[PATCH net v3 1/5] ip_tunnel: do not clear the active encap before validating the new one

[Eric Dumazet <edumazet@google.com>]

ip_tunnel_encap_setup() and ip6_tnl_encap_setup() zero out t->encap with
memset() before calling ip_encap_hlen() / ip6_encap_hlen() to validate
the requested encapsulation type and module availability.

When a changelink request supplies an invalid encapsulation type or an
encapsulation whose module is not loaded, ip[6]_encap_hlen() returns
-EINVAL after t->encap has already been cleared to TUNNEL_ENCAP_NONE,
while t->encap_hlen and t->hlen remain at their previous values. A
rejected netlink request thus permanently disables FOU/GUE on a working
tunnel while keeping its reduced MTU and extra headroom.

The memset() is redundant: struct ip_tunnel_encap has exactly four
fields, and all of them are assigned unconditionally once the length
check has passed. A tunnel being created starts from the zeroed private
area of alloc_netdev(). Simply remove it, so that a failed
ip_tunnel_encap_setup() or ip6_tnl_encap_setup() leaves the active
encapsulation untouched. (Callers still invoke ip[6]_tunnel_encap_setup()
before validating the rest of a changelink request.)

net-next already does this for the IPv4 side in commit 88b84cae6b94
("ip_tunnel: use WRITE_ONCE in ip_tunnel_encap_setup"), which is not in
net. Removing the same lines here keeps the merge trivial, and extends
the fix to IPv6, which that commit did not touch.

Fixes: 56328486539d ("net: Changes to ip_tunnel to support foo-over-udp encapsulation")
Fixes: 058214a4d1df ("ip6_tun: Add infrastructure for doing encapsulation")
Cc: stable@vger.kernel.org
Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 net/ipv4/ip_tunnel.c  | 2 --
 net/ipv6/ip6_tunnel.c | 2 --
 2 files changed, 4 deletions(-)

diff --git a/net/ipv4/ip_tunnel.c b/net/ipv4/ip_tunnel.c
index e6bcf01411d0bcd12cc9a88e449d9283c4a83c64..2a313b18134e2f0bafd52d59d8e173fa1a084f03 100644
--- a/net/ipv4/ip_tunnel.c
+++ b/net/ipv4/ip_tunnel.c
@@ -491,8 +491,6 @@ int ip_tunnel_encap_setup(struct ip_tunnel *t,
 {
 	int hlen;
 
-	memset(&t->encap, 0, sizeof(t->encap));
-
 	hlen = ip_encap_hlen(ipencap);
 	if (hlen < 0)
 		return hlen;
diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c
index d5ff50a2ac01760acd1cb176c952f7113f733288..c918c2b0ad81b0161a2a98e4bd861436497c551c 100644
--- a/net/ipv6/ip6_tunnel.c
+++ b/net/ipv6/ip6_tunnel.c
@@ -1818,8 +1818,6 @@ int ip6_tnl_encap_setup(struct ip6_tnl *t,
 {
 	int hlen;
 
-	memset(&t->encap, 0, sizeof(t->encap));
-
 	hlen = ip6_encap_hlen(ipencap);
 	if (hlen < 0)
 		return hlen;
-- 
2.55.0.1082.g2b9226bbc0-goog