mcr@Sandelman.OCUnix.on.ca (Michael Richardson) (05/25/91)
I spend part of Thursday and Friday testing out a SLIP connection
between fts1 (aka fts.ocunix.on.ca) and latour (aka
latour.sandelman.ocunix.on.ca).
fts1 is a 25Mhz AMI Mark II 386 board running ISC 2.2
latour is a Sun 3/60 running SunOS 4.1(.0) with the streams based code from
Rayan Zachariassen (UofT).
Both have 8Meg of ram and a T2500.
We had installed ISC 2.2 in early April, but were plagued by a flaky
3 1/2" boot drive for a couple weeks before it got replaced. The first
time I tried it, I believe fts1 crashed when its swap and root went
offline, and that filled the SLIP with garbage, causing my (latour's)
kernel to PANIC. (There is a patch for this 100149-03.)
I initially configured the fts1 as instructed by the ISC manuals,
with sysadm, etc.. (although I recall there being some files missing!
It wouldn't create the slip login account on its own)
I wasn't impressed that the slip device had to be ifconfig'ed up in
the netd.cf file. In fact, I was so unimpressed that I commented it
out. This later confused me like crazy when I couldn't get the routes
to install properly when I later did it by hand, but forgot to
actually ifconfig it UP :-)
My reason for being less than impressed should be obvious -- you can
only support the one SLIP connection PERIOD. Not one at a time, but
just ONE. If someone can give me details on the magic that sllogin and sldialup
actually do, I'd appreciate it. (ioctl(0,I_PUSH,"slip"); ??) This
would also allow us to set up a dial out facility that knows how to
use the modem. I suspect that sllogin doesn't pay too much attention
to the carrier detect -- a login prompt didn't always show up after
the connection was lost and I reconnected.
(My telnet windows and remote xload did hang around for a couple of
minutes waiting. I found that particularly usefull, although not
unexpected once I thought about it for a minute)
The connection was V.32, no MNP. (I haven't had a chance to reenable
MNP on fts1's modem) I didn't try PEP mode, I don't expect it to do
too well. This was with GA2.00 roms (old).
A telnet window (telnet running in an xterm actually) was quite
responsive, even more so if you put it in line mode (characters echo
locally) -- but it didn't come out of line mode for things like more.
I'm not sure which end's fault it is. I haven't tested either machine
in line mode vs. char mode in loopback yet. Character mode felt
(qualitatively) better than a 2400 baud connection, perhaps as good
as a raw V.32 mode. It is hard to tell -- what counts is the time
between hitting the key and seeing the character (latency) rather than
raw throughput here. [If you type faster, your characters should come
back in spurts as the system assembles larger packets]
I pulled an xterm up across the connection -- that wasn't too bad,
there was a noticable delay though.
I had hopped that I would be able to pull up an xpcterm, and maybe
even get graphics.. Nope :-( Haven't tried a local (fts1) xpcterm
yet though. There is also obvious problems with keyboard mapping.
(where is the SysReq key on an old style 3/60 keyboard?) Finally,
xpcterm complains about not being able to open the ega font. I ftp'ed
that over (more on that in a moment) from
/usr/lib/X11/fonts/misc/ega.snf to latour's equivalent, but wasn't
sure how to let my X server know about it. I suppose terminating and
restarting Xsun might work, but I NEVER do that :-)
I ftp'ed a couple of small (80k) files between the two machines, I
got 0.57k/s on the first file, realized that I had forgot to enable
rts/cts on the Sun, did that, and got about 0.73k/s on the next two
files. The modem lights went on, stayed on, and then SD and CTS went
off, stayed off for a moment and then went steady on again. I like
that kind of performance.
Telnet (character) response wasn't too good during the
meantime. I don't think PPP fixes that problem though.
The biggest problem I had was with MTU sizes. RFC 1055, and my Sun
slip stuff recommend a 1006 byte MTU in order to be compatible -- I
can see the reasoning for increasing this with MNP modems.
How can I decrease it with the ISC slip implementation? How can I
even find out what it is? /etc/conf/mdevice.d/slip/Space.c (? - fts1
is having its super VGA card replaced with a Paradise card. The other
one seems to hose the system when you do a warm boot. Cold boots are
really nasty to drives...)
