[net] net: ip_tunnel: fix mtu calculation

[net] net: ip_tunnel: fix mtu calculation

[Vadim Fedorenko]

dev->hard_header_len for tunnel interface is set only when header_ops
are set too and already contains full overhead of any tunnel encapsulation.
That's why there is not need to use this overhead twice in mtu calc.

Fixes: fdafed459998 ("ip_gre: set dev->hard_header_len and dev->needed_headroom properly")
Reported-by: Slava Bacherikov <mail@slava.cc>
Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru>
---
 net/ipv4/ip_tunnel.c | 18 +++++++++---------
 1 file changed, 9 insertions(+), 9 deletions(-)

diff --git a/net/ipv4/ip_tunnel.c b/net/ipv4/ip_tunnel.c
index 64594aa..ad78825 100644
--- a/net/ipv4/ip_tunnel.c
+++ b/net/ipv4/ip_tunnel.c
@@ -317,7 +317,7 @@ static int ip_tunnel_bind_dev(struct net_device *dev)
 	}
 
 	dev->needed_headroom = t_hlen + hlen;
-	mtu -= (dev->hard_header_len + t_hlen);
+	mtu -= dev->hard_header_len ? : t_hlen;
 
 	if (mtu < IPV4_MIN_MTU)
 		mtu = IPV4_MIN_MTU;
@@ -347,7 +347,7 @@ static struct ip_tunnel *ip_tunnel_create(struct net *net,
 	nt = netdev_priv(dev);
 	t_hlen = nt->hlen + sizeof(struct iphdr);
 	dev->min_mtu = ETH_MIN_MTU;
-	dev->max_mtu = IP_MAX_MTU - dev->hard_header_len - t_hlen;
+	dev->max_mtu = IP_MAX_MTU - dev->hard_header_len ? : t_hlen;
 	ip_tunnel_add(itn, nt);
 	return nt;
 
@@ -488,11 +488,11 @@ static int tnl_update_pmtu(struct net_device *dev, struct sk_buff *skb,
 	int mtu;
 
 	tunnel_hlen = md ? tunnel_hlen : tunnel->hlen;
-	pkt_size = skb->len - tunnel_hlen - dev->hard_header_len;
+	pkt_size = skb->len - dev->hard_header_len ? : tunnel_hlen;
 
 	if (df)
-		mtu = dst_mtu(&rt->dst) - dev->hard_header_len
-					- sizeof(struct iphdr) - tunnel_hlen;
+		mtu = dst_mtu(&rt->dst) - dev->hard_header_len ? :
+					 (sizeof(struct iphdr) + tunnel_hlen);
 	else
 		mtu = skb_valid_dst(skb) ? dst_mtu(skb_dst(skb)) : dev->mtu;
 
@@ -972,7 +972,7 @@ int __ip_tunnel_change_mtu(struct net_device *dev, int new_mtu, bool strict)
 {
 	struct ip_tunnel *tunnel = netdev_priv(dev);
 	int t_hlen = tunnel->hlen + sizeof(struct iphdr);
-	int max_mtu = IP_MAX_MTU - dev->hard_header_len - t_hlen;
+	int max_mtu = IP_MAX_MTU - dev->hard_header_len ? : t_hlen;
 
 	if (new_mtu < ETH_MIN_MTU)
 		return -EINVAL;
@@ -1149,10 +1149,10 @@ int ip_tunnel_newlink(struct net_device *dev, struct nlattr *tb[],
 
 	mtu = ip_tunnel_bind_dev(dev);
 	if (tb[IFLA_MTU]) {
-		unsigned int max = IP_MAX_MTU - dev->hard_header_len - nt->hlen;
+		unsigned int max = IP_MAX_MTU - dev->hard_header_len ? :
+						 (nt->hlen + sizeof(struct iphdr));
 
-		mtu = clamp(dev->mtu, (unsigned int)ETH_MIN_MTU,
-			    (unsigned int)(max - sizeof(struct iphdr)));
+		mtu = clamp(dev->mtu, (unsigned int)ETH_MIN_MTU, max);
 	}
 
 	err = dev_set_mtu(dev, mtu);
-- 
1.8.3.1

Re: [net] net: ip_tunnel: fix mtu calculation

[Willem de Bruijn]

