Linux Ethernet Bridging - Jul 2021 - [Bridge] [PATCH] netfilter: nf_conntrack_bridge: Fix not free when error

It should be added kfree_skb_list() when err is not equal to zero
in nf_br_ip_fragment().

Fixes: 3c171f496ef5 ("netfilter: bridge: add connection tracking
system")
Signed-off-by: Yajun Deng <yajun.deng at linux.dev>
---
 net/bridge/netfilter/nf_conntrack_bridge.c | 12 ++++++++----
 1 file changed, 8 insertions(+), 4 deletions(-)

diff --git a/net/bridge/netfilter/nf_conntrack_bridge.c
b/net/bridge/netfilter/nf_conntrack_bridge.c
index 8d033a75a766..059f53903eda 100644
--- a/net/bridge/netfilter/nf_conntrack_bridge.c
+++ b/net/bridge/netfilter/nf_conntrack_bridge.c
@@ -83,12 +83,16 @@ static int nf_br_ip_fragment(struct net *net, struct sock
*sk,
 
 			skb->tstamp = tstamp;
 			err = output(net, sk, data, skb);
-			if (err || !iter.frag)
-				break;
-
+			if (err) {
+				kfree_skb_list(iter.frag);
+				return err;
+			}
+
+			if (!iter.frag)
+				return 0;
+
 			skb = ip_fraglist_next(&iter);
 		}
-		return err;
 	}
 slow_path:
 	/* This is a linearized skbuff, the original geometry is lost for us.
-- 
2.32.0

On Mon, Jul 26, 2021 at 11:57:02AM +0800, Yajun Deng
wrote:
> It should be added kfree_skb_list() when err is not equal to zero
> in nf_br_ip_fragment().
> 
> Fixes: 3c171f496ef5 ("netfilter: bridge: add connection tracking
system")
> Signed-off-by: Yajun Deng <yajun.deng at linux.dev>
> ---
>  net/bridge/netfilter/nf_conntrack_bridge.c | 12 ++++++++----
>  1 file changed, 8 insertions(+), 4 deletions(-)
> 
> diff --git a/net/bridge/netfilter/nf_conntrack_bridge.c
b/net/bridge/netfilter/nf_conntrack_bridge.c
> index 8d033a75a766..059f53903eda 100644
> --- a/net/bridge/netfilter/nf_conntrack_bridge.c
> +++ b/net/bridge/netfilter/nf_conntrack_bridge.c
> @@ -83,12 +83,16 @@ static int nf_br_ip_fragment(struct net *net, struct
sock *sk,
>  
>  			skb->tstamp = tstamp;
>  			err = output(net, sk, data, skb);
> -			if (err || !iter.frag)
> -				break;
> -
> +			if (err) {
> +				kfree_skb_list(iter.frag);
> +				return err;
> +			}
> +
> +			if (!iter.frag)
> +				return 0;
> +
>  			skb = ip_fraglist_next(&iter);
>  		}
> -		return err;
Why removing this line above? It enters slow_path: on success.

This patch instead will keep this aligned with IPv6.
>  	}
>  slow_path:
>  	/* This is a linearized skbuff, the original geometry is lost for us.
> -- 
> 2.32.0
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: x.patch
Type: text/x-diff
Size: 460 bytes
Desc: not available
URL:
<http://lists.linuxfoundation.org/pipermail/bridge/attachments/20210728/cf34880e/attachment.bin>

July 29, 2021 12:18 AM, "Pablo Neira Ayuso" <pablo at
netfilter.org> wrote:
> On Mon, Jul 26, 2021 at 11:57:02AM +0800, Yajun Deng wrote:
> 
>> It should be added kfree_skb_list() when err is not equal to zero
>> in nf_br_ip_fragment().
>> 
>> Fixes: 3c171f496ef5 ("netfilter: bridge: add connection tracking
system")
>> Signed-off-by: Yajun Deng <yajun.deng at linux.dev>
>> ---
>> net/bridge/netfilter/nf_conntrack_bridge.c | 12 ++++++++----
>> 1 file changed, 8 insertions(+), 4 deletions(-)
>> 
>> diff --git a/net/bridge/netfilter/nf_conntrack_bridge.c
>> b/net/bridge/netfilter/nf_conntrack_bridge.c
>> index 8d033a75a766..059f53903eda 100644
>> --- a/net/bridge/netfilter/nf_conntrack_bridge.c
>> +++ b/net/bridge/netfilter/nf_conntrack_bridge.c
>> @@ -83,12 +83,16 @@ static int nf_br_ip_fragment(struct net *net,
struct sock *sk,
>> 
>> skb->tstamp = tstamp;
>> err = output(net, sk, data, skb);
>> - if (err || !iter.frag)
>> - break;
>> -
>> + if (err) {
>> + kfree_skb_list(iter.frag);
>> + return err;
>> + }
>> +
>> + if (!iter.frag)
>> + return 0;
>> +
>> skb = ip_fraglist_next(&iter);
>> }
>> - return err;
> 
> Why removing this line above? It enters slow_path: on success.
> 
I used return rather than break, it wouldn't enter the
slow_path.
> This patch instead will keep this aligned with IPv6.
> 
I think err and !iter.frag are not related, there is no need to put them in an
if statement,
We still need to separate them after loop. So I separate them in loop and use
return instead
of break. In addition, if you insist, I will accept your
patch.
>> }
>> slow_path:
>> /* This is a linearized skbuff, the original geometry is lost for us.
>> --
>> 2.32.0