Message ID | 20191123145817.GA22321@fuckup |
---|---|
State | Changes Requested |
Delegated to: | David Miller |
Headers | show |
Series | net: ip/tnl: Set iph->id only when don't fragment is not set | expand |
On 11/23/19 6:58 AM, Oliver Herms wrote: > In IPv4 the identification field ensures that fragments of different datagrams > are not mixed by the receiver. Packets with Don't Fragment (DF) flag set are not > to be fragmented in transit and thus don't need an identification. Official sources for this assertion please, so that we can double check if you implemented the proper avoidance ? > Calculating the identification takes significant CPU time. > This patch will increase IP tunneling performance by ~10% unless DF is not set. > However, DF is set by default which is best practice. > > Signed-off-by: Oliver Herms <oliver.peter.herms@gmail.com> > --- > net/ipv4/ip_tunnel_core.c | 4 +++- > 1 file changed, 3 insertions(+), 1 deletion(-) > > diff --git a/net/ipv4/ip_tunnel_core.c b/net/ipv4/ip_tunnel_core.c > index 1452a97914a0..8636c1e0e7b7 100644 > --- a/net/ipv4/ip_tunnel_core.c > +++ b/net/ipv4/ip_tunnel_core.c > @@ -73,7 +73,9 @@ void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb, > iph->daddr = dst; > iph->saddr = src; > iph->ttl = ttl; > - __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1); > + > + if (unlikely((iph->frag_off & htons(IP_DF)) == false)) This unlikely() seems wrong to me. You do not know what are the odds of IP_DF being set or not. > + __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1); > > err = ip_local_out(net, sk, skb); > > So we are going to send 2 bytes with garbage if we do not call __ip_select_ident() This would cause various security threats, since the garbage might reveal a secret.
On Sat, 2019-11-23 at 09:53 -0800, Eric Dumazet wrote: > > On 11/23/19 6:58 AM, Oliver Herms wrote: > > In IPv4 the identification field ensures that fragments of different datagrams > > are not mixed by the receiver. Packets with Don't Fragment (DF) flag set are not > > to be fragmented in transit and thus don't need an identification. > > Official sources for this assertion please, so that we can double check if you > implemented the proper avoidance ? > > > Calculating the identification takes significant CPU time. > > This patch will increase IP tunneling performance by ~10% unless DF is not set. > > However, DF is set by default which is best practice. > > > > Signed-off-by: Oliver Herms <oliver.peter.herms@gmail.com> > > --- > > net/ipv4/ip_tunnel_core.c | 4 +++- > > 1 file changed, 3 insertions(+), 1 deletion(-) > > > > diff --git a/net/ipv4/ip_tunnel_core.c b/net/ipv4/ip_tunnel_core.c > > index 1452a97914a0..8636c1e0e7b7 100644 > > --- a/net/ipv4/ip_tunnel_core.c > > +++ b/net/ipv4/ip_tunnel_core.c > > @@ -73,7 +73,9 @@ void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb, > > iph->daddr = dst; > > iph->saddr = src; > > iph->ttl = ttl; > > - __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1); > > + > > + if (unlikely((iph->frag_off & htons(IP_DF)) == false)) > > This unlikely() seems wrong to me. as does the comparison to false.
On 23.11.19 18:53, Eric Dumazet wrote: > Official sources for this assertion please, so that we can double check if you > implemented the proper avoidance ? From RFC 6864 Section 4.1: "The IPv4 ID field MUST NOT be used for purposes other than fragmentation and reassembly." >> net/ipv4/ip_tunnel_core.c | 4 +++- >> 1 file changed, 3 insertions(+), 1 deletion(-) >> >> diff --git a/net/ipv4/ip_tunnel_core.c b/net/ipv4/ip_tunnel_core.c >> index 1452a97914a0..8636c1e0e7b7 100644 >> --- a/net/ipv4/ip_tunnel_core.c >> +++ b/net/ipv4/ip_tunnel_core.c >> @@ -73,7 +73,9 @@ void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb, >> iph->daddr = dst; >> iph->saddr = src; >> iph->ttl = ttl; >> - __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1); >> + >> + if (unlikely((iph->frag_off & htons(IP_DF)) == false)) > > This unlikely() seems wrong to me. > > You do not know what are the odds of IP_DF being set or not. Right. I'll send a corrected patch. >> + __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1); > So we are going to send 2 bytes with garbage if we do not call __ip_select_ident() > > This would cause various security threats, since the garbage might reveal a secret. So we should set it to zero then.
diff --git a/net/ipv4/ip_tunnel_core.c b/net/ipv4/ip_tunnel_core.c index 1452a97914a0..8636c1e0e7b7 100644 --- a/net/ipv4/ip_tunnel_core.c +++ b/net/ipv4/ip_tunnel_core.c @@ -73,7 +73,9 @@ void iptunnel_xmit(struct sock *sk, struct rtable *rt, struct sk_buff *skb, iph->daddr = dst; iph->saddr = src; iph->ttl = ttl; - __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1); + + if (unlikely((iph->frag_off & htons(IP_DF)) == false)) + __ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1); err = ip_local_out(net, sk, skb);
In IPv4 the identification field ensures that fragments of different datagrams are not mixed by the receiver. Packets with Don't Fragment (DF) flag set are not to be fragmented in transit and thus don't need an identification. Calculating the identification takes significant CPU time. This patch will increase IP tunneling performance by ~10% unless DF is not set. However, DF is set by default which is best practice. Signed-off-by: Oliver Herms <oliver.peter.herms@gmail.com> --- net/ipv4/ip_tunnel_core.c | 4 +++- 1 file changed, 3 insertions(+), 1 deletion(-)