[PATCH v2 net-next] bridge: Fix incorrect judgment of promisc

[PATCH v2 net-next] bridge: Fix incorrect judgment of promisc

[Toshiaki Makita]

br_manage_promisc() incorrectly expects br_auto_port() to return only 0
or 1, while it actually returns flags, i.e., a subset of BR_AUTO_MASK.

Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
---
 net/bridge/br_if.c | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c
index a08d2b8..3eca3fd 100644
--- a/net/bridge/br_if.c
+++ b/net/bridge/br_if.c
@@ -153,7 +153,8 @@ void br_manage_promisc(struct net_bridge *br)
 			 * This lets us disable promiscuous mode and write
 			 * this config to hw.
 			 */
-			if (br->auto_cnt <= br_auto_port(p))
+			if (br->auto_cnt == 0 ||
+			    (br->auto_cnt == 1 && br_auto_port(p)))
 				br_port_clear_promisc(p);
 			else
 				br_port_set_promisc(p);
-- 
1.8.1.2

Re: [PATCH v2 net-next] bridge: Fix incorrect judgment of promisc

[Michael S. Tsirkin]

On Thu, Jun 05, 2014 at 08:53:32PM +0900, Toshiaki Makita wrote:
> br_manage_promisc() incorrectly expects br_auto_port() to return only 0
> or 1, while it actually returns flags, i.e., a subset of BR_AUTO_MASK.
> 
> Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
> ---
>  net/bridge/br_if.c | 3 ++-
>  1 file changed, 2 insertions(+), 1 deletion(-)
> 
> diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c
> index a08d2b8..3eca3fd 100644
> --- a/net/bridge/br_if.c
> +++ b/net/bridge/br_if.c
> @@ -153,7 +153,8 @@ void br_manage_promisc(struct net_bridge *br)
>  			 * This lets us disable promiscuous mode and write
>  			 * this config to hw.
>  			 */
> -			if (br->auto_cnt <= br_auto_port(p))
> +			if (br->auto_cnt == 0 ||
> +			    (br->auto_cnt == 1 && br_auto_port(p)))
>  				br_port_clear_promisc(p);
>  			else
>  				br_port_set_promisc(p);

It's all a nasty side-effect of using macros IMHO.

How about we just make these inline functions returning bool?

The bugfix will fall out naturally.

Signed-off-by: Michael S. Tsirkin <mst@redhat.com>

Warning: untested.


diff --git a/net/bridge/br_private.h b/net/bridge/br_private.h
index 53d6e32..5818dd2 100644
--- a/net/bridge/br_private.h
+++ b/net/bridge/br_private.h
@@ -200,8 +200,15 @@ struct net_bridge_port
 #endif
 };
 
-#define br_auto_port(p) ((p)->flags & BR_AUTO_MASK)
-#define br_promisc_port(p) ((p)->flags & BR_PROMISC)
+static inline bool br_auto_port(struct net_bridge_port *p)
+{
+	return p->flags & BR_AUTO_MASK;
+}
+
+static inline bool br_promisc_port(struct net_bridge_port *p)
+{
+	return p->flags & BR_PROMISC;
+}
 
 #define br_port_exists(dev) (dev->priv_flags & IFF_BRIDGE_PORT)
 

Re: [PATCH v2 net-next] bridge: Fix incorrect judgment of promisc

[Toshiaki Makita]

