[comp.unix.ultrix] uucp protocol problem

marc@okeeffe.Berkeley.EDU (Marc Teitelbaum) (02/03/88)

You didn't need to convince me there was a problem, decvax did 
a good job of that!  I had a line monitor and was able to see the
problem clearly.  It was interesting because the problem is inherent
in *many* uucp implementations, but it was the different timout values
which caused the bug to surface on ultrix/non-ultrix connections
over noisy phone lines.

I was surprised it wasen't found earlier because i think the person
who bumped the timouts did so to improve communications between
ucbvax and decvax, and this was clearly ultrix/non-ultrix and noisy.
But that was before i came on the scene and i believe that person
has since left the group, so the details are lost in antiquity.
Actually, Jim Bloom would probably remember.  At any rate, I put
the 4.3 g-proto in Ultrix 2.0 so the problem should go away.

As for the adb patches on 1.2 uucico's: just make sure the numbers
your changing look like the ones George listed.  I can't officially
endorse adb'ing binaries (ultrix support would kill me), but 1.2
isn't officially supported anymore anyway (they want you to be
running 2.0 or later), so have fun.  Of course its not nearly as
fun as "adb -k -w /vmunix /dev/mem"...

Marc

-------------------------------
Marc Teitelbaum			+1-415-643-6448
457 Evans Hall
Computer Systems Research Group,  CSRG
University of California
Berkeley, CA 94720

grr@cbmvax.UUCP (George Robbins) (02/05/88)

In article <20348@felix.UUCP> marc@okeeffe.Berkeley.EDU (Marc Teitelbaum) writes:
> You didn't need to convince me there was a problem, decvax did 
> a good job of that!
...
>					 	 At any rate, I put
> the 4.3 g-proto in Ultrix 2.0 so the problem should go away.

That's good to hear!  I tried to send you E-mail to ask what you had come
up with, but address you gave me seemed to be unreachable from here and
I had done the unforgivable act of losing the phone number of somebody
on the inside.  8-)

My current big bitch with Ultrix uucp is that when you are running in
"shared" modem mode, uucp doesn't drop DTR after extracting the line
from the clutches of getty.  This is never, err well, almost never a
bad idea, and is most beneficial with the newer "smart" modems like
the Trailblazers.  This is 1.2 again, so perhaps it's already fixed.

There's a fair amount of stuff over in comp.dcom.modems about fun with
trailblazers, some of which I will repost here once I think I have a
complete "solution".

BTW, for the person who was asking about the relative desirabity of Ultrix,
the answer depends greatly on the circumstances at your site, of which you
said zilch.  In my opinion, the choice depends mostly on what style and
degree of support you are looking for, whether you can obtain a 4.3 source
distribution, what purposes the system is supposed to serve and whether you
tolerate "improvements" made with the best of intentions.

Finally, I guess I should include the mandatory patch of the day.  This
one is for changing the "window size" for uucp for improving thruput over
X.25 style connections.  I don't know whether this has any beneficial
effect with the high-speed modems, and it may actually be detrimental
to the reliability of normal 1200-2400 baud modem lines.  In other words,
you milage may vary...

The following patch is applicable to Ultrix 1.2 uucio, other versions
are probably pretty much the same but you'll have to locate the right
instructions yourself.  You can try other window sizes, but I believe
the legit values are limited to those between 1 and 7.

1) you must be root (or the setuid bits will go away!)
2) uucico must not be running (ps -aux | grep uu)
3) you should know what you are doing...
4) copy the old uucico to some safe place
5) make sure the numbers match

Script started on Thu Feb  4 03:04:30 1988
# cd /usr/lib/uucp
# cp uucico uucico.nopatch
# adb -w uucico
pkopen+58?i
_pkopen+58:	movl	fffffffc(fp),r0

_pkopen+5c:	cvtlb	$3,29(r0)

_pkopen+60:	movl	fffffffc(fp),r1

_pkopen+64:	movb	29(r0),28(r1)

pkopen+5c?i
_pkopen+5c:	cvtlb	$3,29(r0)
?x
_pkopen+5c:	3f6
?w 07f6
_pkopen+5c:	3f6	=	7f6
?i
_pkopen+5c:	cvtlb	$7,29(r0)
$q
# ^D
script done on Thu Feb  4 03:06:15 1988