188eu: replace rtw_netdev_priv define with inline function

The function guarantees type checking of arguments and return value.

Signed-off-by: Ivan Safonov <[email protected]>
---
drivers/staging/rtl8188eu/include/osdep_service.h | 6 ++++--
1 file changed, 4 insertions(+), 2 deletions(-)

diff --git a/drivers/staging/rtl8188eu/include/osdep_service.h b/drivers/staging/rtl8188eu/include/osdep_service.h
index 31d897f1d21f..e0ccafdea9cd 100644
--- a/drivers/staging/rtl8188eu/include/osdep_service.h
+++ b/drivers/staging/rtl8188eu/include/osdep_service.h
@@ -71,8 +71,10 @@ struct rtw_netdev_priv_indicator {
};
struct net_device *rtw_alloc_etherdev_with_old_priv(void *old_priv);

-#define rtw_netdev_priv(netdev) \
- (((struct rtw_netdev_priv_indicator *)netdev_priv(netdev))->priv)
+static inline struct adapter *rtw_netdev_priv(struct net_device *dev)
+{
+ return (((struct rtw_netdev_priv_indicator *)netdev_priv(dev))->priv);
+}
void rtw_free_netdev(struct net_device *netdev);

#define FUNC_NDEV_FMT "%s(%s)"
--
2.26.2


2020-08-01 09:55:06

by Greg Kroah-Hartman

[permalink] [raw]
Subject: Re: [PATCH] staging: r8188eu: replace rtw_netdev_priv define with inline function

On Sat, Aug 01, 2020 at 12:47:07PM +0300, Ivan Safonov wrote:
> The function guarantees type checking of arguments and return value.
>
> Signed-off-by: Ivan Safonov <[email protected]>
> ---
> drivers/staging/rtl8188eu/include/osdep_service.h | 6 ++++--
> 1 file changed, 4 insertions(+), 2 deletions(-)
>
> diff --git a/drivers/staging/rtl8188eu/include/osdep_service.h b/drivers/staging/rtl8188eu/include/osdep_service.h
> index 31d897f1d21f..e0ccafdea9cd 100644
> --- a/drivers/staging/rtl8188eu/include/osdep_service.h
> +++ b/drivers/staging/rtl8188eu/include/osdep_service.h
> @@ -71,8 +71,10 @@ struct rtw_netdev_priv_indicator {
> };
> struct net_device *rtw_alloc_etherdev_with_old_priv(void *old_priv);
>
> -#define rtw_netdev_priv(netdev) \
> - (((struct rtw_netdev_priv_indicator *)netdev_priv(netdev))->priv)
> +static inline struct adapter *rtw_netdev_priv(struct net_device *dev)
> +{
> + return (((struct rtw_netdev_priv_indicator *)netdev_priv(dev))->priv);
> +}
> void rtw_free_netdev(struct net_device *netdev);

Is the cast really needed?

And no blank line before the next function prototype?

And is this really needed? Type checking is already happening as this
is a "simple" macro, right?

thanks,

greg k-h

2020-08-01 13:08:51

by Ivan Safonov

[permalink] [raw]
Subject: Re: [PATCH] staging: r8188eu: replace rtw_netdev_priv define with inline function

