netdev - Re: Massive hash collisions on FIB

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: <CANn89iL0Cy0sEiYZnFbHFAJpj1dUD-Z93wLyHJyr=f-xuLzZtQ@mail.gmail.com>
Date: Wed, 25 Sep 2024 21:25:50 +0200
From: Eric Dumazet <edumazet@...gle.com>
To: Alexandre Ferrieux <alexandre.ferrieux@...il.com>
Cc: Simon Horman <horms@...nel.org>, Przemek Kitszel <przemyslaw.kitszel@...el.com>, 
	nicolas.dichtel@...nd.com, netdev@...r.kernel.org
Subject: Re: Massive hash collisions on FIB

On Wed, Sep 25, 2024 at 9:06 PM Alexandre Ferrieux
<alexandre.ferrieux@...il.com> wrote:
>
> On 24/09/2024 19:18, Alexandre Ferrieux wrote:
> >
> > I see you did the work for the two other hashes (laddr and devindex).
> > I tried to inject the dispersion the same way as you did, just before the final
> > scrambling. Is this what you'd do ?
> >
> > diff --git a/net/ipv4/fib_semantics.c b/net/ipv4/fib_semantics.c
> Doing this, I discovered with surprise that the job was already done in IPv6, as
> inet6_addr_hash() uses net_hash_mix() too. Which leads to the question: why are
> the IPv4 and IPv6 FIB-exact-lookup implementations different/duplicated ?

inet_addr_hash() is also using net_hash_mix()

fib_info_laddrhash_bucket() is also using net_hash_mix()

You know we make these kinds of changes whenever they are needed for
our workload.

For example

d07418afea8f1d9896aaf9dc5ae47ac4f45b220c ipv4: avoid quadratic
behavior in netns dismantle

Just submit a patch, stop wondering why it was not already done.