[PATCH] udp: avoid ufo handling on IP payload compression packets

commit c146066ab802 ("ipv4: Don't use ufo handling on later transformed
packets") and commit f89c56ce710a ("ipv6: Don't use ufo handling on
later transformed packets") added a check that 'rt->dst.header_len' isn't
zero in order to skip UFO, but it doesn't include IPcomp in transport mode
where it equals zero.

Packets, after payload compression, may not require further fragmentation,
and if original length exceeds MTU, later compressed packets will be
transmitted incorrectly. This can be reproduced with LTP udp_ipsec.sh test
on veth device with enabled UFO, MTU is 1500 and UDP payload is 2000:

* IPv4 case, offset is wrong + unnecessary fragmentation
    udp_ipsec.sh -p comp -m transport -s 2000 &
    tcpdump -ni ltp_ns_veth2
    ...
    IP (tos 0x0, ttl 64, id 45203, offset 0, flags [+],
      proto Compressed IP (108), length 49)
      10.0.0.2 > 10.0.0.1: IPComp(cpi=0x1000)
    IP (tos 0x0, ttl 64, id 45203, offset 1480, flags [none],
      proto UDP (17), length 21) 10.0.0.2 > 10.0.0.1: ip-proto-17

* IPv6 case, sending small fragments
    udp_ipsec.sh -6 -p comp -m transport -s 2000 &
    tcpdump -ni ltp_ns_veth2
    ...
    IP6 (flowlabel 0x6b9ba, hlim 64, next-header Compressed IP (108)
      payload length: 37) fd00::2 > fd00::1: IPComp(cpi=0x1000)
    IP6 (flowlabel 0x6b9ba, hlim 64, next-header Compressed IP (108)
      payload length: 21) fd00::2 > fd00::1: IPComp(cpi=0x1000)

Fix it by checking 'rt->dst.xfrm' pointer to 'xfrm_state' struct, skip UFO
if xfrm is set. So the new check will include both cases: IPcomp and IPsec.

Fixes: c146066ab802 ("ipv4: Don't use ufo handling on later transformed packets")
Fixes: f89c56ce710a ("ipv6: Don't use ufo handling on later transformed packets")
Signed-off-by: Alexey Kodanev <redacted>
---
 net/ipv4/ip_output.c  |    5 ++++-
 net/ipv6/ip6_output.c |    5 ++++-
 2 files changed, 8 insertions(+), 2 deletions(-)

diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index b67719f..18383ef 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c

@@ -960,7 +960,10 @@ static int __ip_append_data(struct sock *sk,
 	cork->length += length;
 	if ((((length + fragheaderlen) > mtu) || (skb && skb_is_gso(skb))) &&
 	    (sk->sk_protocol == IPPROTO_UDP) &&
-	    (rt->dst.dev->features & NETIF_F_UFO) && !rt->dst.header_len &&
+	    (rt->dst.dev->features & NETIF_F_UFO) &&
+#ifdef CONFIG_XFRM
+	    !rt->dst.xfrm &&
+#endif
 	    (sk->sk_type == SOCK_DGRAM) && !sk->sk_no_check_tx) {
 		err = ip_ufo_append_data(sk, queue, getfrag, from, length,
 					 hh_len, fragheaderlen, transhdrlen,

diff --git a/net/ipv6/ip6_output.c b/net/ipv6/ip6_output.c
index 7cebee5..2a4d5ab 100644
--- a/net/ipv6/ip6_output.c
+++ b/net/ipv6/ip6_output.c

@@ -1381,7 +1381,10 @@ static int __ip6_append_data(struct sock *sk,
 	if ((((length + fragheaderlen) > mtu) ||
 	     (skb && skb_is_gso(skb))) &&
 	    (sk->sk_protocol == IPPROTO_UDP) &&
-	    (rt->dst.dev->features & NETIF_F_UFO) && !rt->dst.header_len &&
+	    (rt->dst.dev->features & NETIF_F_UFO) &&
+#ifdef CONFIG_XFRM
+	    !rt->dst.xfrm &&
+#endif
 	    (sk->sk_type == SOCK_DGRAM) && !udp_get_no_check6_tx(sk)) {
 		err = ip6_ufo_append_data(sk, queue, getfrag, from, length,
 					  hh_len, fragheaderlen, exthdrlen,

-- 
1.7.1

On Fri, Mar 03, 2017 at 03:37:32PM +0300, Alexey Kodanev wrote:

commit c146066ab802 ("ipv4: Don't use ufo handling on later transformed
packets") and commit f89c56ce710a ("ipv6: Don't use ufo handling on
later transformed packets") added a check that 'rt->dst.header_len' isn't
zero in order to skip UFO, but it doesn't include IPcomp in transport mode
where it equals zero.

IPcomp has an additional header and should better set rt->dst.header_len
instead of hacking around that internally.

Please fix IPcomp to use rt->dst.header_len instead off adding
this ifdef to the generic networking code.

On Mon, Mar 06, 2017 at 07:16:57AM +0100, Steffen Klassert wrote:

quoted

diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index b67719f..18383ef 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c

@@ -960,7 +960,10 @@ static int __ip_append_data(struct sock *sk,
 	cork->length += length;
 	if ((((length + fragheaderlen) > mtu) || (skb && skb_is_gso(skb))) &&
 	    (sk->sk_protocol == IPPROTO_UDP) &&
-	    (rt->dst.dev->features & NETIF_F_UFO) && !rt->dst.header_len &&
+	    (rt->dst.dev->features & NETIF_F_UFO) &&
+#ifdef CONFIG_XFRM
+	    !rt->dst.xfrm &&
+#endif

Please fix IPcomp to use rt->dst.header_len instead off adding
this ifdef to the generic networking code.

It's not that simple though.  IPComp's header_len is set to zero
because we opportunistically drop the IPComp header when the total
compressed length exceeds the original packet length.  That is,
we only ever do IPComp when it does not cause the packet to expand.

So it seems that we need another way of indicating the presence of
XFRM.

Cheers,
-- 
Email: Herbert Xu [off-list ref]
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt

On Tue, Mar 07, 2017 at 03:01:50PM +0800, Herbert Xu wrote:

On Mon, Mar 06, 2017 at 07:16:57AM +0100, Steffen Klassert wrote:

quoted

quoted

diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index b67719f..18383ef 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c

@@ -960,7 +960,10 @@ static int __ip_append_data(struct sock *sk,
 	cork->length += length;
 	if ((((length + fragheaderlen) > mtu) || (skb && skb_is_gso(skb))) &&
 	    (sk->sk_protocol == IPPROTO_UDP) &&
-	    (rt->dst.dev->features & NETIF_F_UFO) && !rt->dst.header_len &&
+	    (rt->dst.dev->features & NETIF_F_UFO) &&
+#ifdef CONFIG_XFRM
+	    !rt->dst.xfrm &&
+#endif

Please fix IPcomp to use rt->dst.header_len instead off adding
this ifdef to the generic networking code.

It's not that simple though.  IPComp's header_len is set to zero
because we opportunistically drop the IPComp header when the total
compressed length exceeds the original packet length.  That is,
we only ever do IPComp when it does not cause the packet to expand.

Yes, indeed. And it makes sense of course.

So it seems that we need another way of indicating the presence of
XFRM.

Then we probably have to use the !rt->dst.xfrm check. But then
we should use dst_xfrm() accessor, this avoids the ifdef here.

From: Alexey Kodanev <redacted>
Date: Fri,  3 Mar 2017 15:37:32 +0300

As Steffen has suggested, please use dst_xfrm().