netdev - Re: [PATCH net-next 1/2] tcp: conditionally call ip_icmp_error() from tcp_v4

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: <20240418110909.091b0550@kernel.org>
Date: Thu, 18 Apr 2024 11:09:09 -0700
From: Jakub Kicinski <kuba@...nel.org>
To: Eric Dumazet <edumazet@...gle.com>, David Ahern <dsahern@...nel.org>
Cc: Paolo Abeni <pabeni@...hat.com>, "David S . Miller"
 <davem@...emloft.net>, netdev@...r.kernel.org, Neal Cardwell
 <ncardwell@...gle.com>, Dragos Tatulea <dtatulea@...dia.com>,
 eric.dumazet@...il.com, Maciej Żenczykowski
 <maze@...gle.com>, Willem de Bruijn <willemb@...gle.com>, Shachar Kagan
 <skagan@...dia.com>
Subject: Re: [PATCH net-next 1/2] tcp: conditionally call ip_icmp_error()
 from tcp_v4_err()

On Thu, 18 Apr 2024 19:47:51 +0200 Eric Dumazet wrote:
> > You have a kernel patch that makes a test fail, and your solution is
> > changing userspace? The tests are examples of userspace applications and
> > how they can use APIs, so if the patch breaks a test it is by definition
> > breaking userspace which is not allowed.  

Tests are often overly sensitive to kernel behavior, while this is
obviously a red flag it's not an automatic nack. The more tests we
have the more often we'll catch tiny changes. A lot of tests started
flaking with 6.9 because of the optimizations in the timer subsystem.
You know where I'm going with this..

> I think the userspace program relied on a bug added in linux in 2020
> 
> Jakub, I will stop trying to push the patches, this is a lost battle.

If you have the patches ready - please post them.
I'm happy to take the blame if they actually regress something in 
the wild :(

We're pursuing this because real application suffer real problems
when routing changes cause transient ICMP errors. Users read the RFC
and then come shouting at us that Linux is buggy.