ipv4: fix path MTU discovery with connection tracking

IPv4 conntrack defragments incoming packet at the PRE_ROUTING hook and
(in case of forwarded packets) refragments them at POST_ROUTING
independent of the IP_DF flag. Refragmentation uses the dst_mtu() of
the local route without caring about the original fragment sizes,
thereby breaking PMTUD.

This patch fixes this by keeping track of the largest received fragment
with IP_DF set and generates an ICMP fragmentation required error during
refragmentation if that size exceeds the MTU.

Signed-off-by: Patrick McHardy <kaber@trash.net>
Acked-by: Eric Dumazet <edumazet@google.com>
Acked-by: David S. Miller <davem@davemloft.net>

4 files changed, 14 insertions, 2 deletions

diff --git a/include/net/inet_frag.h b/include/net/inet_frag.h
index 2431cf83aeca..5098ee7b7e0e 100644
--- a/include/net/inet_frag.h
+++ b/include/net/inet_frag.h
@@ -29,6 +29,8 @@ struct inet_frag_queue {
 #define INET_FRAG_COMPLETE      4
 #define INET_FRAG_FIRST_IN      2
 #define INET_FRAG_LAST_IN       1
+
+        u16                     max_size;
 };
 
 #define INETFRAGS_HASHSZ                64
diff --git a/include/net/ip.h b/include/net/ip.h
index 5a5d84d3d2c6..0707fb9551aa 100644
--- a/include/net/ip.h
+++ b/include/net/ip.h
@@ -42,6 +42,8 @@ struct inet_skb_parm {
 #define IPSKB_XFRM_TRANSFORMED  4
 #define IPSKB_FRAG_COMPLETE     8
 #define IPSKB_REROUTED          16
+
+        u16                     frag_max_size;
 };
 
 static inline unsigned int ip_hdrlen(const struct sk_buff *skb)
diff --git a/net/ipv4/ip_fragment.c b/net/ipv4/ip_fragment.c
index 8d07c973409c..fa6a12c51066 100644
--- a/net/ipv4/ip_fragment.c
+++ b/net/ipv4/ip_fragment.c
@@ -523,6 +523,10 @@ found:
         if (offset == 0)
                 qp->q.last_in |= INET_FRAG_FIRST_IN;
 
+        if (ip_hdr(skb)->frag_off & htons(IP_DF) &&
+            skb->len + ihl > qp->q.max_size)
+                qp->q.max_size = skb->len + ihl;
+
         if (qp->q.last_in == (INET_FRAG_FIRST_IN | INET_FRAG_LAST_IN) &&
             qp->q.meat == qp->q.len)
                 return ip_frag_reasm(qp, prev, dev);
@@ -646,9 +650,11 @@ static int ip_frag_reasm(struct ipq *qp, struct sk_buff *prev,
         head->next = NULL;
         head->dev = dev;
         head->tstamp = qp->q.stamp;
+        IPCB(head)->frag_max_size = qp->q.max_size;
 
         iph = ip_hdr(head);
-        iph->frag_off = 0;
+        /* max_size != 0 implies at least one fragment had IP_DF set */
+        iph->frag_off = qp->q.max_size ? htons(IP_DF) : 0;
         iph->tot_len = htons(len);
         iph->tos |= ecn;
         IP_INC_STATS_BH(net, IPSTATS_MIB_REASMOKS);
diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index c196d749daf2..a5beab1dc958 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -467,7 +467,9 @@ int ip_fragment(struct sk_buff *skb, int (*output)(struct sk_buff *))
 
         iph = ip_hdr(skb);
 
-        if (unlikely((iph->frag_off & htons(IP_DF)) && !skb->local_df)) {
+        if (unlikely(((iph->frag_off & htons(IP_DF)) && !skb->local_df) ||
+                     (IPCB(skb)->frag_max_size &&
+                      IPCB(skb)->frag_max_size > dst_mtu(&rt->dst)))) {
                 IP_INC_STATS(dev_net(dev), IPSTATS_MIB_FRAGFAILS);
                 icmp_send(skb, ICMP_DEST_UNREACH, ICMP_FRAG_NEEDED,
                           htonl(ip_skb_dst_mtu(skb)));