[net-next fragmentation icmp v4 0/4] fragmentation ICMP

[net-next fragmentation icmp v4 0/4] fragmentation ICMP

[Andy Zhou]

Currently, we send ICMP packets when errors occur during fragmentation or
de-fragmentation.  However, it is a bug when sending those ICMP packets
in the context of using netfilter for bridging.

Those ICMP packets are only expected in the context of routing, not in
bridging mode.

The local stack is not involved in bridging forward decisions, thus
should be not used for deciding the reverse path for those ICMP messages.

This bug only affects IPV4, not in IPv6.

---
v1->v2:  restructure the patches into two patches that fix defragmentation and
         fragmentation respectively. 

	 A bit is add in IPCB to control whether ICMP packet should be
	 generated for defragmentation.

	 Fragmentation ICMP is now removed by restructuring the
	 ip_fragment() API.

v2->v3:  Add droping icmp for bridging contrack users
         drop exporting ip_fragment() API.

v3->v4:  Remove unnecessary parentheses in 'return' statements

Andy Zhou (4):
  ipv4: introduce frag_expire_skip_icmp()
  IPv4: skip ICMP for bridge contrack users when defrag expires
  bridge_netfilter: No ICMP packet on IPv4 defragmentation timeout
  bridge_netfilter: No ICMP packet on IPv4 fragmentation error

 include/net/inet_frag.h   |  4 +++-
 include/net/ip.h          | 15 +++++++++++++--
 net/bridge/br_netfilter.c | 26 +++++++++++++++++++++++++-
 net/ipv4/ip_fragment.c    | 25 ++++++++++++++++++++-----
 net/ipv4/ip_output.c      | 40 ++++++++++++++++++++++++++++------------
 5 files changed, 89 insertions(+), 21 deletions(-)

-- 
1.9.1

[net-next fragmentation icmp v4 1/4] ipv4: introduce frag_expire_skip_icmp()

[Andy Zhou]

Improve readability of skip ICMP for de-fragmentation expiration logic.
This change will also make the logic easier to maintain when the
following patches in this series are applied.

Signed-off-by: Andy Zhou <azhou@nicira.com>
---
 include/net/ip.h       | 10 ++++++++++
 net/ipv4/ip_fragment.c | 13 +++++++++----
 2 files changed, 19 insertions(+), 4 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index 0ed6d76..43f6f39 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -478,6 +478,16 @@ enum ip_defrag_users {
 	IP_DEFRAG_MACVLAN,
 };
 
+/* Return true if the value of 'user' is between 'lower_bond'
+ * and 'upper_bond' inclusively.
+ */
+static inline bool ip_defrag_user_in_between(u32 user,
+					     enum ip_defrag_users lower_bond,
+					     enum ip_defrag_users upper_bond)
+{
+	return user >= lower_bond && user <= upper_bond;
+}
+
 int ip_defrag(struct sk_buff *skb, u32 user);
 #ifdef CONFIG_INET
 struct sk_buff *ip_check_defrag(struct sk_buff *skb, u32 user);
diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c
index cc1da6d..83424f1 100644
--- a/net/ipv4/ip_fragment.c
+++ b/net/ipv4/ip_fragment.c
@@ -173,6 +173,13 @@ static void ipq_kill(struct ipq *ipq)
 	inet_frag_kill(&ipq->q, &ip4_frags);
 }
 
+static bool frag_expire_skip_icmp(u32 user)
+{
+	return user == IP_DEFRAG_AF_PACKET ||
+	       ip_defrag_user_in_between(user, IP_DEFRAG_CONNTRACK_IN,
+					 __IP_DEFRAG_CONNTRACK_IN_END);
+}
+
 /*
  * Oops, a fragment queue timed out.  Kill it and send an ICMP reply.
  */
@@ -217,10 +224,8 @@ static void ip_expire(unsigned long arg)
 		/* Only an end host needs to send an ICMP
 		 * "Fragment Reassembly Timeout" message, per RFC792.
 		 */
-		if (qp->user == IP_DEFRAG_AF_PACKET ||
-		    ((qp->user >= IP_DEFRAG_CONNTRACK_IN) &&
-		     (qp->user <= __IP_DEFRAG_CONNTRACK_IN_END) &&
-		     (skb_rtable(head)->rt_type != RTN_LOCAL)))
+		if (frag_expire_skip_icmp(qp->user) &&
+		    (skb_rtable(head)->rt_type != RTN_LOCAL))
 			goto out_rcu_unlock;
 
 		/* Send an ICMP "Fragment Reassembly Timeout" message. */
-- 
1.9.1

[net-next fragmentation icmp v4 2/4] IPv4: skip ICMP for bridge contrack users when defrag expires

[Andy Zhou]

users in [IP_DEFRAG_CONNTRACK_BRIDGE_IN, __IP_DEFRAG_CONNTRACK_BR_IN]
should not ICMP message also.

