Skip to contentSkip to conversation

net/ipv4: return the real errno instead of -EINVAL

PATCH In review [email protected] 3 participants 3 messages Aug 4, 2023 – Aug 7, 2023

Subsystems (3)
lore
Share & tools
lore
#0 lore raw

PATCH 1/1 net/ipv4: return the real errno instead of -EINVAL

From: xu xin <[email protected]>

For now, no matter what error pointer ip_neigh_for_gw() returns, ip_finish_output2() always return -EINVAL, which may mislead the upper users.

For exemple, an application uses sendto to send an UDP packet, but when the neighbor table overflows, sendto() will get a value of -EINVAL, and it will cause users to waste a lot of time checking parameters for errors.

Return the real errno instead of -EINVAL.

---
 net/ipv4/ip_output.c | 9 ++++-----
 1 file changed, 4 insertions(+), 5 deletions(-)
net/ipv4/ip_output.c +4 -5
index 6ba1a0fafbaa..2d7cf083dff9 100644
@@ -201,6 +201,7 @@ static int ip_finish_output2(struct net *net, struct sock *sk, struct sk_buff *s
 	unsigned int hh_len = LL_RESERVED_SPACE(dev);
 	struct neighbour *neigh;
 	bool is_v6gw = false;
+	int res;
 
 	if (rt->rt_type == RTN_MULTICAST) {
 		IP_UPD_PO_STATS(net, IPSTATS_MIB_OUTMCAST, skb->len);
@@ -214,8 +215,7 @@ static int ip_finish_output2(struct net *net, struct sock *sk, struct sk_buff *s
 	}
 
 	if (lwtunnel_xmit_redirect(dst->lwtstate)) {
-		int res = lwtunnel_xmit(skb);
-
+		res = lwtunnel_xmit(skb);
 		if (res < 0 || res == LWTUNNEL_XMIT_DONE)
 			return res;
 	}
@@ -223,8 +223,6 @@ static int ip_finish_output2(struct net *net, struct sock *sk, struct sk_buff *s
 	rcu_read_lock();
 	neigh = ip_neigh_for_gw(rt, skb, &is_v6gw);
 	if (!IS_ERR(neigh)) {
-		int res;
-
 		sock_confirm_neigh(skb, neigh);
 		/* if crossing protocols, can not use the cached header */
 		res = neigh_output(neigh, skb, is_v6gw);
@@ -236,7 +234,8 @@ static int ip_finish_output2(struct net *net, struct sock *sk, struct sk_buff *s
 	net_dbg_ratelimited("%s: No header cache and no neighbour!\n",
 			    __func__);
 	kfree_skb_reason(skb, SKB_DROP_REASON_NEIGH_CREATEFAIL);
-	return -EINVAL;
+	res = PTR_ERR(neigh);
+	return res;
 }
 
 static int ip_finish_output_gso(struct net *net, struct sock *sk,
Signature · 1 line
2.15.2
#1 lore raw
Quoted 53 lines · On 04/08/2023 09:48, [email protected] wrote:

From: xu xin <[email protected]>

For now, no matter what error pointer ip_neigh_for_gw() returns, ip_finish_output2() always return -EINVAL, which may mislead the upper users.

For exemple, an application uses sendto to send an UDP packet, but when the neighbor table overflows, sendto() will get a value of -EINVAL, and it will cause users to waste a lot of time checking parameters for errors.

Return the real errno instead of -EINVAL.

---
  net/ipv4/ip_output.c | 9 ++++-----
  1 file changed, 4 insertions(+), 5 deletions(-)
net/ipv4/ip_output.c +2 -4
index 6ba1a0fafbaa..2d7cf083dff9 100644
@@ -201,6 +201,7 @@ static int ip_finish_output2(struct net *net, struct sock *sk, struct sk_buff *s
  	unsigned int hh_len = LL_RESERVED_SPACE(dev);
  	struct neighbour *neigh;
  	bool is_v6gw = false;
+	int res;
 
  	if (rt->rt_type == RTN_MULTICAST) {
  		IP_UPD_PO_STATS(net, IPSTATS_MIB_OUTMCAST, skb->len);
@@ -214,8 +215,7 @@ static int ip_finish_output2(struct net *net, struct sock *sk, struct sk_buff *s
  	}
 
  	if (lwtunnel_xmit_redirect(dst->lwtstate)) {
-		int res = lwtunnel_xmit(skb);
-
+		res = lwtunnel_xmit(skb);
  		if (res < 0 || res == LWTUNNEL_XMIT_DONE)
  			return res;
  	}
@@ -223,8 +223,6 @@ static int ip_finish_output2(struct net *net, struct sock *sk, struct sk_buff *s
  	rcu_read_lock();
  	neigh = ip_neigh_for_gw(rt, skb, &is_v6gw);
  	if (!IS_ERR(neigh)) {
-		int res;
-
  		sock_confirm_neigh(skb, neigh);
  		/* if crossing protocols, can not use the cached header */
  		res = neigh_output(neigh, skb, is_v6gw);

net/ipv4/ip_output.c static int ip_finish_output2(struct net *net, struct sock *sk, struct sk_buff *s · line 238

Quoted 6 lines
net/ipv4/ip_output.c +2 -1
@@ -236,7 +234,8 @@ static int ip_finish_output2(struct net *net, struct sock *sk, struct sk_buff *s
  	net_dbg_ratelimited("%s: No header cache and no neighbour!\n",
  			    __func__);
  	kfree_skb_reason(skb, SKB_DROP_REASON_NEIGH_CREATEFAIL);
-	return -EINVAL;
+	res = PTR_ERR(neigh);
+	return res;

I believe this part could be written as

- return -EINVAL;
+ return PTR_ERR(neigh);

and there is no need to change the code is other places. This will be easier to track/backport.

Quoted 3 lines
net/ipv4/ip_output.c +0 -0
  }
 
  static int ip_finish_output_gso(struct net *net, struct sock *sk,
#2 lore raw
Quoted 14 lines

net/ipv4/ip_output.c static int ip_finish_output2(struct net *net, struct sock *sk, struct sk_buff *s · line 238

Quoted 7 lines
net/ipv4/ip_output.c +2 -1
@@ -236,7 +234,8 @@ static int ip_finish_output2(struct net *net, struct sock *sk, struct sk_buff *s
  	net_dbg_ratelimited("%s: No header cache and no neighbour!\n",
  			    __func__);
  	kfree_skb_reason(skb, SKB_DROP_REASON_NEIGH_CREATEFAIL);
-	return -EINVAL;
+	res = PTR_ERR(neigh);
+	return res;

I believe this part could be written as

(quoted diff) +1 -1
-	return -EINVAL;
+	return PTR_ERR(neigh);
Quoted 2 lines

and there is no need to change the code is other places. This will be easier to track/backport.

Sure, I initially wrote the same thing with you, but considering the convenience of maintainer review, I deliberately modified some other parts of the code to present a clear context of the core modifications.

From your reply, I can see you agree with my key idea of this patch, so I have sent a v2 patch[1] according to your view.

[1] https://lore.kernel.org/all/[email protected]/