🚀 go-pugleaf

In article <0eR3UD200VsnQrw1om@andrew.cmu.edu> Robert Andrew Ryan <rr2b+@andrew.cmu.edu> writes:
>I am trying to configure some RTs as X terminals to be connected to our
>network via slip.  They are floor models and have two built in serial
>ports, however I have been unable to get reliable results at over 4800
>baud, I get garbage and dropped characters.

I suspect that the RT is overflowing your annex box.  
Or your annex box is overflowing the RT. 

>Also if I have hardware flow control set on in the modem once I connect
>to our annex box it just sits there, if I switch the modem back to
>XON/XOFF and disconnect and re-connect to the modem (before it drops the
>annex connection) I get a bunch of backed up data from the annex.  When
>connected to an RS/6000 9600 baud and hardware flow control seem to work
>ok with the same annex.

Under AIX, it is my understanding that the RT serial port device 
driver does not support hardware flow control.  I have not looked at 
the AOS code, so I am not sure if it suffers the same limitation.

>I seem to remember reading a post here maybe a year ago about the chip
>driving the serial line having only a 1 character buffer, but that it
>was pin compatible with one with a 16 character buffer.  Or was that
>about the 4 port serial card?
>-Rob Ryan

Not sure about that rumor.  

With the VRM/AIX combination, there is quite a bit of buffer space (slack)
to store receiving characters (and its configurable!).  Unfortunately, on
sending, there's no way to throttle the RT.

Regards, Rich Coe
-----
"Pizza with everything means jalapenos"
UUCP: ...!uunet!ibmsupt!rhcoe                         IBM AIX BOS Field Quality
Internet: rhcoe@aixwiz.austin.ibm.com                 Austin, TX
internal: rhcoe@buscfg.austin.ibm.com                 VNET: rhcoe@ausvm6

Robert Andrew Ryan <rr2b+@andrew.cmu.edu> writes:
> I seem to remember reading a post here maybe a year ago about the chip
> driving the serial line having only a 1 character buffer, but that it
> was pin compatible with one with a 16 character buffer.  Or was that
> about the 4 port serial card?

The UART chip with the 16 character buffer is the NS16450.  The two 4
port serial cards I have all have these chips.  I don't know what the
1 character buffer UART chip number is, but I wasn't able to find any
NS16450's on the visible side of the tower model motherboard.

I run SLIP at 19.2kbaud, but I use a different type of SLIP server
than the annex.  Even using the NS16450, I get lots of overruns and
dropped characters.  Occasionally the serial driver just stops sending
out characters.  I modified the serial driver to not notify me about
overruns and to trigger at 4 characters (instead of 14), but it's
still losing.

I wasn't able to find any obvious problems with the serial driver's
code, so I suspect the machine is taking an inordinate amount of time
servicing higher-priority interrupts.  Then again, I have no real
information about what is going on.

In general, I'd have to say the RT makes a poor SLIP server.

				_.John

In article <weRrcgC00WBw08jnFY@andrew.cmu.edu> John Gardiner Myers <jgm+@CMU.EDU> writes:
>In general, I'd have to say the RT makes a poor SLIP server.

I never had a problem with 19,2kb SLIP using the buffered 4-port card
as long as I ran AOS 4.3 from the Sep '88 release.  The Dec 88 (final)
release, which attempts to split character processing into fast and
slow interrupt routines, seems too smart for its own good.  I never got
overruns on the Sep 88 kernel, but couldn't avoid them on the Dec 88 kernel.


-- 
Steve Dyer
dyer@ursa-major.spdcc.com aka {ima,harvard,rayssd,linus,m2c}!spdcc!dyer

Robert Andrew Ryan <rr2b+@andrew.cmu.edu> writes:
> I am trying to configure some RTs as X terminals to be connected to our
> network via slip.  They are floor models and have two built in serial
> ports, however I have been unable to get reliable results at over 4800
> baud, I get garbage and dropped characters.

The device driver in the kernel for the serial ports on the mother board
does not work reliably at speed over 2400 baud.  The device driver needs
to be rewritten (and I doubt that anyone will do it.)

> Also if I have hardware flow control set on in the modem once I connect
> to our annex box it just sits there, if I switch the modem back to
> XON/XOFF and disconnect and re-connect to the modem (before it drops the
> annex connection) I get a bunch of backed up data from the annex.  When
> connected to an RS/6000 9600 baud and hardware flow control seem to work
> ok with the same annex.

The device drivers for either type of serial port did not support hardware
flow control.

> I seem to remember reading a post here maybe a year ago about the chip
> driving the serial line having only a 1 character buffer, but that it
> was pin compatible with one with a 16 character buffer.  Or was that
> about the 4 port serial card?

The AT Serial/Parallel card and the 4 port (unbufferd) card uses a NS16450
uart.  This can be replaced with a NS16550AFN.  (I just clipped the old
uart out of an AT Serial/Parallel card and soldered in a socket and inserted
the NS16550AFN into the socket.)  This will provide reliable connections
up to 9600 baud.  To achieve even faster speeds, compile the kernel with
the a new copy of asy.c that was hacked over by Peter Honeyman and then
lightly modified by me.  The new chip and software together reliably handle
38,400 bps and hardware flow control.

Hope this helps, ...Tim

          Tim Theisen           Systems Programmer
Internet: tim@cs.wisc.edu       Department of Computer Sciences
    UUCP: uwvax!tim             University of Wisconsin-Madison
   Phone: (608)262-0438         1210 West Dayton Street
     FAX: (608)262-9777         Madison, WI   53706

John Gardiner Myers <jgm+@CMU.EDU> writes:
> In general, I'd have to say the RT makes a poor SLIP server.

I have to disagree strongly with this statement.  With the right hardware
and software the slip performance is excellent.  Here is my setup.

I have and RT at home.  I took an IBM AT Serial/Parallel card and replaced
the NS16450 with an NS16550AFN.  I ported Van Jacobson's compressed slip and
installed a new asy device driver from Peter Honeyman.  I connect to work
via two WorldBlazer V.32bis modems with V.42 and V.42bis turned on.  If I
ftp a /etc/termcap, ftp reports a transfer rate of 2kbytes/second in either
direction.  The modems are locked in at 38,400 bps with hardware flow control
on.

I usually use local xterms and rlogin to hosts at work.  However, I find it
more convenient to use xmh at work and display at home.  I don't mind waiting
a couple of seconds with wxmap draws the current weather map either.

I can give limited advice on how to achieve the same success.

Hope this helps,  ...Tim

          Tim Theisen           Systems Programmer
Internet: tim@cs.wisc.edu       Department of Computer Sciences
    UUCP: uwvax!tim             University of Wisconsin-Madison
   Phone: (608)262-0438         1210 West Dayton Street
     FAX: (608)262-9777         Madison, WI   53706

tim@appenzell.cs.wisc.edu (Tim Theisen) writes:

>John Gardiner Myers <jgm+@CMU.EDU> writes:
....
>I have and RT at home.  I took an IBM AT Serial/Parallel card and replaced
>the NS16450 with an NS16550AFN.  I ported Van Jacobson's compressed slip and
>installed a new asy device driver from Peter Honeyman.  I connect to work
....

A couple questions:

Where can I find the honeyman asy.c (i assume this is AOS we're
talking about)?

Also, has anyone had any luck getting a SECOND AT serial/parallel
card to work (for the serial port)?  I just can't get it to work at
all (/dev/ttyc1).   There's nothing obviously wrong with the driver
as far as i can tell, so i wonder if i'm missing something obvious.
--