On 6/3/19 11:59 PM, Yang Xiao wrote:

Sorry, I don't get your point. Why is xfrm6_transport_output() buggy?
The point is that there would be out-of-bound access in
mip6_destopt_offset() and mip6_destopt_offset(), since there is no
sanity check for offset.

There is chance that offset + sizeof(struct ipv6_opt_hdr) > packet_len.

As described in CVE-2017-9074:  "The IPv6 fragmentation implementation
in the Linux kernel through 4.11.1 does not consider that the nexthdr
field may be associated with an invalid option, which allows local
users to cause a denial of service (out-of-bounds read and BUG)".

At the same time, there are bugs in  mip6_destopt_offset() and
mip6_destopt_offset(), which is similar to CVE-2017-7542.

I suggest that you stop the nonsense.

As explained by Herbert, your patch is not needed at all.

If this was needed, then we would have to fix the callers, which you did not.

Citing arbitrary CVE is of no use, we do not copy/paste patches or CVE.

On Sat, Jun 1, 2019 at 1:35 AM Eric Dumazet [off-list ref] wrote:

quoted

On 5/30/19 8:04 PM, Yang Xiao wrote:

quoted

On Fri, May 31, 2019 at 1:17 AM Eric Dumazet [off-list ref] wrote:

quoted

On 5/30/19 8:28 AM, Young Xiao wrote:

quoted

The fragmentation code tries to parse the header options in order
to figure out where to insert the fragment option.  Since nexthdr points
to an invalid option, the calculation of the size of the network header
can made to be much larger than the linear section of the skb and data
is read outside of it.

This vulnerability is similar to CVE-2017-9074.

Signed-off-by: Young Xiao <redacted>
---
 net/ipv6/mip6.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)

diff --git a/net/ipv6/mip6.c b/net/ipv6/mip6.c
index 64f0f7b..30ed1c5 100644
--- a/net/ipv6/mip6.c
+++ b/net/ipv6/mip6.c

@@ -263,8 +263,6 @@ static int mip6_destopt_offset(struct xfrm_state *x, struct sk_buff *skb,
                             u8 **nexthdr)
 {
      u16 offset = sizeof(struct ipv6hdr);
-     struct ipv6_opt_hdr *exthdr =
-                                (struct ipv6_opt_hdr *)(ipv6_hdr(skb) + 1);
      const unsigned char *nh = skb_network_header(skb);
      unsigned int packet_len = skb_tail_pointer(skb) -
              skb_network_header(skb);

@@ -272,7 +270,8 @@ static int mip6_destopt_offset(struct xfrm_state *x, struct sk_buff *skb,

      *nexthdr = &ipv6_hdr(skb)->nexthdr;

-     while (offset + 1 <= packet_len) {
+     while (offset <= packet_len) {
+             struct ipv6_opt_hdr *exthdr;

              switch (**nexthdr) {
              case NEXTHDR_HOP:

@@ -299,12 +298,15 @@ static int mip6_destopt_offset(struct xfrm_state *x, struct sk_buff *skb,
                      return offset;
              }

+             if (offset + sizeof(struct ipv6_opt_hdr) > packet_len)
+                     return -EINVAL;
+
+             exthdr = (struct ipv6_opt_hdr *)(nh + offset);
              offset += ipv6_optlen(exthdr);
              *nexthdr = &exthdr->nexthdr;
-             exthdr = (struct ipv6_opt_hdr *)(nh + offset);
      }

-     return offset;
+     return -EINVAL;
 }

Ok, but have you checked that callers have been fixed ?

I've checked the callers. There are two callers:
xfrm6_transport_output() and xfrm6_ro_output(). There are checks in
both function.

------------------------------------------------------------------------------
        hdr_len = x->type->hdr_offset(x, skb, &prevhdr);
        if (hdr_len < 0)
                return hdr_len;
------------------------------------------------------------------------------

quoted

xfrm6_transport_output() seems buggy as well,
unless the skbs are linearized before entering these functions ?

I can not understand what you mean about this comment.
Could you explain it in more detail.

If we had a problem, then the memmove(ipv6_hdr(skb), iph, hdr_len);
 in xfrm6_transport_output() would be buggy, since iph could also point to freed memory.