[PATCH net] udp: consistently apply ufo or fragmentation

[PATCH net] udp: consistently apply ufo or fragmentation

[Vasily Averin]

Dear Willem,

ip*_ufo_append_data() is called now for each skb_is_gso() packet.
Before the patch it was called for UDP packets only.

Could you please clarify, is it correct?

Thank you,
	Vasily Averin

Re: [PATCH net] udp: consistently apply ufo or fragmentation

[Willem de Bruijn]

> ip*_ufo_append_data() is called now for each skb_is_gso() packet.
> Before the patch it was called for UDP packets only.
>
> Could you please clarify, is it correct?

__ip_append_data is called for udp, raw, icmp and ping sockets.
Of these, only udp packets have a path to get skb_is_gso() set:
when __ip_append_data has been called before on this packet
(using MSG_MORE) and the path entered ip_ufo_append_data
in that call through the other conditions in the branch, which
include sk->sk_type == SOCK_DGRAM.

This function is not reached from other segmentable protocols
like tcp, which call ip_queue_xmit in the ip layer. The exceptions
are sending in ip_send_unicast_reply, but those will not have gso.

[PATCH net] udp: consistently apply ufo or fragmentation

[Willem de Bruijn]

From: Willem de Bruijn <willemb@google.com>

When iteratively building a UDP datagram with MSG_MORE and that
datagram exceeds MTU, consistently choose UFO or fragmentation.

Once skb_is_gso, always apply ufo. Conversely, once a datagram is
split across multiple skbs, do not consider ufo.

Sendpage already maintains the first invariant, only add the second.
IPv6 does not have a sendpage implementation to modify.

A gso skb must have a partial checksum, do not follow sk_no_check_tx
in udp_send_skb.

Found by syzkaller.

Fixes: e89e9cf539a2 ("[IPv4/IPv6]: UFO Scatter-gather approach")
Reported-by: Andrey Konovalov <andreyknvl@google.com>
Signed-off-by: Willem de Bruijn <willemb@google.com>
---
 net/ipv4/ip_output.c  | 8 +++++---
 net/ipv4/udp.c        | 2 +-
 net/ipv6/ip6_output.c | 7 ++++---
 3 files changed, 10 insertions(+), 7 deletions(-)

diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
index 50c74cd890bc..e153c40c2436 100644
--- a/net/ipv4/ip_output.c
+++ b/net/ipv4/ip_output.c
@@ -965,11 +965,12 @@ static int __ip_append_data(struct sock *sk,
 		csummode = CHECKSUM_PARTIAL;
 
 	cork->length += length;
-	if ((((length + (skb ? skb->len : fragheaderlen)) > mtu) ||
-	     (skb && skb_is_gso(skb))) &&
+	if ((skb && skb_is_gso(skb)) ||
+	    (((length + (skb ? skb->len : fragheaderlen)) > mtu) &&
+	    (skb_queue_len(queue) <= 1) &&
 	    (sk->sk_protocol == IPPROTO_UDP) &&
 	    (rt->dst.dev->features & NETIF_F_UFO) && !dst_xfrm(&rt->dst) &&
-	    (sk->sk_type == SOCK_DGRAM) && !sk->sk_no_check_tx) {
+	    (sk->sk_type == SOCK_DGRAM) && !sk->sk_no_check_tx)) {
 		err = ip_ufo_append_data(sk, queue, getfrag, from, length,
 					 hh_len, fragheaderlen, transhdrlen,
 					 maxfraglen, flags);
@@ -1288,6 +1289,7 @@ ssize_t	ip_append_page(struct sock *sk, struct flowi4 *fl4, struct page *page,
 		return -EINVAL;
 
 	if ((size + skb->len > mtu) &&
+	    (skb_queue_len(&sk->sk_write_queue) == 1) &&
 	    (sk->sk_protocol == IPPROTO_UDP) &&
 	    (rt->dst.dev->features & NETIF_F_UFO)) {
 		if (skb->ip_summed != CHECKSUM_PARTIAL)
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index e6276fa3750b..a7c804f73990 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -802,7 +802,7 @@ static int udp_send_skb(struct sk_buff *skb, struct flowi4 *fl4)
 	if (is_udplite)  				 /*     UDP-Lite      */
 		csum = udplite_csum(skb);
 
-	else if (sk->sk_no_check_tx) {   /* UDP csum disabled */
+	else if (sk->sk_no_check_tx && !skb_is_gso(skb)) {   /* UDP csum off */
 
 		skb->ip_summed = CHECKSUM_NONE;
 		goto send;
diff --git a/net/ipv6/ip6_output.c b/net/ipv6/ip6_output.c
index 162efba0d0cd..2dfe50d8d609 100644
--- a/net/ipv6/ip6_output.c
+++ b/net/ipv6/ip6_output.c
@@ -1381,11 +1381,12 @@ static int __ip6_append_data(struct sock *sk,
 	 */
 
 	cork->length += length;
-	if ((((length + (skb ? skb->len : headersize)) > mtu) ||
-	     (skb && skb_is_gso(skb))) &&
+	if ((skb && skb_is_gso(skb)) ||
+	    (((length + (skb ? skb->len : headersize)) > mtu) &&
+	    (skb_queue_len(queue) <= 1) &&
 	    (sk->sk_protocol == IPPROTO_UDP) &&
 	    (rt->dst.dev->features & NETIF_F_UFO) && !dst_xfrm(&rt->dst) &&
-	    (sk->sk_type == SOCK_DGRAM) && !udp_get_no_check6_tx(sk)) {
+	    (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,
 					  transhdrlen, mtu, flags, fl6);
-- 
2.14.0.434.g98096fd7a8-goog

Re: [PATCH net] udp: consistently apply ufo or fragmentation

[David Miller]

From: Willem de Bruijn <willemdebruijn.kernel@gmail.com>
Date: Thu, 10 Aug 2017 12:29:19 -0400

> From: Willem de Bruijn <willemb@google.com>
> 
> When iteratively building a UDP datagram with MSG_MORE and that
> datagram exceeds MTU, consistently choose UFO or fragmentation.
> 
> Once skb_is_gso, always apply ufo. Conversely, once a datagram is
> split across multiple skbs, do not consider ufo.
> 
> Sendpage already maintains the first invariant, only add the second.
> IPv6 does not have a sendpage implementation to modify.
> 
> A gso skb must have a partial checksum, do not follow sk_no_check_tx
> in udp_send_skb.
> 
> Found by syzkaller.
> 
> Fixes: e89e9cf539a2 ("[IPv4/IPv6]: UFO Scatter-gather approach")
> Reported-by: Andrey Konovalov <andreyknvl@google.com>
> Signed-off-by: Willem de Bruijn <willemb@google.com>

Applied and queued up for -stable.