Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 31 Dec 2003 10:20:23 -0500 (EST)
From:      John Von Essen <essenz@bjork.quonix.net>
To:        Matthew Seaman <m.seaman@infracaninophile.co.uk>
Cc:        freebsd-questions@freebsd.org
Subject:   Re: (2) rsh and rcp problems between Solaris and FreeBSD
Message-ID:  <Pine.BSF.4.10.10312311009560.6247-100000@bjork.quonix.net>
In-Reply-To: <20031231095115.GA98228@happy-idiot-talk.infracaninophile.co.uk>

Next in thread | Previous in thread | Raw E-Mail | Index | Archive | Help
One more thing. Apparently, if I do 'rsh -n host cmd' on the Solaris box,
it no longer hangs, and I can do it back to back indefinitely. Say I do
ten of them, 5 secs apart. I still see the following 10 times in netstat:

tcp4       0      0  mx100.841              embryo.bluebell..1014
TIME_WAIT

After 30 secs they go away.

On Solaris 2.6, the -n to rsh is:

     -n             Redirect the input of rsh to /dev/null.   You
                    sometimes  need  this  option to avoid unfor-
                    tunate interactions between rsh and the shell
                    which  invokes  it.   For example, if you are
                    running rsh and invoke a  rsh  in  the  back-
                    ground  without  redirecting  its  input away
                    from the terminal, it will block even  if  no
                    reads  are posted by the remote command.  The
                    -n option will prevent this.


This doesn't affect rcp, so those are still slow. The only other thing is
that I am going through a firewall, from an internal network to a dmz.


-John

On Wed, 31 Dec 2003, Matthew Seaman wrote:

> On Tue, Dec 30, 2003 at 11:42:41PM -0500, John Von Essen wrote:
> > 
> > I have a Solaris 2.6 box that has been sending data to a Solaris 8 box 
> > via rsh and rcp.
> > 
> > I finally changed the Solaris 8 box to a FreeBSD 4.9-STABLE machine.
> > 
> > Unfortunately, I am noticing alot of problems with my rsh and rcp 
> > calls. Again, the rsh/rcp calls are being initiated on my Solaris 2.6 
> > and are hitting a FreeBSD 4.9-STABLE box.
> > 
> > Here is what happens:
> > 
> > My first rsh works, but if I try another rsh within a few seconds it 
> > takes a really long time (30 - 60 sec) to return - but it does return 
> > successful. If I issue my rsh calls every 2 minutes, it returns quick 
> > everytime. But if I do rsh calls to close together (5 sec delays) they 
> > hang for a long time.
> 
> Now that is weird.  30-60 second delay sounds like classic DNS
> breakage, but in that case you'ld see it the first time you connected
> and probably subsequent times.
> 
> How are you doing name resolution on this system -- host files, NIS,
> DNS, something else?  Are you using Kerberos at all?  Does toggling
> the use of the '-D' and '-n' flags in inetd.conf on the FreeBSD side
> make any difference?
> 
> Hmmm... does this happen all of the time, or do you get a grace period
> of a few minutes immediately after rebooting the FreeBSD box?  Are you
> perhaps ending up with an awful lot of connections sitting in
> CLOSE_WAIT stage on the FBSD box?
>  
> > The rcp behaves the same way - but with an added oddity... I can't seem 
> > to 'rcp -r' directories. For example, say I have /tmp/test and in there 
> > I have three files (a, b, and c.). When I try to rcp -r that directory, 
> > I get the following:
> > 
> > # rcp -r /tmp/test host:/tmp
> > rcp: /tmp/test/a/b: Not a directory
> > rcp: /tmp/test/a/b/c: Not a directory
> > 
> > Very weird!
> 
> Does saying:
> 
>     # rcp -r /tmp/test host:/tmp/
> 
> (note the trailing '/') make a difference?  This is by analogy to
> cp(1) where trailing slashes do have a similar sort of effect -- I
> think that's a feature of BSD-ish Unices but not SysV-ish flavours.
>  
> > Anyone have any ideas? If I can't get this resolved I am going to have 
> > to go back to the old SUN to SUN setup and scrap the FreeBSD machine.
> 
> rcp(1) and rsh(1) are really considered as legacy stuff on FreeBSD
> nowadays.  Most people will strongly advise you to use ssh(1) and
> scp(1) instead -- those are standard on Solaris 9 but you'll have to
> compile yourself up a copy on Solaris 2.6.  You can use key based
> authentication with ssh-agent(1) in order to avoid having to put in
> passwords all the time: see the SSH FAQ at
> 
>     http://www.snailbook.com/faq/no-passphrase.auto.html
> 
> Note too that sshd(8) under FreeBSD disallows root access by default,
> but there's a pretty obvious control in the /etc/ssh/sshd.conf config
> file.
> 
> 	Cheers,
> 
> 	Matthew
> 
> -- 
> Dr Matthew J Seaman MA, D.Phil.                       26 The Paddocks
>                                                       Savill Way
> PGP: http://www.infracaninophile.co.uk/pgpkey         Marlow
> Tel: +44 1628 476614                                  Bucks., SL7 1TH UK
> 



Want to link to this message? Use this URL: <http://docs.FreeBSD.org/cgi/mid.cgi?Pine.BSF.4.10.10312311009560.6247-100000>