Re: [PATCH 00/13 net-next] Allow compiling an IPv6-only kernel network stack

[Fernando Fernandez Mancera <fmancera@suse.de>]

On 9/28/26 7:08 PM, Willem de Bruijn wrote:
> Fernando Fernandez Mancera wrote:
>> On 9/15/26 2:24 AM, Jakub Kicinski wrote:
>>> On Thu, 10 Sep 2026 16:48:25 +0200 Fernando Fernandez Mancera wrote:
>>>> The primary goal of this patch series is to enable the compilation of an
>>>> IPv6-only kernel by decoupling the core networking infrastructure from the IPv4
>>>> protocol.
>>>
>>> Could you useful to know for example how many of the recent CVEs were
>>> IPv4 specific, if you have enough token budget to get such data?
>>>
>>
>> Hi Jakub I will include the numbers in the v2 cover letter but in essence:
>>
>> 0.3% (32) of the released CVEs since 2025. FWIW, this only includes
>> strictly IPv4-only code. The percentage is also stable over time if we
>> extend the search up to 2020, 0.3% (57) of the released CVEs.
>>
>>> High level, I don't have must to say beyond what you already know -
>>> the code movement may be too painful for backports. There are too
>>> many ifdefs (as Stan pointed out).
>>>
>>
>> Yes, the code movement is the biggest pain. Regarding the ifdefs, they
>> are being reduced greatly in the v2 as I have decided to rely on DCE
>> (Dead Code Elimination) to avoid using them. The code looks much better
>> this way.
> 
> On code movement: are the large code block movements into separate ipv4
> files needed, or could those just as well stay in place, behind an if
> enabled check?

It depends. Most of it could be guarded with ifdefs but that would IMHO 
make the code ugly and quite confusing. That code is 100% IPv4-only so 
it should be moved out..

In addition, this happened in the past too, e.g net/ipv4/tcp_ipv4.c was 
splitted from net/ipv4/tcp.c even before this series.

Of course, I trust maintainers' taste here so if you all insist that it 
is much better to put ifdefs.. so be it! But in my opinion, this make 
completes sense (while acknowledging the pain it might introduce for 
backports). FWIW, the code does not change.. so finding the function in 
the other file should be rather easy.

Thanks,
Fernando.