Re: [PATCH net v1 1/1] netfilter: nf_reject: Fix build error when CONFIG_BRIDGE_NETFILTER=n

[Simon Horman <horms@kernel.org>]

On Fri, Sep 06, 2024 at 05:55:13PM +0300, Andy Shevchenko wrote:
> In some cases (CONFIG_BRIDGE_NETFILTER=n) the pointer to IP header
> is set but not used, it prevents kernel builds with clang, `make W=1`
> and CONFIG_WERROR=y:
> 
> ipv6: split nf_send_reset6() in smaller functions
> netfilter: nf_reject_ipv4: split nf_send_reset() in smaller functions
> 
> net/ipv4/netfilter/nf_reject_ipv4.c:243:16: error: variable 'niph' set but not used [-Werror,-Wunused-but-set-variable]
>   243 |         struct iphdr *niph;
>       |                       ^
> net/ipv6/netfilter/nf_reject_ipv6.c:286:18: error: variable 'ip6h' set but not used [-Werror,-Wunused-but-set-variable]
>   286 |         struct ipv6hdr *ip6h;
>       |                         ^
> 
> Fix these by marking respective variables with __maybe_unused as it
> seems more complicated to address that in a better way due to ifdeffery.
> 
> Fixes: 8bfcdf6671b1 ("netfilter: nf_reject_ipv6: split nf_send_reset6() in smaller functions")
> Fixes: 052b9498eea5 ("netfilter: nf_reject_ipv4: split nf_send_reset() in smaller functions")

Hi Andy,

As mentioned in relation to another similar patch,
I'm not sure that resolution of W=1 warnings are fixes.

> Signed-off-by: Andy Shevchenko <andriy.shevchenko@linux.intel.com>
> ---
>  net/ipv4/netfilter/nf_reject_ipv4.c | 2 +-
>  net/ipv6/netfilter/nf_reject_ipv6.c | 2 +-
>  2 files changed, 2 insertions(+), 2 deletions(-)

> 
> diff --git a/net/ipv4/netfilter/nf_reject_ipv4.c b/net/ipv4/netfilter/nf_reject_ipv4.c
> index 04504b2b51df..0af42494ac66 100644
> --- a/net/ipv4/netfilter/nf_reject_ipv4.c
> +++ b/net/ipv4/netfilter/nf_reject_ipv4.c
> @@ -240,7 +240,7 @@ void nf_send_reset(struct net *net, struct sock *sk, struct sk_buff *oldskb,
>  		   int hook)
>  {
>  	struct sk_buff *nskb;
> -	struct iphdr *niph;
> +	struct iphdr *niph __maybe_unused;
>  	const struct tcphdr *oth;
>  	struct tcphdr _oth;
>  

Possibly it is broken for some reason - like reading nskb too late -
but I wonder if rather than annotating niph it's scope can be reduced
to the code that is only compiled if CONFIG_BRIDGE_NETFILTER is enabled.

This also addreses what appears to be an assingment of niph without
the value being used - the first assingment.

E.g., for the ipv4 case (compile tested only!):

diff --git a/net/ipv4/netfilter/nf_reject_ipv4.c b/net/ipv4/netfilter/nf_reject_ipv4.c
index 04504b2b51df..87fd945a0d27 100644
--- a/net/ipv4/netfilter/nf_reject_ipv4.c
+++ b/net/ipv4/netfilter/nf_reject_ipv4.c
@@ -239,9 +239,8 @@ static int nf_reject_fill_skb_dst(struct sk_buff *skb_in)
 void nf_send_reset(struct net *net, struct sock *sk, struct sk_buff *oldskb,
 		   int hook)
 {
-	struct sk_buff *nskb;
-	struct iphdr *niph;
 	const struct tcphdr *oth;
+	struct sk_buff *nskb;
 	struct tcphdr _oth;
 
 	oth = nf_reject_ip_tcphdr_get(oldskb, &_oth, hook);
@@ -266,14 +265,12 @@ void nf_send_reset(struct net *net, struct sock *sk, struct sk_buff *oldskb,
 	nskb->mark = IP4_REPLY_MARK(net, oldskb->mark);
 
 	skb_reserve(nskb, LL_MAX_HEADER);
-	niph = nf_reject_iphdr_put(nskb, oldskb, IPPROTO_TCP,
-				   ip4_dst_hoplimit(skb_dst(nskb)));
+	nf_reject_iphdr_put(nskb, oldskb, IPPROTO_TCP,
+			    ip4_dst_hoplimit(skb_dst(nskb)));
 	nf_reject_ip_tcphdr_put(nskb, oldskb, oth);
 	if (ip_route_me_harder(net, sk, nskb, RTN_UNSPEC))
 		goto free_nskb;
 
-	niph = ip_hdr(nskb);
-
 	/* "Never happens" */
 	if (nskb->len > dst_mtu(skb_dst(nskb)))
 		goto free_nskb;
@@ -290,6 +287,7 @@ void nf_send_reset(struct net *net, struct sock *sk, struct sk_buff *oldskb,
 	 */
 	if (nf_bridge_info_exists(oldskb)) {
 		struct ethhdr *oeth = eth_hdr(oldskb);
+		struct iphdr *niph = ip_hdr(nskb);
 		struct net_device *br_indev;
 
 		br_indev = nf_bridge_get_physindev(oldskb, net);