Re: [PATCH] update raw patch in POM

[Patrick McHardy <kaber@trash.net>]

Roberto Nibali wrote:
> Well, maybe we're not in sync anymore regarding my test conducts:
> 
> Given, that
> a1: wait-for-untracked-refs patch
> a2: nf_reset patch
> 
> 
> I have following test chart:
> 
> 
>            |   a1 enabled   |  a1 disabled   |
> -----------+----------------+----------------+
> a2 enabled |  fix, no hang  |  not tested    |
> -----------+----------------+----------------+
> a2 disabled|    30s hang    |     crash      |
> -----------+----------------+----------------+

Yes, I got that.

>>If the nf_reset patch helps it must be somewhere after
>>ip_finish_output2(). This could be an arp queue, a device queue
>>or a packet socket queue. You could add an nf_reset to dev_queue_xmit
>>and see if it changes anything.
> 
> Maybe the above sketch explains the confusion I might have created.

I think both patches together are the correct fix, I just wanted
to understand where the hang comes from to be sure.

>>There's no raw table in 2.4, so the wait-for-untracked-refs patch should
>>go in the 2.4 raw table patch.
> 
> 
> Pablo, shall I do it and submit for POM?

Yes, please.

>>For 2.6 I'm not sure yet, Pablo has a
>>patch that changes how untracked works and makes this patch unnecessary.
> 
> Hmm, is it possible to backport this to 2.4.x or do the changes need the 2.6.x
> netfilter framework?

It is possible, but it is not yet decided if we're going to
submit that change.

Regards
Patrick