[PATCH net-next v3] ip6_vti: adjust vti mtu according to mtu of output device

[PATCH net-next v3] ip6_vti: adjust vti mtu according to mtu of output device

[Alexey Kodanev]

LTP/udp6_ipsec_vti tests fail when sending large UDP datagrams that
require fragmentation and the underlying device has MTU <= 1500. This
happens because ip6_vti sets mtu to ETH_DATA_LEN and not updating it
depending on a destination address or link parameter.

Further attempts to send UDP packets may succeed because pmtu gets
updated on ICMPV6_PKT_TOOBIG in vti6_err().

Here is the example when the output device MTU is set to 9000:

  # ip a sh ltp_ns_veth2
      ltp_ns_veth2@if7: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 9000 ...
        inet 10.0.0.2/24 scope global ltp_ns_veth2
        inet6 fd00::2/64 scope global

  # ip li add vti6 type vti6 local fd00::2 remote fd00::1
  # ip li show vti6
      vti6@NONE: <POINTOPOINT,NOARP> mtu 1500 ...
        link/tunnel6 fd00::2 peer fd00::1

After the patch:
  # ip li add vti6 type vti6 local fd00::2 remote fd00::1
  # ip li show vti6
      vti6@NONE: <POINTOPOINT,NOARP> mtu 8832 ...
        link/tunnel6 fd00::2 peer fd00::1

Regarding ip_vti, it already tunes MTU with ip_tunnel_bind_dev().

Reported-by: Petr Vorel <pvorel@suse.cz>
Signed-off-by: Alexey Kodanev <alexey.kodanev@oracle.com>
Acked-by: Shannon Nelson <shannon.nelson@oracle.com>
---
v3: * fix style issue with curly braces around single-statement if block

v2: * cleanup commit message issues (thanks to Shannon)

    * handle the case when we don't have route but have device parameter

    * cast new MTU to int and then check the maximum (tdev->mtu can be
      less than dev->hard_header_len)

 net/ipv6/ip6_vti.c |   21 +++++++++++++++++++++
 1 files changed, 21 insertions(+), 0 deletions(-)

diff --git a/net/ipv6/ip6_vti.c b/net/ipv6/ip6_vti.c
index dbb74f3..5404443 100644
--- a/net/ipv6/ip6_vti.c
+++ b/net/ipv6/ip6_vti.c
@@ -626,6 +626,7 @@ static void vti6_link_config(struct ip6_tnl *t)
 {
 	struct net_device *dev = t->dev;
 	struct __ip6_tnl_parm *p = &t->parms;
+	struct net_device *tdev = NULL;
 
 	memcpy(dev->dev_addr, &p->laddr, sizeof(struct in6_addr));
 	memcpy(dev->broadcast, &p->raddr, sizeof(struct in6_addr));
@@ -638,6 +639,26 @@ static void vti6_link_config(struct ip6_tnl *t)
 		dev->flags |= IFF_POINTOPOINT;
 	else
 		dev->flags &= ~IFF_POINTOPOINT;
+
+	if (p->flags & IP6_TNL_F_CAP_XMIT) {
+		int strict = (ipv6_addr_type(&p->raddr) &
+			      (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
+
+		struct rt6_info *rt = rt6_lookup(t->net,
+						 &p->raddr, &p->laddr,
+						 p->link, strict);
+
+		if (rt)
+			tdev = rt->dst.dev;
+		ip6_rt_put(rt);
+	}
+
+	if (!tdev && p->link)
+		tdev = __dev_get_by_index(t->net, p->link);
+
+	if (tdev)
+		dev->mtu = max_t(int, tdev->mtu - dev->hard_header_len,
+				 IPV6_MIN_MTU);
 }
 
 /**
-- 
1.7.1

Re: [PATCH net-next v3] ip6_vti: adjust vti mtu according to mtu of output device

[David Miller]

From: Alexey Kodanev <alexey.kodanev@oracle.com>
Date: Thu, 14 Dec 2017 18:37:50 +0300

Two minor pieces of feedback:

> LTP/udp6_ipsec_vti tests fail when sending large UDP datagrams that
> require fragmentation and the underlying device has MTU <= 1500. This
> happens because ip6_vti sets mtu to ETH_DATA_LEN and not updating it
> depending on a destination address or link parameter.
> 
> Further attempts to send UDP packets may succeed because pmtu gets
> updated on ICMPV6_PKT_TOOBIG in vti6_err().
> 
> Here is the example when the output device MTU is set to 9000:

You are fixing a problem that occurs when the underlying device has
an MTU smaller than 1500, yet you show an example involving an MTU
of 9000.

Care to adjust that inconsistency or explain why it's legit here?

> +	if (p->flags & IP6_TNL_F_CAP_XMIT) {
> +		int strict = (ipv6_addr_type(&p->raddr) &
> +			      (IPV6_ADDR_MULTICAST | IPV6_ADDR_LINKLOCAL));
> +
> +		struct rt6_info *rt = rt6_lookup(t->net,
> +						 &p->raddr, &p->laddr,
> +						 p->link, strict);

Please do not place empty lines between local variable declarations.

Thank you.