Re: [PATCH] TCP simultaneous open support

[Patrick McHardy <kaber@trash.net>]

Jozsef Kadlecsik wrote:
> Hi Patrick,
> 
> On Tue, 2 Jun 2009, Patrick McHardy wrote:
> 
>> Thanks Jozsef. Just one question:
>>
>>> -	TCP_CONNTRACK_LISTEN,
>>> +	TCP_CONNTRACK_SYN_SENT2,
>> is this constant already used in userspace? From your conntrack
>> output above I'd assume the answer is yes, so we should keep it
>> around as an alias to avoid compilation errors.
> 
> libnetfilter_conntrack has got it's own header file which defines 
> TCP_CONNTRACK_LISTEN and it's string interpretation as "LISTEN" state. 
> 'conntrack' relies on libnetfilter_conntrack when translating the 
> numerical id received from the kernel to the string value, that is the 
> reason for the "misinterpretation" of the state. So from 
> libnetfilter_conntrack and conntrack-tools point of view we can safely 
> remove TCP_CONNTRACK_LISTEN.
> 
> I'm not aware of any project which would use the kernel header file 
> directly, but that does not mean there isn't any. We might keep the 
> symbol, of course. Shall I resend the patch with TCP_CONNTRACK_LISTEN 
> preserved?

Thanks for the explanation. I think we should keep it around just
to make sure. No need to resend though, I'll just add it myself.

Thanks.