>Number:         1769
>Category:       kern
>Synopsis:       repeated TCP connections from a HP-UX to Netbsd causes timeouts
>Confidential:   no
>Severity:       serious
>Priority:       medium
>Responsible:    kern-bug-people (Kernel Bug People)
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Thu Nov 16 19:20:05 1995
>Last-Modified:
>Originator:     Erik Berls <cyber@ecst.csuchico.edu>
>Organization:
	
>Release:        951109
>Environment:
	
System: NetBSD lestat 1.0A NetBSD 1.0A (GENERIC) #7: Sun Sep 3 01:53:23 PDT 1995 thorpej@basalt:/work/netbsd/src/sys/arch/hp300/compile/GENERIC hp300

NetBSD cindy 1.1_ALPHA NetBSD 1.1_ALPHA (Cindy) #0: Mon Nov 13 12:55:07 PST 1995     cyber@cindy:/usr/src/sys/arch/hp300/compile/Cindy hp300

>Description:
Repeated remsh's from a HP-UX 9.05 and 10.01 box to NetBSD boxes causes
timeouts.  Target netbsd boxes include: netbsd/sparc and netbsd/hp300.
On the netbsd/hp300 both 1.0 and -current were tested.
At least one remsh in ten will time out, sometimes as many as 3.

This has also been tried from SunOS 4.1.3 boxes and IRIX 5.3 boxes with
no problems.

HP had the following to say when i gave them a dump of traffic to anylyse:
> After looking over the traces of the HP system trying to do
> repeated remshells to the customer's netbsd system it appears that
> the HP is sending out TCP SYN packets to the netbsd system at
> the time of the failure and the netbsd system is not responding.
> The only thing of interest is that we are making all of our
> connections from source TCP port 1023 but this is perfectly valid.
> 
> There does not appear to be much else we can do here. It would be 
> up to someone to take a look at what is going on on the Netbsd system
> since it is not responding to our initial TCP SYN packets. We may be
> able to persue this a bit further but strictly on a Time and Materials
> basis and even then we may not be able to get much more information.
> 
> I suspect that the netbsd system is running into some limitation or
> is having problems clearing out it's old connections.


>How-To-Repeat:
The script:
#other names have been substituted, this is just the one that was used
#when sending a report to HP.
remsh lestat w
remsh lestat w
remsh lestat w
remsh lestat w
remsh lestat w
remsh lestat w
remsh lestat w
remsh lestat w
remsh lestat w

run on the HP-UX box:
# nettl -tn pduin pduout -e ns_ls_driver -f /tmp/raw
# ./foo
Permission denied.
Permission denied.
Permission denied.
Permission denied.
Permission denied.
Permission denied.
rcmd: connect: lestat.ecst.csuchico.edu: Connection timed out
Permission denied.
Permission denied.
# nettl -tf -e ns_ls_driver

The 'nettl' commands turn on tracing from the HP-UX side.  I still have this
trace, but it is quite large.

>Fix:
I wish i knew.
>Audit-Trail:
>Unformatted: