I haven't seen any logging about fragments being blocked. I just now tried
with 'keep frags' added in and got the same behaviour.

On Fri Oct 11 12:41:31 2002, Wes Zuber wrote:
> 
> hmm, I always use keep state keep frags in my pass out statement. Could 
> this be an issue?
> 
> --Wes
> 
> On Friday, October 11, 2002, at 12:09  PM, Iain Morgan wrote:
> 
> > Hello,
> >
> > We have ipf v3.4.29 running on a Solaris 8 system. For the most part,
> > things work correctly. However under some circumstances established
> > TCP connections seem to fail par way through.
> >
> > This is particularly noticable when the system is backed up by piping 
> > the
> > output of ufsdump through ssh to our mass storage system. What appears 
> > to
> > be happening is that the incoming packets cease to match the 
> > state-table
> > entry. The outgoing packets continue to match the entry.
> >
> > I have a kludge that works around the issue, but I was wondering if
> > anyone had any similar problems and perhpas a better solution.
> >
> > The relevant rules in the ipf.conf are:
> >
> > block in log all
> > pass out proto tcp from any to 129.99.0.0/16 port = 22 flags S keep 
> > state
> >
> > # Special handling for lou to allow backups to work
> > pass in  proto tcp from 129.99.248.41 port = 22 to any
> >
> > I've used ipfstat -t to confirm that an entry for the connection does
> > get created and is mode 4/4. I've tried using ipmon and snoop to 
> > pinpoint
> > the problem to no avail.
> >
> > Thsi problem occurs with both 3.4.28 and 3.4.29. The Sun Workshop 6.0U1
> > compiler was used to build ipf.
> >
> > --
> > Iain Morgan
> > NAS Desktop Support Group
> >
> 


--
Iain Morgan
NAS Desktop Support Group

Reply via email to