Thread (10 messages) 10 messages, 2 authors, 2009-07-07

Re: [RFC] Fixing up TCP/UDP checksum for UDP encap. ESP4 packets in transport mode

From: David Miller <davem@davemloft.net>
Date: 2009-07-07 02:02:28

From: Herbert Xu <herbert@gondor.apana.org.au>
Date: Tue, 7 Jul 2009 09:58:50 +0800
On Mon, Jul 06, 2009 at 06:54:11PM -0700, David Miller wrote:
quoted
Hmmm, aren't we talking about packets which were protected by either a
hash, strong encryption, or both at some point?
Only if there is no inner NAT, i.e., only if this patch isn't
needed.  Otherwise

Source  ---  GW1  ----  GW2  ---  Dest

the path between Source and GW1, will be unprotected if transport
mode is used between GW1 and GW2.  The only bit protected by IPsec's
hash is between GW1 and GW2.

When the traffic comes back, then the bit between Dest and GW2 will
be unprotected.  This is why the only safe way to use this would be
if your traffic is one-way and you only had inner NAT at the other
end.
I see.  So people are using transport IPSEC + NAT as a kind of
specialized tunnel.

Indeed, there is no way to handle checksums sanely.  The whole
end-to-end protection of the checksum would be entirely subverted
if we fixed it up.
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help