Re: no reassembly for outgoing packets on RAW socket

[Patrick McHardy <kaber@trash.net>]

Jiri Olsa wrote:
> On Thu, Jun 10, 2010 at 11:14:04AM +0200, Patrick McHardy wrote:
>   
>> Jiri Olsa wrote:
>>     
>>> On Wed, Jun 09, 2010 at 04:16:42PM +0200, Patrick McHardy wrote:
>>>   
>>>       
>>>>> If this is not the way, I'd appreciatte any hint..  my goal is
>>>>> to put malformed packet on the wire (more frags bit set for a
>>>>> non fragmented packet)
>>>>>       
>>>>>           
>>>> I don't have any good suggestions besides adding a flag to the IPCB
>>>> and skipping defragmentation based on that.
>>>>     
>>>>         
>>> ok,
>>>
>>> I can see a way when I set this via setsockopt to the socket,
>>> and check the value before the defragmentation..  would such a new
>>> setsock option be acceptable?
>>>
>>> I'm not sure I can see a way via IPCB, AFAICS it's for skb bound flags
>>> which arise during the skb processing.
>>>   
>>>       
>> Yes, a socket option is basically what I was suggesting, using the
>> IPCB to mark the packet. But just marking the socket is fine of
>> course.
>>
>>
>>     
>
> one last thought before the socket option.. :)
>
> there's IP_HDRINCL option which is enabled for RAW sockets
> (can be disabled later by setsockopt)
>
> The 'man 7 ip' says:
> 	"the user supplies an IP header in front of the user data"
>
> but does not mention the outgoing defragmentation.
>
> It kind of looks to me more appropriate to preserve the user suplied
> IP header.. moreover if there's a way to switch this off and have
> netfilter defragmentation + connection tracking for RAW socket.
>
> please check the following patch..
> (there's no special need for the IPSKB_NODEFRAG, it could check the
> socket->hdrincl flag directly..)
>
> thoughts?

My main concern is that users might expect netfilter to properly
track fragmented packets created using IP_HDRINCL.