[1277] in linux-net channel archive
Re: Solution for dosemu disconnecting during ipx/novell sessions.
daemon@ATHENA.MIT.EDU (G.W. Wettstein)
Sat Oct 28 13:06:29 1995
From: greg@wind.rmcc.com (G.W. Wettstein)
Date: Sat, 28 Oct 1995 08:48:01 CDT
In-Reply-To: Raul Miller <rdm@tad.micro.umn.edu>
"Re: Solution for dosemu disconnecting during ipx/novell sessions." (Oct 27, 10:44pm)
To: Raul Miller <rdm@tad.micro.umn.edu>
Cc: linux-net@vger.rutgers.edu, linux-msdos@vger.rutgers.edu
On Oct 27, 10:44pm, Raul Miller wrote:
} Subject: Re: Solution for dosemu disconnecting during ipx/novell sessions.
Good morning Raul, thank you for the note.
> It is almost that simple, but not quite.
>
> Hints:
>
> If you have multiple dos sessions, with multiple novell sessions open
> between the linux box and the novell server, they time out
> independently.
>
> ipx is normally a dos tsr, rather than a linux kernel service.
>
> netx is a dos tsr, but it's not handling this situation.
These hints are all very helpful. I had assumed that the timeout was
probably unique to the DOS/Novell session. I really hadn't had a
chance to try this out on a single machine.
We actually use vlm for our DOS networking layer. I suspect that just
like netx it is not responsible for doing the keep-alive stuff,
whatever that may be.
I think that the keep-alive stuff is probably in the ipx layer
someplace. On our DOS-only workstations (the ones left) a protocol
stack is loaded which is something like this:
device driver -> packet driver -> ipxodi (ipx/spx) -> vlm
The kernel driver/ipx layer essentialy takes the first three layers
out of the stack and vlm thinks that its talking directly to an IPX
layer. This would thus implicate either something in the kernel or
the ipx stuff in the dos emulator.
What I probably need to do next is to kick up a DOS/Novell session in
the emulator which uses pdether -> ipxodi -> vlm. If this session
does not timeout this would confirm that something is missing in
either the core emulator code or at the kernel ipx level. I will
probably conduct this test today.
The next step from there would be to put a DOS machine on our network
with an ethernet driver and the standard packet layer. It should be
possible to use tcpdump at this point to watch for packets to/from
this machine and see what happens when the machine sits idle for an
extended period of time. I was in the process of rigging this setup
when I stumbled across the problem with tcpdump and the 'not ip'
problem I outlined in my other note.
> Potential cures:
>
> (1) put something into the kernel which is informed enough of novell
> structures that it can form a reasonable transaction with the server
> every five minutes or so.
>
> (2) write a dos tsr that does something analogous.
>
> I suspect that (1) would be more robust -- dos tsrs tend to be flakey.
My druthers would be to find out exactly what a native DOS/ipxodi/vlm
stack 'knows' that we don't know about. There is of course the third
alternative and this is to modify the dosemu ipx core to help support
this functionality. If this problem is native only to the DOS stuff
it would be cleaner to incorporate the functionality in the emulator
core rather than the kernel.
I agree with the premise of a DOS tsr not being the route to go. Of
course providing the support in the emulator core is basically a
reduction of the DOS tsr solution.
> I've not looked at the kernel ipx code, and I don't have the expertise
> to write good tsr code for dos, so this is about as far as I go.
I appreciate the input. It confirms what I am experiencing and has
provided me with some important clues to guide my thinking. Have a
pleasant weekend.
> Raul
}-- End of excerpt from Raul Miller
As always,
Dr. G.W. Wettstein Oncology Research Div. Computing Facility
Roger Maris Cancer Center INTERNET: greg@wind.rmcc.com
820 4th St. N.
Fargo, ND 58122
Phone: 701-234-7556
----------------------------------------------------------------------
`The truest mark of a man's wisdom is his ability to listen to other
men expound their wisdom.' -- GWW