(2014/06/05 21:30), Michael S. Tsirkin wrote:
> On Thu, Jun 05, 2014 at 08:53:32PM +0900, Toshiaki Makita wrote:
>> br_manage_promisc() incorrectly expects br_auto_port() to return only 0
>> or 1, while it actually returns flags, i.e., a subset of BR_AUTO_MASK.
>>
>> Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
>> ---
>>  net/bridge/br_if.c | 3 ++-
>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>
>> diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c
>> index a08d2b8..3eca3fd 100644
>> --- a/net/bridge/br_if.c
>> +++ b/net/bridge/br_if.c
>> @@ -153,7 +153,8 @@ void br_manage_promisc(struct net_bridge *br)
>>  			 * This lets us disable promiscuous mode and write
>>  			 * this config to hw.
>>  			 */
>> -			if (br->auto_cnt <= br_auto_port(p))
>> +			if (br->auto_cnt == 0 ||
>> +			    (br->auto_cnt == 1 && br_auto_port(p)))
>>  				br_port_clear_promisc(p);
>>  			else
>>  				br_port_set_promisc(p);
> 
> It's all a nasty side-effect of using macros IMHO.
> 
> How about we just make these inline functions returning bool?
> 
> The bugfix will fall out naturally.
> 
> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
> 
> Warning: untested.
> 
> 
> diff --git a/net/bridge/br_private.h b/net/bridge/br_private.h
> index 53d6e32..5818dd2 100644
> --- a/net/bridge/br_private.h
> +++ b/net/bridge/br_private.h
> @@ -200,8 +200,15 @@ struct net_bridge_port
>  #endif
>  };
>  
> -#define br_auto_port(p) ((p)->flags & BR_AUTO_MASK)
> -#define br_promisc_port(p) ((p)->flags & BR_PROMISC)
> +static inline bool br_auto_port(struct net_bridge_port *p)
> +{
> +	return p->flags & BR_AUTO_MASK;
> +}
> +
> +static inline bool br_promisc_port(struct net_bridge_port *p)
> +{
> +	return p->flags & BR_PROMISC;
> +}
>  
>  #define br_port_exists(dev) (dev->priv_flags & IFF_BRIDGE_PORT)

This also looks good.

IMHO, the caller side should not assume these macros (or inline
functions) return boolean value.  There exists similar macro such as
br_port_exists() that doesn't return boolean.

Ohterwise, we should change all macros into boolean functions, but it
might affect performance a little if such a macro is used in fast path?
(I'm worried about the cost of casting non-zero values into 1.)

Thanks,
Toshiaki Makita

Re: [PATCH v2 net-next] bridge: Fix incorrect judgment of promisc

[Vlad Yasevich]

On 06/05/2014 09:01 AM, Toshiaki Makita wrote:
> (2014/06/05 21:30), Michael S. Tsirkin wrote:
>> On Thu, Jun 05, 2014 at 08:53:32PM +0900, Toshiaki Makita wrote:
>>> br_manage_promisc() incorrectly expects br_auto_port() to return only 0
>>> or 1, while it actually returns flags, i.e., a subset of BR_AUTO_MASK.
>>>
>>> Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
>>> ---
>>>  net/bridge/br_if.c | 3 ++-
>>>  1 file changed, 2 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/net/bridge/br_if.c b/net/bridge/br_if.c
>>> index a08d2b8..3eca3fd 100644
>>> --- a/net/bridge/br_if.c
>>> +++ b/net/bridge/br_if.c
>>> @@ -153,7 +153,8 @@ void br_manage_promisc(struct net_bridge *br)
>>>  			 * This lets us disable promiscuous mode and write
>>>  			 * this config to hw.
>>>  			 */
>>> -			if (br->auto_cnt <= br_auto_port(p))
>>> +			if (br->auto_cnt == 0 ||
>>> +			    (br->auto_cnt == 1 && br_auto_port(p)))
>>>  				br_port_clear_promisc(p);
>>>  			else
>>>  				br_port_set_promisc(p);
>>
>> It's all a nasty side-effect of using macros IMHO.
>>
>> How about we just make these inline functions returning bool?
>>
>> The bugfix will fall out naturally.
>>
>> Signed-off-by: Michael S. Tsirkin <mst@redhat.com>
>>
>> Warning: untested.
>>
>>
>> diff --git a/net/bridge/br_private.h b/net/bridge/br_private.h
>> index 53d6e32..5818dd2 100644
>> --- a/net/bridge/br_private.h
>> +++ b/net/bridge/br_private.h
>> @@ -200,8 +200,15 @@ struct net_bridge_port
>>  #endif
>>  };
>>  
>> -#define br_auto_port(p) ((p)->flags & BR_AUTO_MASK)
>> -#define br_promisc_port(p) ((p)->flags & BR_PROMISC)
>> +static inline bool br_auto_port(struct net_bridge_port *p)
>> +{
>> +	return p->flags & BR_AUTO_MASK;
>> +}
>> +
>> +static inline bool br_promisc_port(struct net_bridge_port *p)
>> +{
>> +	return p->flags & BR_PROMISC;
>> +}
>>  
>>  #define br_port_exists(dev) (dev->priv_flags & IFF_BRIDGE_PORT)
> 
> This also looks good.
> 
> IMHO, the caller side should not assume these macros (or inline
> functions) return boolean value.  There exists similar macro such as
> br_port_exists() that doesn't return boolean.
> 
> Ohterwise, we should change all macros into boolean functions, but it
> might affect performance a little if such a macro is used in fast path?
> (I'm worried about the cost of casting non-zero values into 1.)

