cvs commit: src/sys/netinet in_pcb.c

Andre Oppermann andre at freebsd.org
Mon Apr 26 11:43:00 PDT 2004


Mike Silbersack wrote:
> 
> On Fri, 23 Apr 2004, Mike Tancsa wrote:
> 
> > Hi, thanks for this commit.  I notice,
> >
> > net.inet.ip.portrange.lowfirst: 1023
> > net.inet.ip.portrange.lowlast: 600
> > net.inet.ip.portrange.first: 1024
> > net.inet.ip.portrange.last: 5000
> > net.inet.ip.portrange.hifirst: 49152
> > net.inet.ip.portrange.hilast: 65535
> > net.inet.ip.portrange.randomized: 1
> >
> >
> > How are the ranges allocated to applications ?  A quick test of 30
> > sequential outbound connections to another box on port 25 shows all
> > allocations in the 1024 to 5000 range.  Is this the way it is supposed to
> > be ?  Also, is there any caveats about moving that range from say 4000 to
> > 20000 ?
> >
> >          ---Mike
> 
> The randomization is within the selected range, not randomization between
> ranges.
> 
> You can change the first -> last range to anything you like, as long as
> you stay above 1024.  The RFCs say to use 49152->65535, but some OSes use
> 1024->32768, some use 32768->65535, so anything you pick will be equally
> ok. :)

We should change our defaults for first/last form the very limited space
to something much larger:

 net.inet.ip.portrange.first: 1024 -> 1024
 net.inet.ip.portrange.last: 5000 -> 49151

At least OpenBSD has it this big for quite some time.  AFAIK NetBSD too.

> Ignore the hifirst/hilast range, that's really only used by ftpd, it was a
> way to give ftpd a larger range of ports without changing first/last.

-- 
Andre


More information about the cvs-src mailing list