Reported-by: Florian Westphal <fw@strlen.de>
Signed-off-by: Andy Zhou <azhou@nicira.com>
---
 net/ipv4/ip_fragment.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c
index 83424f1..47fa64e 100644
--- a/net/ipv4/ip_fragment.c
+++ b/net/ipv4/ip_fragment.c
@@ -177,7 +177,9 @@ static bool frag_expire_skip_icmp(u32 user)
 {
 	return user == IP_DEFRAG_AF_PACKET ||
 	       ip_defrag_user_in_between(user, IP_DEFRAG_CONNTRACK_IN,
-					 __IP_DEFRAG_CONNTRACK_IN_END);
+					 __IP_DEFRAG_CONNTRACK_IN_END) ||
+	       ip_defrag_user_in_between(user, IP_DEFRAG_CONNTRACK_BRIDGE_IN,
+					 __IP_DEFRAG_CONNTRACK_BRIDGE_IN);
 }
 
 /*
-- 
1.9.1

[net-next fragmentation icmp v4 3/4] bridge_netfilter: No ICMP packet on IPv4 defragmentation timeout

[Andy Zhou]

Currently, on defragmentation timeout error, ICMP error message
will be generated. This is fine when they are used in a routing context,
but does not make sense in the context of bridging netfilter.

This patch adds a bit (IPSKB_NO_FRAG_ICMP) in IPCB to control
whether ICMP error message should be generated. br_netfiler sets
this bit.

Signed-off-by: Andy Zhou <azhou@nicira.com>
---
 include/net/inet_frag.h   |  4 +++-
 include/net/ip.h          |  1 +
 net/bridge/br_netfilter.c |  5 +++++
 net/ipv4/ip_fragment.c    | 12 ++++++++++--
 4 files changed, 19 insertions(+), 3 deletions(-)

diff --git a/include/net/inet_frag.h b/include/net/inet_frag.h
index 8d17655..e3c8840 100644
--- a/include/net/inet_frag.h
+++ b/include/net/inet_frag.h
@@ -22,12 +22,14 @@ struct netns_frags {
  * @INET_FRAG_LAST_IN: final fragment has arrived
  * @INET_FRAG_COMPLETE: frag queue has been processed and is due for destruction
  * @INET_FRAG_EVICTED: frag queue is being evicted
+ * @INET_FRAG_NO_ICMP: Do not send icmp message on incomplete defrag
  */
 enum {
 	INET_FRAG_FIRST_IN	= BIT(0),
 	INET_FRAG_LAST_IN	= BIT(1),
 	INET_FRAG_COMPLETE	= BIT(2),
-	INET_FRAG_EVICTED	= BIT(3)
+	INET_FRAG_EVICTED	= BIT(3),
+	INET_FRAG_NO_ICMP	= BIT(4)
 };
 
 /**
diff --git a/include/net/ip.h b/include/net/ip.h
index 43f6f39..af44c9f 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -45,6 +45,7 @@ struct inet_skb_parm {
 #define IPSKB_FRAG_COMPLETE	BIT(3)
 #define IPSKB_REROUTED		BIT(4)
 #define IPSKB_DOREDIRECT	BIT(5)
+#define IPSKB_NO_FRAG_ICMP	BIT(6)
 
 	u16			frag_max_size;
 };
diff --git a/net/bridge/br_netfilter.c b/net/bridge/br_netfilter.c
index ab55e24..6a2adba 100644
--- a/net/bridge/br_netfilter.c
+++ b/net/bridge/br_netfilter.c
@@ -663,6 +663,11 @@ static unsigned int br_nf_pre_routing(const struct nf_hook_ops *ops,
 	if (br_parse_ip_options(skb))
 		return NF_DROP;
 
