netdev - Re: [RFC] ipv6: dst_allfrag() not taken into account by TCP

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <1326824111.2259.54.camel@edumazet-HP-Compaq-6005-Pro-SFF-PC>
Date:	Tue, 17 Jan 2012 19:15:11 +0100
From:	Eric Dumazet <eric.dumazet@...il.com>
To:	David Miller <davem@...emloft.net>
Cc:	netdev@...r.kernel.org, tore@....no
Subject: Re: [RFC] ipv6: dst_allfrag() not taken into account by TCP

Le mardi 17 janvier 2012 à 12:34 -0500, David Miller a écrit :
> From: Eric Dumazet <eric.dumazet@...il.com>
> Date: Tue, 17 Jan 2012 17:28:19 +0100
> 
> > It seems that dst_allfrag() will force us to use ip6_fragment() and
> > reduce effective MSS to :
> 
> Don't use TCP with forced fragmentation.
> 
> Use path MTU discovery or an artificially reduced route MTU on paths
> where ICMP blocking is causing problems.

We receive a correct ICMP message and we process it correctly.

It should work without fragments, if we hadnt a bug, when we hit the
minimum IPv6 MTU (1280) ?

Problem is we build too large TCP frames (by 8 bytes) and they must be
fragmented in ip6_finish_output(), wasting cpu and network bandwidth.

We do the following in tcp_mtu_to_mss()

mss_now = pmtu - icsk->icsk_af_ops->net_header_len - sizeof(struct tcphdr);

(net_header_len being a constant value : sizeof(struct ipv6hdr) for IPv6)

For IPv6, effective network header len would be either :

1)	sizeof(struct ipv6hdr) 
2)	sizeof(struct ipv6hdr) + sizeof(frag_hdr)

We could change net_header_len to be a method, and for ipv6 do :

static unsigned int ipv6_net_header_len(struct sock *sk)
{
	const struct dst_entry *dst = __sk_dst_get(sk);

	return sizeof(struct ipv6hdr) +
		(dst && dst_allfrag(dst)) ? sizeof(struct frag_hdr) : 0; 
}

/* for ipv4 : */
static unsigned int ipv4_net_header_len(struct sock *sk)
{
	return sizeof(struct iphdr);
}



--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html