Ethereal-users: RE: [Ethereal-users] Runing ethereal with a remote xwindow displa y

Note: This archive is from the project's previous web site, ethereal.com. This list is no longer active.

From: "HOOD, Andy" <ahood@xxxxxxxxxxxxxx>
Date: Fri, 17 Oct 2003 12:21:34 +1000
Guy,

Pardon for the top-post, but I think you might have mis-interpreted.

My interpretation of David's problem was that Ethereal works if his X server
is on the same box as he is running Ethereal (Linux), but crashes if the X
server is on Solaris. That suggests the problem isn't in Ethereal.

If he was running Ethereal on Solaris he would get sigbus errors for
unaligned accesses.

> -----Original Message-----
> From: Guy Harris
> 
> On Wed, Oct 15, 2003 at 07:25:13PM -0700, Kaas, David D wrote:
> > I connect from a Sun (2.6 or 2.7) to a Linux rehat 9 
> system.  I start
> > etheral (0.9.13) and the display opens on my sun 
> workstation ok.  I can
> > start a capture but when I select Stop the capture windows 
> closes and just
> > after the packets start to appear in the ethereal windown 
> that window also
> > closes.  At times another windows pops open for just a 
> second, an error
> > maybe, but closes to fast to read it.  If I start a capture 
> and selecte
> > "Update list of packets in real time" both windows close as 
> soon as the
> > first packets appear.
> 
> "Windows close" usually means "Ethereal exits immediately", which
> usually means "Ethereal crashes with a bad pointer dereference" or
> something such as that.
> 
> If you're running Ethereal from the command line, are you getting any
> "core dumped" messages?
> 
> > It runs ok if I run directly from the linux system console.
> 
> If you run it directly from the Linux console, you're running it on a
> little-endian machine whose processor has no problem dereferencing
> unaligned pointers.  If you run it on the Sun - *regardless of whether
> you're displaying to a remote machine or displaying on a Sun
> workstation's console* - you're running it on a big-endian 
> machine whose
> processor traps (causing a signal that kills the process) when
> dereferencing unaligned pointers.
> 
> I.e., this probably has nothing to do with whether the 
> display is remote
> or local, and everything to do with whether the machine on 
> which you're
> doing it can dereference unaligned pointers (or is big-endian, or is
> running Solaris, or...) - if you had a Sun workstation on 
> your desktop,
> and were running Etheral on a Linux server, I suspect it 
> wouldn't crash,
> but would crash with the same traffic on your workstation.  
> (The traffic
> you're seeing might also be an issue - if the bug is in the dissector
> for the XXX protocol, it'd show up only if you captured XXX protocol
> traffic that triggered the bug.)
> 
> For better or worse, I suspect most Ethereal developers are working on
> machines with x86 processors (which are little-endian processors that
> have no problem dereferencing unaligned pointers), so bugs where
> Ethereal derefereences unaligned pointers (or, less likely, doesn't do
> the right byte-order-canonicalization) don't show up.
> 
> There may be files in "/tmp" or "/var/tmp" on your Sun with names
> beginning with "ether" and having a bunch of letters and numbers after
> them.  Try reading those capture files in Ethereal; if the window
> disappears when you try that, those files are capture files containing
> packets that, due to Ethereal bugs, cause it to crash.  If 
> so, then, if
> you could send us those files (note that they contain packets 
> from your
> network; try reading them on your Linux box instead, and, if you can
> read them, see whether they contain any data you don't want 
> us to read),
> that'd help us try to find the bug.

Regards,
Andrew Hood
A distributed system is one in which the failure of a computer you
didn't even know existed can render your own computer unusable. --
Leslie Lamport, as quoted in CACM, June 1992

<<application/ms-tnef>>