On 8/1/20 12:51 PM, Greg Kroah-Hartman wrote:
> On Sat, Aug 01, 2020 at 12:47:07PM +0300, Ivan Safonov wrote:
>> The function guarantees type checking of arguments and return value.
>>
>> Signed-off-by: Ivan Safonov <[email protected]>
>> ---
>> drivers/staging/rtl8188eu/include/osdep_service.h | 6 ++++--
>> 1 file changed, 4 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/staging/rtl8188eu/include/osdep_service.h b/drivers/staging/rtl8188eu/include/osdep_service.h
>> index 31d897f1d21f..e0ccafdea9cd 100644
>> --- a/drivers/staging/rtl8188eu/include/osdep_service.h
>> +++ b/drivers/staging/rtl8188eu/include/osdep_service.h
>> @@ -71,8 +71,10 @@ struct rtw_netdev_priv_indicator {
>> };
>> struct net_device *rtw_alloc_etherdev_with_old_priv(void *old_priv);
>>
>> -#define rtw_netdev_priv(netdev) \
>> - (((struct rtw_netdev_priv_indicator *)netdev_priv(netdev))->priv)
>> +static inline struct adapter *rtw_netdev_priv(struct net_device *dev)
>> +{
>> + return (((struct rtw_netdev_priv_indicator *)netdev_priv(dev))->priv);
>> +}
>> void rtw_free_netdev(struct net_device *netdev);
>
> Is the cast really needed?

(struct rtw_netdev_priv_indicator *) cast needed for access to '->priv'.

The (struct adapter *) return type is a starting point for simplify
struct adapter *padapter = (struct adapter *)rtw_netdev_priv(dev);
constructions.

>
> And no blank line before the next function prototype?

In v2.

>
> And is this really needed? Type checking is already happening as this
> is a "simple" macro, right?

1. The flexibility of macros is not needed here.
2. The macro silently assigns ->priv to any pointer, while the function
indicates such a compile-time error.

Ivan Safonov.

2020-08-01 13:13:35

by Greg Kroah-Hartman

[permalink] [raw]
Subject: Re: [PATCH] staging: r8188eu: replace rtw_netdev_priv define with inline function

On Sat, Aug 01, 2020 at 04:11:38PM +0300, Ivan Safonov wrote:
> On 8/1/20 12:51 PM, Greg Kroah-Hartman wrote:
> > On Sat, Aug 01, 2020 at 12:47:07PM +0300, Ivan Safonov wrote:
> > > The function guarantees type checking of arguments and return value.
> > >
> > > Signed-off-by: Ivan Safonov <[email protected]>
> > > ---
> > > drivers/staging/rtl8188eu/include/osdep_service.h | 6 ++++--
> > > 1 file changed, 4 insertions(+), 2 deletions(-)
> > >
> > > diff --git a/drivers/staging/rtl8188eu/include/osdep_service.h b/drivers/staging/rtl8188eu/include/osdep_service.h
> > > index 31d897f1d21f..e0ccafdea9cd 100644
> > > --- a/drivers/staging/rtl8188eu/include/osdep_service.h
> > > +++ b/drivers/staging/rtl8188eu/include/osdep_service.h
> > > @@ -71,8 +71,10 @@ struct rtw_netdev_priv_indicator {
> > > };
> > > struct net_device *rtw_alloc_etherdev_with_old_priv(void *old_priv);
> > > -#define rtw_netdev_priv(netdev) \
> > > - (((struct rtw_netdev_priv_indicator *)netdev_priv(netdev))->priv)
> > > +static inline struct adapter *rtw_netdev_priv(struct net_device *dev)
> > > +{
> > > + return (((struct rtw_netdev_priv_indicator *)netdev_priv(dev))->priv);
> > > +}
> > > void rtw_free_netdev(struct net_device *netdev);
> >
> > Is the cast really needed?
>
> (struct rtw_netdev_priv_indicator *) cast needed for access to '->priv'.
>
> The (struct adapter *) return type is a starting point for simplify
> struct adapter *padapter = (struct adapter *)rtw_netdev_priv(dev);
> constructions.

Ah, sorry, missed that I thought this was a cast of netdev_priv(dev)
itself, and not the additional structure.

> >
> > And no blank line before the next function prototype?
>
> In v2.
>
> >
> > And is this really needed? Type checking is already happening as this
> > is a "simple" macro, right?
>
> 1. The flexibility of macros is not needed here.
> 2. The macro silently assigns ->priv to any pointer, while the function
> indicates such a compile-time error.

Ok, put that somewhere in the changelog please, saying how the existing
macro does not provide any error checking.

thanks,

greg k-h