Hi Alin, On Thu, Dec 13, 2018 at 06:36:56AM +0100, Alin Năstac wrote: > On Thu, Dec 13, 2018 at 2:28 AM Pablo Neira Ayuso <pablo@xxxxxxxxxxxxx> wrote: [...] > > I can see in the existing code: > > > > #1 do DST manip case: so range.flags, .min_proto and .min_addr are set. > > #2 In case of paired expectation in place, we overwrite these range fields > > again. > > #3 No override for the _SRC case. > > > > If your patch can help disentagle this code a bit by: > > > > #1 check for paired expectation, if so do handling, return. > > #2 check for _SRC manip needed, return. > > #3 existing _DST manip case. > > DST manip handling is needed at least in one RTP case: on the > expectation created by the INVITE send from LAN. So you see, I can't > return from this function just because one manip logic applies, doing > so will break the case where the applied manip is equivalent to no > operation. > > > It may be larger patch, but we skip 'range_set_for_snat'? > > What is wrong with using booleans? If you want, I can use range.flags > for the same purpose (I would have to reset it to 0 after doing DST > manip), but I thought this would make code reading more difficult. > Besides, if I would use return in the loop I would have to duplicate > the spin unlocking within the loop and this will look ugly. Right. No worries. I was just exploring. Please, just submit a v2 patch that restricts this to RTP/RCTP and we go. Thanks!