Finally, concerning routing, I would like to use gated rather than
set up static routes (you can't add a static route until the interface
has been ifconfig'ed. And I don't want to do that at boot time)
I have gated on both machines, but can't convince the Sun one to
stay up as a daemon, and I haven't begun to experiment with fts1's
version. (The default version hoses the wd0 and sl0 interfaces since
it hasn't received any routes from them)
[I'm going to post more details to comp.protocols.tcp-ip later.
Please respond to this part there.]
I've seen (and had at one time) some PPP code for SunOS 4.1, however
it seemed to be Sparc specific (which makes no sense as I thought it
was all source) so I didn't investigate. If anyone has any pointers to
a ISC 2.2 PPP or SunOS 4.1 PPP, I'd appreciate it. I can ftp (the real
thing) and/or anonuucp if needed.
Thanks.
--
:!mcr!: | The postmaster never | So much mail,
Michael Richardson | resolves twice. | so little time.
HOME: mcr@sandelman.ocunix.on.ca Bell: (613) 237-5629
Small Ottawa nodes contact me about joining ocunix.on.ca!mcr@Sandelman.OCUnix.on.ca (Michael Richardson) (05/28/91)
A day or two ago I wrote: > I spend part of Thursday and Friday testing out a SLIP connection >between fts1 (aka fts.ocunix.on.ca) and latour (aka >latour.sandelman.ocunix.on.ca). > fts1 is a 25Mhz AMI Mark II 386 board running ISC 2.2 > latour is a Sun 3/60 running SunOS 4.1(.0) with the streams based code from >Rayan Zachariassen (UofT). > Both have 8Meg of ram and a T2500. > The biggest problem I had was with MTU sizes. RFC 1055, and my Sun >slip stuff recommend a 1006 byte MTU in order to be compatible -- I >can see the reasoning for increasing this with MNP modems. > How can I decrease it with the ISC slip implementation? How can I >even find out what it is? /etc/conf/mdevice.d/slip/Space.c The above file revealed nothing at all. This problem is really annoying me. I was about to increase the MTU to 8192 on the Sun end (which just involves editing /usr/include/sys/slip.h and recompiling) when I realized that it may now be too big the other way. Short of poking around the live kernel (which I may start doing in a moment) how can I find out the value of ifnet->if_mtu? Neither ifconfig nor netstat seem to allow me to do that. Can anyone offer any advice on poking around ISC 386/ix kernels? :!mcr!: -- :!mcr!: | The postmaster never | So much mail, Michael Richardson | resolves twice. | so little time. HOME: mcr@sandelman.ocunix.on.ca Bell: (613) 237-5629 Small Ottawa nodes contact me about joining ocunix.on.ca!
Michael.Richardson@sunbrk.FidoNet.Org (Michael Richardson) (05/28/91)
A day or two ago I wrote: > I spend part of Thursday and Friday testing out a SLIP connection >between fts1 (aka fts.ocunix.on.ca) and latour (aka >latour.sandelman.ocunix.on.ca). > fts1 is a 25Mhz AMI Mark II 386 board running ISC 2.2 > latour is a Sun 3/60 running SunOS 4.1(.0) with the streams based code from >Rayan Zachariassen (UofT). > Both have 8Meg of ram and a T2500. > The biggest problem I had was with MTU sizes. RFC 1055, and my Sun >slip stuff recommend a 1006 byte MTU in order to be compatible -- I >can see the reasoning for increasing this with MNP modems. > How can I decrease it with the ISC slip implementation? How can I >even find out what it is? /etc/conf/mdevice.d/slip/Space.c The above file revealed nothing at all. This problem is really annoying me. I was about to increase the MTU to 8192 on the Sun end (which just involves editing /usr/include/sys/slip.h and recompiling) when I realized that it may now be too big the other way. Short of poking around the live kernel (which I may start doing in a moment) how can I find out the value of ifnet->if_mtu? Neither ifconfig nor netstat seem to allow me to do that. Can anyone offer any advice on poking around ISC 386/ix kernels? :!mcr!: -- :!mcr!: | The postmaster never | So much mail, Michael Richardson | resolves twice. | so little time. HOME: mcr@sandelman.ocunix.on.ca Bell: (613) 237-5629 Small Ottawa nodes contact me about joining ocunix.on.ca! * Origin: Seaeast - Fidonet<->Usenet Gateway - sunbrk (1:343/15.0)