+	/* In case this is a fragmented packet, do not send icmp packet on
+	 * defragmentation error
+	 */
+	IPCB(skb)->flags |= IPSKB_NO_FRAG_ICMP;
+
 	nf_bridge_put(skb->nf_bridge);
 	if (!nf_bridge_alloc(skb))
 		return NF_DROP;
diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c
index 47fa64e..60afaa7 100644
--- a/net/ipv4/ip_fragment.c
+++ b/net/ipv4/ip_fragment.c
@@ -226,7 +226,8 @@ static void ip_expire(unsigned long arg)
 		/* Only an end host needs to send an ICMP
 		 * "Fragment Reassembly Timeout" message, per RFC792.
 		 */
-		if (frag_expire_skip_icmp(qp->user) &&
+		if ((qp->q.flags & INET_FRAG_NO_ICMP) ||
+		    frag_expire_skip_icmp(qp->user) ||
 		    (skb_rtable(head)->rt_type != RTN_LOCAL))
 			goto out_rcu_unlock;
 
@@ -330,6 +331,7 @@ static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
 	int ihl, end;
 	int err = -ENOENT;
 	u8 ecn;
+	bool no_icmp;
 
 	if (qp->q.flags & INET_FRAG_COMPLETE)
 		goto err;
@@ -347,6 +349,7 @@ static int ip_frag_queue(struct ipq *qp, struct sk_buff *skb)
 	offset &= IP_OFFSET;
 	offset <<= 3;		/* offset is in 8-byte chunks */
 	ihl = ip_hdrlen(skb);
+	no_icmp = IPCB(skb)->flags & IPSKB_NO_FRAG_ICMP;
 
 	/* Determine the position of this fragment. */
 	end = offset + skb->len - ihl;
@@ -485,7 +488,12 @@ found:
 	    skb->len + ihl > qp->q.max_size)
 		qp->q.max_size = skb->len + ihl;
 
-	if (qp->q.flags == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
+	if (no_icmp) {
+		qp->q.flags |= INET_FRAG_NO_ICMP;
+	}
+
+	if (((qp->q.flags & ~INET_FRAG_NO_ICMP) ==
+		(INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN)) &&
 	    qp->q.meat == qp->q.len) {
 		unsigned long orefdst = skb->_skb_refdst;
 
-- 
1.9.1

Re: [net-next fragmentation icmp v4 3/4] bridge_netfilter: No ICMP packet on IPv4 defragmentation timeout

[Florian Westphal]

Andy Zhou <azhou@nicira.com> wrote:
> Currently, on defragmentation timeout error, ICMP error message
> will be generated. This is fine when they are used in a routing context,
> but does not make sense in the context of bridging netfilter.
> 
> This patch adds a bit (IPSKB_NO_FRAG_ICMP) in IPCB to control
> whether ICMP error message should be generated. br_netfiler sets
> this bit.

Could you please explain why we need this patch?
After the previous change (patch 2 in your series), we will already
bail out before hitting

/* Send an ICMP "Fragment Reassembly Timeout" message. */
icmp_send(head, ICMP_TIME_EXCEEDED, ICMP_EXC_FRAGTIME, 0);

in ip_expire() based on qp->user value test?

Re: [net-next fragmentation icmp v4 3/4] bridge_netfilter: No ICMP packet on IPv4 defragmentation timeout

[Andy Zhou]

On Thu, May 14, 2015 at 1:59 AM, Florian Westphal <fw@strlen.de> wrote:
> Andy Zhou <azhou@nicira.com> wrote:
>> Currently, on defragmentation timeout error, ICMP error message
>> will be generated. This is fine when they are used in a routing context,
>> but does not make sense in the context of bridging netfilter.
>>
>> This patch adds a bit (IPSKB_NO_FRAG_ICMP) in IPCB to control
>> whether ICMP error message should be generated. br_netfiler sets
>> this bit.
>
> Could you please explain why we need this patch?
> After the previous change (patch 2 in your series), we will already
> bail out before hitting
>
> /* Send an ICMP "Fragment Reassembly Timeout" message. */
> icmp_send(head, ICMP_TIME_EXCEEDED, ICMP_EXC_FRAGTIME, 0);
>
> in ip_expire() based on qp->user value test?

Letting caller making the decision seems to be a better design choice
rather than implicit
logic embedded within the function.  It is also more flexible and
robust when we need to extend 'user'
range down the road.

