Re: [PATCH] sctp: allow authenticating DATA chunks that are bundled with COOKIE_ECHO

[Vlad Yasevich <vyasevich@gmail.com>]

On 06/12/2015 07:26 AM, Neil Horman wrote:
> On Thu, Jun 11, 2015 at 05:27:45PM -0700, David Miller wrote:
>> From: mleitner@redhat.com
>> Date: Thu, 11 Jun 2015 14:49:46 -0300
>>
>>> From: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
>>>
>>> Currently, we can ask to authenticate DATA chunks and we can send DATA
>>> chunks on the same packet as COOKIE_ECHO, but if you try to combine
>>> both, the DATA chunk will be sent unauthenticated and peer won't accept
>>> it, leading to a communication failure.
>>>
>>> This happens because even though the data was queued after it was
>>> requested to authenticate DATA chunks, it was also queued before we
>>> could know that remote peer can handle authenticating, so
>>> sctp_auth_send_cid() returns false.
>>>
>>> The fix is whenever we set up an active key, re-check send queue for
>>> chunks that now should be authenticated. As a result, such packet will
>>> now contain COOKIE_ECHO + AUTH + DATA chunks, in that order.
>>>
>>> Reported-by: Liu Wei <weliu@redhat.com>
>>> Signed-off-by: Marcelo Ricardo Leitner <marcelo.leitner@gmail.com>
>>
>> Vlad/Neil, please review.
>>
> 
> sorry Dave, though I had sent email on that already.
> 
> I had an initial concern that there could be a race in which a previous
> iteration of sctp_outq_flush would move some chunks to a packet, but not flush
> it to the network layer yet (due to not being full), and that would result in
> the same condition.  But since this only happens with a COOKIE_ECHO chunk (which
> is a control chunk), we should be ok, as those are sent immediately.

Neil.  I don't think this race can happen since outq manipulation always happens under
a socket lock and so do socket options.  So, we are guaranteed that outq will not change
in this case.

Acked-by: Vlad Yasevich <vyasevich@gmail.com>

-vlad

> Acked-by: Neil Horman <nhorman@tuxdriver.com>
>