netdev - Re: [PATCH nf-next RFC 1/3] xt_statistic: taking GRO/GSO into account for nth-match

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: <aTbeKpBL0qVtjyY-@strlen.de>
Date: Mon, 8 Dec 2025 15:18:25 +0100
From: Florian Westphal <fw@...len.de>
To: Nick Wood <nwood@...udflare.com>
Cc: Jesper Dangaard Brouer <hawk@...nel.org>,
	netfilter-devel@...r.kernel.org,
	Pablo Neira Ayuso <pablo@...filter.org>, netdev@...r.kernel.org,
	phil@....cc, Eric Dumazet <eric.dumazet@...il.com>,
	"David S. Miller" <davem@...emloft.net>,
	Jakub Kicinski <kuba@...nel.org>, Paolo Abeni <pabeni@...hat.com>,
	kernel-team@...udflare.com, mfleming@...udflare.com,
	matt@...dmodwrite.com, aforster@...udflare.com
Subject: Re: [PATCH nf-next RFC 1/3] xt_statistic: taking GRO/GSO into
 account for nth-match

Nick Wood <nwood@...udflare.com> wrote:
> > It was easier to change the nth code (and easier for me to reason about)
> > than dealing with converting the the formula to use an integer
> > approximation (given we don't have floating point calc in kernel).
> 
> with s = integer sample rate (i.e s=100 if we're sampling 1/100)
> and n = nr_packets:
> 
> sample_threshold = [ 2**32 //s //s ] * [n*(s - (n-1))] ;
> 
> if get_random_u32 < sample_threshold {
>     sample_single_subpacket
> }*

Thanks for clarifying.  But we can't do that within the limitations
imposed by the netfilter framework.

We can only flag the entire skb/aggregate as either matching or not
matching.  We can't signal a 'match this subpacket'.

While we could split the aggregate within the conditional, we can't
(re)inject splitted packets back into the processing pipeline.

If there is a way, I can't think of anything.  At least not without
major rework of the entire netfilter stack :-|