netdev - Re: Change in IPv4 TCP reset behavior after kernel upgrade

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: <20131205132733.GA14095@order.stressinduktion.org>
Date:	Thu, 5 Dec 2013 14:27:33 +0100
From:	Hannes Frederic Sowa <hannes@...essinduktion.org>
To:	Eric Dumazet <eric.dumazet@...il.com>
Cc:	Fiedler Roman <Roman.Fiedler@....ac.at>,
	Neal Cardwell <ncardwell@...gle.com>,
	"davem@...emloft.net" <davem@...emloft.net>,
	"netdev@...r.kernel.org" <netdev@...r.kernel.org>
Subject: Re: Change in IPv4 TCP reset behavior after kernel upgrade

On Wed, Dec 04, 2013 at 03:37:01PM -0800, Eric Dumazet wrote:
> On Wed, 2013-12-04 at 23:02 +0100, Hannes Frederic Sowa wrote:
> 
> > Oh yes, actually looking at the code we definitely run into __ip_select_ident
> > and generate an inetpeer and use the derived ipid from that.
> > 
> > Roman, I would be very interested why you only experienced that with
> > 3.10.21, could you share the specific vserver patch?
> 
> BTW I think we might have an opportunity to respect no_pmtu_disc sysctl
> as in :
> 
> (ie have DF bit set or unset for RST packets, and ACK packets sent on behalf
> of TIMEWAIT sockets)
> 
> Still we use a non 0 id for these packets.
> 
> diff --git a/net/ipv4/ip_output.c b/net/ipv4/ip_output.c
> index 9124027..a1e05a9 100644
> --- a/net/ipv4/ip_output.c
> +++ b/net/ipv4/ip_output.c
> @@ -1473,7 +1473,6 @@ static DEFINE_PER_CPU(struct inet_sock, unicast_sock) = {
>                 .sk_allocation  = GFP_ATOMIC,
>                 .sk_flags       = (1UL << SOCK_USE_WRITE_QUEUE),
>         },
> -       .pmtudisc       = IP_PMTUDISC_WANT,
>         .uc_ttl         = -1,
>  };
>  
> @@ -1519,6 +1518,7 @@ void ip_send_unicast_reply(struct net *net, struct sk_buff *skb, __be32 daddr,
>         inet = &get_cpu_var(unicast_sock);
>  
>         inet->tos = arg->tos;
> +       inet->pmtudisc = ipv4_config.no_pmtu_disc ? IP_PMTUDISC_DONT : IP_PMTUDISC_WANT;
>         sk = &inet->sk;
>         sk->sk_priority = skb->priority;
>         sk->sk_protocol = ip_hdr(skb)->protocol;
> 

Ack, the patch looks good. I guess we can do the same for icmp sockets.

Because we select per-destination ipids for those packets I don't know if it
is actually worth to try to change this behaviour again. It could reduce the
creation of new inetpeers a bit but we also generate one for each outgoing
icmp error, so maybe it is not worth the effort.

Greetings,

  Hannes

--
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