OSDN Git Service

ipv6: Fix dangling pointer when ipv6 fragment
authorJunwei Hu <hujunwei4@huawei.com>
Tue, 2 Apr 2019 11:38:04 +0000 (19:38 +0800)
committerGreg Kroah-Hartman <gregkh@linuxfoundation.org>
Sat, 27 Apr 2019 07:33:54 +0000 (09:33 +0200)
[ Upstream commit ef0efcd3bd3fd0589732b67fb586ffd3c8705806 ]

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: 405c92f7a541 ("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>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
net/ipv6/ip6_output.c

index f8cca81..e39dc94 100644 (file)
@@ -575,7 +575,7 @@ 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;
@@ -586,6 +586,7 @@ 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);
 
@@ -620,6 +621,7 @@ 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)) {
                int first_len = skb_pagelen(skb);