ipv6: Fix dangling pointer when ipv6 fragment

At the beginning of ip6_fragment func, the prevhdr pointer is
obtained in the ip6_find_1stfragopt func.
However, all the pointers pointing into skb header may change
when calling skb_checksum_help func with
skb->ip_summed = CHECKSUM_PARTIAL condition.
The prevhdr pointe will be dangling if it is not reloaded after
calling __skb_linearize func in skb_checksum_help func.

Here, I add a variable, nexthdr_offset, to evaluate the offset,
which does not changes even after calling __skb_linearize func.

Fixes: 405c92f ("ipv6: add defensive check for CHECKSUM_PARTIAL skbs in ip_fragment")
Signed-off-by: Junwei Hu <hujunwei4@huawei.com>
Reported-by: Wenhao Zhang <zhangwenhao8@huawei.com>
Reported-by: syzbot+e8ce541d095e486074fc@syzkaller.appspotmail.com
Reviewed-by: Zhiqiang Liu <liuzhiqiang26@huawei.com>
Acked-by: Martin KaFai Lau <kafai@fb.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

net/ipv6/ip6_output.c

@@ -601,7 +601,7 @@ int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
 				inet6_sk(skb->sk) : NULL;
 	struct ipv6hdr *tmp_hdr;
 	struct frag_hdr *fh;
-	unsigned int mtu, hlen, left, len;
+	unsigned int mtu, hlen, left, len, nexthdr_offset;
 	int hroom, troom;
 	__be32 frag_id;
 	int ptr, offset = 0, err = 0;
@@ -612,6 +612,7 @@ int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
 		goto fail;
 	hlen = err;
 	nexthdr = *prevhdr;
+	nexthdr_offset = prevhdr - skb_network_header(skb);
 
 	mtu = ip6_skb_dst_mtu(skb);
 
@@ -646,6 +647,7 @@ int ip6_fragment(struct net *net, struct sock *sk, struct sk_buff *skb,
 	    (err = skb_checksum_help(skb)))
 		goto fail;
 
+	prevhdr = skb_network_header(skb) + nexthdr_offset;
 	hroom = LL_RESERVED_SPACE(rt->dst.dev);
 	if (skb_has_frag_list(skb)) {
 		unsigned int first_len = skb_pagelen(skb);