>

Re: [net-next fragmentation icmp v4 3/4] bridge_netfilter: No ICMP packet on IPv4 defragmentation timeout

[David Miller]

From: Andy Zhou <azhou@nicira.com>
Date: Thu, 14 May 2015 12:54:23 -0700

> On Thu, May 14, 2015 at 1:59 AM, Florian Westphal <fw@strlen.de> wrote:
>> Andy Zhou <azhou@nicira.com> wrote:
>>> Currently, on defragmentation timeout error, ICMP error message
>>> will be generated. This is fine when they are used in a routing context,
>>> but does not make sense in the context of bridging netfilter.
>>>
>>> This patch adds a bit (IPSKB_NO_FRAG_ICMP) in IPCB to control
>>> whether ICMP error message should be generated. br_netfiler sets
>>> this bit.
>>
>> Could you please explain why we need this patch?
>> After the previous change (patch 2 in your series), we will already
>> bail out before hitting
>>
>> /* Send an ICMP "Fragment Reassembly Timeout" message. */
>> icmp_send(head, ICMP_TIME_EXCEEDED, ICMP_EXC_FRAGTIME, 0);
>>
>> in ip_expire() based on qp->user value test?
> 
> Letting caller making the decision seems to be a better design choice
> rather than implicit
> logic embedded within the function.  It is also more flexible and
> robust when we need to extend 'user'
> range down the road.

You're doing useless testing for a hypothetical situation which may
never come to fruitation at all.

I don't think that ever qualifies as a legitimate objection.

Re: [net-next fragmentation icmp v4 3/4] bridge_netfilter: No ICMP packet on IPv4 defragmentation timeout

[Andy Zhou]

O.K. I will drop this patch in the next Rev.

On Thu, May 14, 2015 at 2:42 PM, David Miller <davem@davemloft.net> wrote:
> From: Andy Zhou <azhou@nicira.com>
> Date: Thu, 14 May 2015 12:54:23 -0700
>
>> On Thu, May 14, 2015 at 1:59 AM, Florian Westphal <fw@strlen.de> wrote:
>>> Andy Zhou <azhou@nicira.com> wrote:
>>>> Currently, on defragmentation timeout error, ICMP error message
>>>> will be generated. This is fine when they are used in a routing context,
>>>> but does not make sense in the context of bridging netfilter.
>>>>
>>>> This patch adds a bit (IPSKB_NO_FRAG_ICMP) in IPCB to control
>>>> whether ICMP error message should be generated. br_netfiler sets
>>>> this bit.
>>>
>>> Could you please explain why we need this patch?
>>> After the previous change (patch 2 in your series), we will already
>>> bail out before hitting
>>>
>>> /* Send an ICMP "Fragment Reassembly Timeout" message. */
>>> icmp_send(head, ICMP_TIME_EXCEEDED, ICMP_EXC_FRAGTIME, 0);
>>>
>>> in ip_expire() based on qp->user value test?
>>
>> Letting caller making the decision seems to be a better design choice
>> rather than implicit
>> logic embedded within the function.  It is also more flexible and
>> robust when we need to extend 'user'
>> range down the road.
>
> You're doing useless testing for a hypothetical situation which may
> never come to fruitation at all.
>
> I don't think that ever qualifies as a legitimate objection.

[net-next fragmentation icmp v4 4/4] bridge_netfilter: No ICMP packet on IPv4 fragmentation error

[Andy Zhou]

When bridge netfilter re-fragments an IP packet for output, all
packets that can not be re-fragmented to their original input size
should be silently discarded.

However, current bridge netfilter output path generates an ICMP packet
with 'size exceeded MTU' message for such packets, this is a bug.

This patch refactors the ip_fragment() API to allow two separate
use cases. The bridge netfilter user case will not
send ICMP, the routing output will, as before.

Signed-off-by: Andy Zhou <azhou@nicira.com>
---
 include/net/ip.h          |  4 ++--
 net/bridge/br_netfilter.c | 21 ++++++++++++++++++++-
 net/ipv4/ip_output.c      | 40 ++++++++++++++++++++++++++++------------
 3 files changed, 50 insertions(+), 15 deletions(-)

diff --git a/include/net/ip.h b/include/net/ip.h
index af44c9f..47e62ea 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -109,8 +109,8 @@ int ip_local_deliver(struct sk_buff *skb);
 int ip_mr_input(struct sk_buff *skb);
 int ip_output(struct sock *sk, struct sk_buff *skb);
 int ip_mc_output(struct sock *sk, struct sk_buff *skb);
-int ip_fragment(struct sock *sk, struct sk_buff *skb,
-		int (*output)(struct sock *, struct sk_buff *));
+int ip_do_fragment(struct sock *sk, struct sk_buff *skb,
+		   int (*output)(struct sock *, struct sk_buff *));
 int ip_do_nat(struct sk_buff *skb);
 void ip_send_check(struct iphdr *ip);
 int __ip_local_out(struct sk_buff *skb);
diff --git a/net/bridge/br_netfilter.c b/net/bridge/br_netfilter.c
index 6a2adba..f83a35c 100644
--- a/net/bridge/br_netfilter.c
+++ b/net/bridge/br_netfilter.c
@@ -849,6 +849,25 @@ static int br_nf_push_frag_xmit(struct sock *sk, struct sk_buff *skb)
 	return br_dev_queue_push_xmit(sk, skb);
 }
 