On Wed, Jan 27, 2021 at 11:14 PM Vadim Fedorenko <vfedorenko@novek.ru> wrote:
>
> dev->hard_header_len for tunnel interface is set only when header_ops
> are set too and already contains full overhead of any tunnel encapsulation.
> That's why there is not need to use this overhead twice in mtu calc.
>
> Fixes: fdafed459998 ("ip_gre: set dev->hard_header_len and dev->needed_headroom properly")
> Reported-by: Slava Bacherikov <mail@slava.cc>
> Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru>
> ---
>  net/ipv4/ip_tunnel.c | 18 +++++++++---------
>  1 file changed, 9 insertions(+), 9 deletions(-)
>
> diff --git a/net/ipv4/ip_tunnel.c b/net/ipv4/ip_tunnel.c
> index 64594aa..ad78825 100644
> --- a/net/ipv4/ip_tunnel.c
> +++ b/net/ipv4/ip_tunnel.c
> @@ -317,7 +317,7 @@ static int ip_tunnel_bind_dev(struct net_device *dev)
>         }
>
>         dev->needed_headroom = t_hlen + hlen;
> -       mtu -= (dev->hard_header_len + t_hlen);
> +       mtu -= dev->hard_header_len ? : t_hlen;

Safety of this change also depends on whether any other ip tunnels
might have non-zero hard_header_len.

I haven't fully checked yet, but at first scan I only see one other
instance of header_ops, and that ip_tunnel_header_ops does not have a
create implementation.

>
>         if (mtu < IPV4_MIN_MTU)
>                 mtu = IPV4_MIN_MTU;
> @@ -347,7 +347,7 @@ static struct ip_tunnel *ip_tunnel_create(struct net *net,
>         nt = netdev_priv(dev);
>         t_hlen = nt->hlen + sizeof(struct iphdr);
>         dev->min_mtu = ETH_MIN_MTU;
> -       dev->max_mtu = IP_MAX_MTU - dev->hard_header_len - t_hlen;
> +       dev->max_mtu = IP_MAX_MTU - dev->hard_header_len ? : t_hlen;

here and elsewhere: subtraction takes precedence over ternary
conditional, so (IP_MAX_MTU - ..) always true.

Re: [net] net: ip_tunnel: fix mtu calculation

[Vadim Fedorenko]

On 28.01.2021 21:48, Willem de Bruijn wrote:
> On Wed, Jan 27, 2021 at 11:14 PM Vadim Fedorenko <vfedorenko@novek.ru> wrote:
>>
>> dev->hard_header_len for tunnel interface is set only when header_ops
>> are set too and already contains full overhead of any tunnel encapsulation.
>> That's why there is not need to use this overhead twice in mtu calc.
>>
>> Fixes: fdafed459998 ("ip_gre: set dev->hard_header_len and dev->needed_headroom properly")
>> Reported-by: Slava Bacherikov <mail@slava.cc>
>> Signed-off-by: Vadim Fedorenko <vfedorenko@novek.ru>
>> ---
>>   net/ipv4/ip_tunnel.c | 18 +++++++++---------
>>   1 file changed, 9 insertions(+), 9 deletions(-)
>>
>> diff --git a/net/ipv4/ip_tunnel.c b/net/ipv4/ip_tunnel.c
>> index 64594aa..ad78825 100644
>> --- a/net/ipv4/ip_tunnel.c
>> +++ b/net/ipv4/ip_tunnel.c
>> @@ -317,7 +317,7 @@ static int ip_tunnel_bind_dev(struct net_device *dev)
>>          }
>>
>>          dev->needed_headroom = t_hlen + hlen;
>> -       mtu -= (dev->hard_header_len + t_hlen);
>> +       mtu -= dev->hard_header_len ? : t_hlen;
> 
> Safety of this change also depends on whether any other ip tunnels
> might have non-zero hard_header_len.

Yes, sure.

> I haven't fully checked yet, but at first scan I only see one other
> instance of header_ops, and that ip_tunnel_header_ops does not have a
> create implementation.

Yes. The calls to ip_tunnel_setup are in ip_gre.c, ip_tunnel.c, ip_vti.c
and ipip.c. All of them except of ip_gre use ip_tunnel_header_ops which
doesn't have create implementation and have hard_header_len set to 0.

> 
>>
>>          if (mtu < IPV4_MIN_MTU)
>>                  mtu = IPV4_MIN_MTU;
>> @@ -347,7 +347,7 @@ static struct ip_tunnel *ip_tunnel_create(struct net *net,
>>          nt = netdev_priv(dev);
>>          t_hlen = nt->hlen + sizeof(struct iphdr);
>>          dev->min_mtu = ETH_MIN_MTU;
>> -       dev->max_mtu = IP_MAX_MTU - dev->hard_header_len - t_hlen;
>> +       dev->max_mtu = IP_MAX_MTU - dev->hard_header_len ? : t_hlen;
> 
> here and elsewhere: subtraction takes precedence over ternary
> conditional, so (IP_MAX_MTU - ..) always true.

Oh, sure, my bad. Will fix it in v2, thanks!