From: Sabrina Dubroca <sd@queasysnail.net>
Date: Thu, 13 Aug 2020 16:24:04 +0200
Subject: espintcp: restore IP CB before handing the packet to xfrm
Patch-mainline: v5.9
Git-commit: 4eb2e13415757a2bce5bb0d580d22bbeef1f5346
References: bsc#1176447

Xiumei reported a bug with espintcp over IPv6 in transport mode,
because xfrm6_transport_finish expects to find IP6CB data (struct
inet6_skb_cb). Currently, espintcp zeroes the CB, but the relevant
part is actually preserved by previous layers (first set up by tcp,
then strparser only zeroes a small part of tcp_skb_tb), so we can just
relocate it to the start of skb->cb.

Fixes: e27cca96cd68 ("xfrm: add espintcp (RFC 8229)")
Reported-by: Xiumei Mu <xmu@redhat.com>
Signed-off-by: Sabrina Dubroca <sd@queasysnail.net>
Signed-off-by: Steffen Klassert <steffen.klassert@secunet.com>
Acked-by: Thomas Bogendoerfer <tbogendoerfer@suse.de>
---
 net/xfrm/espintcp.c |    6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

--- a/net/xfrm/espintcp.c
+++ b/net/xfrm/espintcp.c
@@ -29,8 +29,12 @@ static void handle_nonesp(struct espintc
 
 static void handle_esp(struct sk_buff *skb, struct sock *sk)
 {
+	struct tcp_skb_cb *tcp_cb = (struct tcp_skb_cb *)skb->cb;
+
 	skb_reset_transport_header(skb);
-	memset(skb->cb, 0, sizeof(skb->cb));
+
+	/* restore IP CB, we need at least IP6CB->nhoff */
+	memmove(skb->cb, &tcp_cb->header, sizeof(tcp_cb->header));
 
 	rcu_read_lock();
 	skb->dev = dev_get_by_index_rcu(sock_net(sk), skb->skb_iif);