netdev - Re: [PATCH net v3] ipv4: allow local fragmentation in ip_finish_output

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] [day] [month] [year] [list]

Message-ID: <1461381070.97911800.1478267353507.JavaMail.zimbra@redhat.com>
Date:   Fri, 4 Nov 2016 09:49:13 -0400 (EDT)
From:   Lance Richardson <lrichard@...hat.com>
To:     Shmulik Ladkani <shmulik.ladkani@...il.com>
Cc:     Hannes Frederic Sowa <hannes@...essinduktion.org>, fw@...len.de,
        netdev@...r.kernel.org, jtluka@...hat.com
Subject: Re: [PATCH net v3] ipv4: allow local fragmentation in
 ip_finish_output_gso()

> From: "Shmulik Ladkani" <shmulik.ladkani@...il.com>
> To: "Hannes Frederic Sowa" <hannes@...essinduktion.org>
> Cc: "Lance Richardson" <lrichard@...hat.com>, fw@...len.de, netdev@...r.kernel.org, jtluka@...hat.com
> Sent: Friday, November 4, 2016 5:40:14 AM
> Subject: Re: [PATCH net v3] ipv4: allow local fragmentation in ip_finish_output_gso()
> 
> On Thu, 3 Nov 2016 22:34:34 +0100 Hannes Frederic Sowa
> <hannes@...essinduktion.org> wrote:
> > Correct, but we should maybe redefine the code a bit. From my
> > understanding we can now create an ICMP storm in case every fragment gets.
> 
> Yes, you are right.
> 
> Each segment gets into ip_fragment, and due to outer DF being set,
> ICMP_FRAG_NEEDED is sent per segment.
> 
> BTW, suppose GRO is off, and sender actually did send a burst of
> (non-gso) packets with outer DF set, and each was tunnel encapsulated,
> resulting in oversized frames.
> 
> Would'nt the stack just send the ICMP_FRAG_NEEDED per encapsulated
> frame?
> 
> If so, then the GRO behaviour is aligned, and there's nothing to fix.
> 

Agree.

> Best,
> Shmulik
>