+static int br_nf_ip_fragment(struct sock *sk, struct sk_buff *skb,
+			     int (*output)(struct sock *, struct sk_buff *))
+{
+	unsigned int mtu = ip_skb_dst_mtu(skb);
+	struct iphdr *iph = ip_hdr(skb);
+	struct rtable *rt = skb_rtable(skb);
+	struct net_device *dev = rt->dst.dev;
+
+	if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
+		     (IPCB(skb)->frag_max_size &&
+		      IPCB(skb)->frag_max_size > mtu))) {
+		IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
+		kfree_skb(skb);
+		return -EMSGSIZE;
+	}
+
+	return ip_do_fragment(sk, skb, output);
+}
+
 static int br_nf_dev_queue_xmit(struct sock *sk, struct sk_buff *skb)
 {
 	int ret;
@@ -880,7 +899,7 @@ static int br_nf_dev_queue_xmit(struct sock *sk, struct sk_buff *skb)
 		skb_copy_from_linear_data_offset(skb, -data->size, data->mac,
 						 data->size);
 
-		ret = ip_fragment(sk, skb, br_nf_push_frag_xmit);
+		ret = br_nf_ip_fragment(sk, skb, br_nf_push_frag_xmit);
 	} else {
 		ret = br_dev_queue_push_xmit(sk, skb);
 	}
diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index 2acc5dc..8d91b92 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -83,6 +83,9 @@
 int sysctl_ip_default_ttl __read_mostly = IPDEFTTL;
 EXPORT_SYMBOL(sysctl_ip_default_ttl);
 
+static int ip_fragment(struct sock *sk, struct sk_buff *skb,
+		       int (*output)(struct sock *, struct sk_buff *));
+
 /* Generate a checksum for an outgoing IP datagram. */
 void ip_send_check(struct iphdr *iph)
 {
@@ -478,6 +481,28 @@ static void ip_copy_metadata(struct sk_buff *to, struct sk_buff *from)
 	skb_copy_secmark(to, from);
 }
 
+static int ip_fragment(struct sock *sk, struct sk_buff *skb,
+		       int (*output)(struct sock *, struct sk_buff *))
+{
+	struct iphdr *iph = ip_hdr(skb);
+	unsigned int mtu = ip_skb_dst_mtu(skb);
+
+	if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
+		     (IPCB(skb)->frag_max_size &&
+		      IPCB(skb)->frag_max_size > mtu))) {
+		struct rtable *rt = skb_rtable(skb);
+		struct net_device *dev = rt->dst.dev;
+
+		IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
+		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
+			  htonl(mtu));
+		kfree_skb(skb);
+		return -EMSGSIZE;
+	}
+
+	return ip_do_fragment(sk, skb, output);
+}
+
 /*
  *	This IP datagram is too large to be sent in one piece.  Break it up into
  *	smaller pieces (each of size equal to IP header plus
@@ -485,8 +510,8 @@ static void ip_copy_metadata(struct sk_buff *to, struct sk_buff *from)
  *	single device frame, and queue such a frame for sending.
  */
 
-int ip_fragment(struct sock *sk, struct sk_buff *skb,
-		int (*output)(struct sock *, struct sk_buff *))
+int ip_do_fragment(struct sock *sk, struct sk_buff *skb,
+		   int (*output)(struct sock *, struct sk_buff *))
 {
 	struct iphdr *iph;
 	int ptr;
@@ -507,15 +532,6 @@ int ip_fragment(struct sock *sk, struct sk_buff *skb,
 	iph = ip_hdr(skb);
 
 	mtu = ip_skb_dst_mtu(skb);
-	if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->ignore_df) ||
-		     (IPCB(skb)->frag_max_size &&
-		      IPCB(skb)->frag_max_size > mtu))) {
-		IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
-		icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
-			  htonl(mtu));
-		kfree_skb(skb);
-		return -EMSGSIZE;
-	}
 
 	/*
 	 *	Setup starting values.
@@ -751,7 +767,7 @@ fail:
 	IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
 	return err;
 }
-EXPORT_SYMBOL(ip_fragment);
+EXPORT_SYMBOL(ip_do_fragment);
 
 int
 ip_generic_getfrag(void *from, char *to, int offset, int len, int odd, struct sk_buff *skb)
-- 
1.9.1