Follow @Openwall on Twitter for new release announcements and other news
[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <1e18c29fe7f9605b399c542ddd9c2de3142c1dd2.camel@kjb.biglobe.ne.jp>
Date: Fri, 13 Nov 2026 04:20:07 +0900
From: Nakashima Seiichi <s-nakashima@....biglobe.ne.jp>
To: Marian Marinov <hackman@...u.biz>, kernel-hardening@...ts.openwall.com, 
	linux-security-module@...ts.linux.dev
Cc: keescook@...omium.org, paul@...l-moore.com, mic@...ikod.net
Subject: Re: [RFC][IDEA] External Monitoring of Kernel Network Stack via BPF
 kprobe (KAF v2/v3 withdrawn)

Dear Marian

Thank you for your comment.

I apologize for taking up your valuable time. I will end it here.

I had hoped that if we could stop the threat somewhere, 
various burdens would be reduced, 
but I have realized that this is difficult to achieve.

I learned a lot, thank you.

Nakashima Seiichi

> Do you realize how slow this would be in a hot path in the kernel?
> 
> Please think this through. A network path on a 10G interface might
> process millions of packets per second.
> Implementing what you want will kill the performance. Performance
> that is tightly optimized around CPU cache lines.
> What you are proposing will introduce slowness and change of cache
> coherency that will have a heavy hit on performance. 
> 
> I think that we should allow the sysadmins to decide if they want to
> reduce their system performance for this kind of security measure. We
> already suffer some performance hit because of CPU mitigations and
> there is nothing the kernel can do about it. 
> 
> Additionally, most of the serious attack vectors would require
> multiple things in the kernel to be monitored. 
> We also know that slowing down some parts of the kernel might open
> other parts of it to race condition attacks. 
> 
> Maybe we should focus more on educating people on how to handle these
> types of attacks with the tools we have at our disposal at the
> moment.
> 
> I really hope that this path you took, thought you more about how the
> kernel and its security works, so you can enhance your idea(s) in the
> future.
> 
> I would also suggest that you look at building this out-of-tree and
> getting some of your friends on board. Obviously you are doing a lot
> of work with an LLM. Force it to build the PoC for you. Show this as
> a viable option. 
> 
> At the moment, personally, I think this proposal is just loosing the
> valuable time of the people in these mailing lists. 
> 
> Marian
> 

Powered by blists - more mailing lists

Confused about mailing lists and their use? Read about mailing lists on Wikipedia and check out these guidelines on proper formatting of your messages.