[PATCH] [media] dvb: avoid warning in dvb_net

[PATCH] [media] dvb: avoid warning in dvb_net

[Arnd Bergmann]

With gcc-5 or higher on x86, we can get a bogus warning in the
dvb-net code:

drivers/media/dvb-core/dvb_net.c: In function ‘dvb_net_ule’:
arch/x86/include/asm/string_32.h:77:14: error: ‘dest_addr’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
drivers/media/dvb-core/dvb_net.c:633:8: note: ‘dest_addr’ was declared here

The problem here is that gcc doesn't track all of the conditions
to prove it can't end up copying uninitialized data.
This changes the logic around so we zero out the destination
address earlier when we determine that it is not set here.
This allows the compiler to figure it out.

Signed-off-by: Arnd Bergmann <arnd@arndb.de>
---
 drivers/media/dvb-core/dvb_net.c | 12 +++++-------
 1 file changed, 5 insertions(+), 7 deletions(-)

diff --git a/drivers/media/dvb-core/dvb_net.c b/drivers/media/dvb-core/dvb_net.c
index 088914c4623f..f1b416de9dab 100644
--- a/drivers/media/dvb-core/dvb_net.c
+++ b/drivers/media/dvb-core/dvb_net.c
@@ -688,6 +688,9 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
 							      ETH_ALEN);
 						skb_pull(priv->ule_skb, ETH_ALEN);
 					}
+				} else {
+					 /* othersie use zero destination address */
+					eth_zero_addr(dest_addr);
 				}
 
 				/* Handle ULE Extension Headers. */
@@ -715,13 +718,8 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
 				if (!priv->ule_bridged) {
 					skb_push(priv->ule_skb, ETH_HLEN);
 					ethh = (struct ethhdr *)priv->ule_skb->data;
-					if (!priv->ule_dbit) {
-						 /* dest_addr buffer is only valid if priv->ule_dbit == 0 */
-						memcpy(ethh->h_dest, dest_addr, ETH_ALEN);
-						eth_zero_addr(ethh->h_source);
-					}
-					else /* zeroize source and dest */
-						memset( ethh, 0, ETH_ALEN*2 );
+					memcpy(ethh->h_dest, dest_addr, ETH_ALEN);
+					eth_zero_addr(ethh->h_source);
 
 					ethh->h_proto = htons(priv->ule_sndu_type);
 				}
-- 
2.9.0

Re: [PATCH] [media] dvb: avoid warning in dvb_net

[Jarod Wilson]

On Thu, Oct 27, 2016 at 03:57:41PM +0200, Arnd Bergmann wrote:
> With gcc-5 or higher on x86, we can get a bogus warning in the
> dvb-net code:
> 
> drivers/media/dvb-core/dvb_net.c: In function ‘dvb_net_ule’:
> arch/x86/include/asm/string_32.h:77:14: error: ‘dest_addr’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
> drivers/media/dvb-core/dvb_net.c:633:8: note: ‘dest_addr’ was declared here
> 
> The problem here is that gcc doesn't track all of the conditions
> to prove it can't end up copying uninitialized data.
> This changes the logic around so we zero out the destination
> address earlier when we determine that it is not set here.
> This allows the compiler to figure it out.
> 
> Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> ---
>  drivers/media/dvb-core/dvb_net.c | 12 +++++-------
>  1 file changed, 5 insertions(+), 7 deletions(-)
> 
> diff --git a/drivers/media/dvb-core/dvb_net.c b/drivers/media/dvb-core/dvb_net.c
> index 088914c4623f..f1b416de9dab 100644
> --- a/drivers/media/dvb-core/dvb_net.c
> +++ b/drivers/media/dvb-core/dvb_net.c
> @@ -688,6 +688,9 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
>  							      ETH_ALEN);
>  						skb_pull(priv->ule_skb, ETH_ALEN);
>  					}
> +				} else {
> +					 /* othersie use zero destination address */

I'm assuming you meant "otherwise" there instead of "othersie".