The following works correctly for me:
#define br_auto_port(p) !!((p)->flags & BR_AUTO_MASK)

Small test shows:

printf("%d\n", br_auto_port(0x20);
1 <-- correct.  learning is set.

printf("%d\n", br_auto_port(0x40);
1 <-- correct.  flooding is set.

printf("%d\n", br_auto_port(0x60);
1 <-- correct, both are set.

printf("%d\n", br_auto_port(0x08);
0 <-- correct.  neither are set.

-vlad

> 
> Thanks,
> Toshiaki Makita
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> 

Re: [PATCH v2 net-next] bridge: Fix incorrect judgment of promisc

[David Miller]

From: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
Date: Thu,  5 Jun 2014 20:53:32 +0900

> br_manage_promisc() incorrectly expects br_auto_port() to return only 0
> or 1, while it actually returns flags, i.e., a subset of BR_AUTO_MASK.
> 
> Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>

I'm applying this as-is for now, even though I saw the other
suggestions in this thread (which BTW didn't get picked up by
patchwork, maybe some of you dropped the Message-Id in your replies by
accident).

RE: [PATCH v2 net-next] bridge: Fix incorrect judgment of promisc

[David Laight]

From: David Miller [mailto:davem@davemloft.net]
> > br_manage_promisc() incorrectly expects br_auto_port() to return only 0
> > or 1, while it actually returns flags, i.e., a subset of BR_AUTO_MASK.
> >
> > Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
> 
> I'm applying this as-is for now, even though I saw the other
> suggestions in this thread (which BTW didn't get picked up by
> patchwork, maybe some of you dropped the Message-Id in your replies by
> accident).

I don't have a problem with that.
The condition looks odd, but it is enabling promiscuous mode
if any other ports are in 'auto' mode.
Possibly the comment above made that clear, but it was truncated
in the diffs.

	David

Re: [PATCH v2 net-next] bridge: Fix incorrect judgment of promisc

[Vlad Yasevich]

On 06/06/2014 04:48 AM, David Laight wrote:
> From: David Miller [mailto:davem@davemloft.net]
>>> br_manage_promisc() incorrectly expects br_auto_port() to return only 0
>>> or 1, while it actually returns flags, i.e., a subset of BR_AUTO_MASK.
>>>
>>> Signed-off-by: Toshiaki Makita <makita.toshiaki@lab.ntt.co.jp>
>>
>> I'm applying this as-is for now, even though I saw the other
>> suggestions in this thread (which BTW didn't get picked up by
>> patchwork, maybe some of you dropped the Message-Id in your replies by
>> accident).
> 
> I don't have a problem with that.
> The condition looks odd, but it is enabling promiscuous mode
> if any other ports are in 'auto' mode.

No, the condition is correct and explicit.  The cases are:
  0 auto ports == all ports are statically configured and non-promisc.
  1 auto port == only this port can be non-promisc.  all others promisc.
  > 1 auto port == all ports promisc.

-vlad

> Possibly the comment above made that clear, but it was truncated
> in the diffs.
> 
> 	David
> 
> 
> 
> --
> To unsubscribe from this list: send the line "unsubscribe netdev" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>