kernel header divergence

kernel header divergence

[Stephen Hemminger]

The headers in iptables command source (include/linux) are different than
the headers produced by the kernel header sanitation process:
  make headers_install

Has someone looked into rectifying this? 

Re: kernel header divergence

[Jan Engelhardt]

On Saturday 2009-02-21 01:30, Stephen Hemminger wrote:

>The headers in iptables command source (include/linux) are different than
>the headers produced by the kernel header sanitation process:
>  make headers_install
>
>Has someone looked into rectifying this? 

They were last resynced before the workshop IIRC, and in fact, the 
headers are allowed to diverge a bit. Is it a large change that gets 
introduced by resyncing to 2.6.29?

Re: kernel header divergence

[Jan Engelhardt]

On Saturday 2009-02-21 02:27, Jan Engelhardt wrote:
>On Saturday 2009-02-21 01:30, Stephen Hemminger wrote:
>
>>The headers in iptables command source (include/linux) are different than
>>the headers produced by the kernel header sanitation process:
>>  make headers_install
>>
>>Has someone looked into rectifying this? 
>
>They were last resynced before the workshop IIRC, and in fact, the 
>headers are allowed to diverge a bit. Is it a large change that gets 
>introduced by resyncing to 2.6.29?

So, I resynced it. The branch is at
	git://dev.medozas.de/iptables header-resync

One thing I noticed was that nf_nat.h lives in
${kernel}/include/net/netfilter/, from which generally no files are
exported (e.g. headers_install) at all. Oh well. So I manually copied
nf_nat.h and hand-edited out the definitely-unneeded parts.
Same goes for include/net/netfilter/nf_conntrack_tuple.h.

As it stands, a number of IPv4 NAT userspace extensions had to be
adjusted here and there a bit because - heh - the iptables headers
still were so old they carried "ip_nat_range" instead of "nf_nat_range".

Closing for today, it's "testers || pull()"; thanks :)