-- 
Jarod Wilson
jarod@redhat.com

Re: [PATCH] [media] dvb: avoid warning in dvb_net

[Arnd Bergmann]

On Thursday, October 27, 2016 10:13:27 AM CEST Jarod Wilson wrote:
> On Thu, Oct 27, 2016 at 03:57:41PM +0200, Arnd Bergmann wrote:
> > With gcc-5 or higher on x86, we can get a bogus warning in the
> > dvb-net code:
> > 
> > drivers/media/dvb-core/dvb_net.c: In function ‘dvb_net_ule’:
> > arch/x86/include/asm/string_32.h:77:14: error: ‘dest_addr’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
> > drivers/media/dvb-core/dvb_net.c:633:8: note: ‘dest_addr’ was declared here
> > 
> > The problem here is that gcc doesn't track all of the conditions
> > to prove it can't end up copying uninitialized data.
> > This changes the logic around so we zero out the destination
> > address earlier when we determine that it is not set here.
> > This allows the compiler to figure it out.
> > 
> > Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> > ---
> >  drivers/media/dvb-core/dvb_net.c | 12 +++++-------
> >  1 file changed, 5 insertions(+), 7 deletions(-)
> > 
> > diff --git a/drivers/media/dvb-core/dvb_net.c b/drivers/media/dvb-core/dvb_net.c
> > index 088914c4623f..f1b416de9dab 100644
> > --- a/drivers/media/dvb-core/dvb_net.c
> > +++ b/drivers/media/dvb-core/dvb_net.c
> > @@ -688,6 +688,9 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
> >                                                             ETH_ALEN);
> >                                               skb_pull(priv->ule_skb, ETH_ALEN);
> >                                       }
> > +                             } else {
> > +                                      /* othersie use zero destination address */
> 
> I'm assuming you meant "otherwise" there instead of "othersie".
> 

Yes, I sent a v2 now, thanks for taking a look. I assume this means
you have no other objections to the patch?

	Arnd

Re: [PATCH] [media] dvb: avoid warning in dvb_net

[Jarod Wilson]

On Thu, Oct 27, 2016 at 05:09:28PM +0200, Arnd Bergmann wrote:
> On Thursday, October 27, 2016 10:13:27 AM CEST Jarod Wilson wrote:
> > On Thu, Oct 27, 2016 at 03:57:41PM +0200, Arnd Bergmann wrote:
> > > With gcc-5 or higher on x86, we can get a bogus warning in the
> > > dvb-net code:
> > > 
> > > drivers/media/dvb-core/dvb_net.c: In function ‘dvb_net_ule’:
> > > arch/x86/include/asm/string_32.h:77:14: error: ‘dest_addr’ may be used uninitialized in this function [-Werror=maybe-uninitialized]
> > > drivers/media/dvb-core/dvb_net.c:633:8: note: ‘dest_addr’ was declared here
> > > 
> > > The problem here is that gcc doesn't track all of the conditions
> > > to prove it can't end up copying uninitialized data.
> > > This changes the logic around so we zero out the destination
> > > address earlier when we determine that it is not set here.
> > > This allows the compiler to figure it out.
> > > 
> > > Signed-off-by: Arnd Bergmann <arnd@arndb.de>
> > > ---
> > >  drivers/media/dvb-core/dvb_net.c | 12 +++++-------
> > >  1 file changed, 5 insertions(+), 7 deletions(-)
> > > 
> > > diff --git a/drivers/media/dvb-core/dvb_net.c b/drivers/media/dvb-core/dvb_net.c
> > > index 088914c4623f..f1b416de9dab 100644
> > > --- a/drivers/media/dvb-core/dvb_net.c
> > > +++ b/drivers/media/dvb-core/dvb_net.c
> > > @@ -688,6 +688,9 @@ static void dvb_net_ule( struct net_device *dev, const u8 *buf, size_t buf_len )
> > >                                                             ETH_ALEN);
> > >                                               skb_pull(priv->ule_skb, ETH_ALEN);
> > >                                       }
> > > +                             } else {
> > > +                                      /* othersie use zero destination address */
> > 
> > I'm assuming you meant "otherwise" there instead of "othersie".
> > 
> 
> Yes, I sent a v2 now, thanks for taking a look. I assume this means
> you have no other objections to the patch?

No objections, but I don't know enough about ULE or it's handling there
to do an informed critique outside of the typo.

-- 
Jarod Wilson
jarod@redhat.com