I wrote:
>
> I am having a sporadic, frequent problem with getty. The configuration is a
> Sun Sparc2 running SunOS v4.1.1 with the jumbo DNI and serial port patches
> installed. About every day or two the following hits:
>
> Jan 7 16:02:06 sun44 getty: ioctl(TCGETS): Bad file number
> Jan 7 16:02:06 sun44 getty: ioctl(TCGETS): Operation not supported on socket
> Jan 7 16:02:06 sun44 getty: ioctl(TCGETS): Operation not supported on socket
>
> But nothing seems to be wrong. I am assuming that getty could not open a
> socket for some reason and gets the above error. When it happens, DTR gets
> dropped (Telebit T2500) and never gets raised, causing all incoming dialups
> to fail (NO ANSWER). Does anyone have any ideas. I will summarize if
> any answers are received (post or mail). Thanks!
My thanks to the following people:
dem@meaddata.com (David Myers) for a whole *slew* of messages and
the correct solution.
vic@cd.com (Vic Serbe)
Kevin Cosgrove 642-2676 <qiclab!solomon!kevinc>
It became apparent that this topic was discussed at length some time ago. As
a result I stongly urge that it appear in the FAQ as SUN, according to one
sysadmin, will not respond to this problem. The problem is that the Telebit
T2500 takes longer to reset than the serial port driver allows for. The
proper method to correct this situation follows:
==============
>From mail supplied by dem@meaddata.com (David Myers) who in turn had it
from eggert@twinsun.com (Paul Eggert)
This is a all-too-well-known problem. Here's how to fix it on a Sparc.
Go to your kernel configuration files, find zs_async.o, and patch it as follows:
# adb -w zs_async.o
zsadtrlow?W 6
$q
Then rebuild your kernel. (You can patch the kernel directly if you like.)
I haven't tried this on a Sun-3 but I wouldn't be surprised if the
same fix worked there, too.
zsadtrlow controls how long DTR is kept low when a connection is dropped.
Sun ships it with the value of 3, but some modems (notably Telebit T2500s)
need a larger value; without it, SunOS sometimes gets confused and keeps DTR
low forever after a connection is dropped.
Question: is it true that in Solaris 2.0, you won't be able to make
patches like this with Sun software without forking over extra bucks?
That is, will adb be unbundled too?
Credits: jh@moon.nbn.com (John Harkin) posted this patch in
<1991Nov6.181150.25181@moon.nbn.com>. He credited Erik J. Murrey
(ejm@riscit.NOC.Vitalink.COM) and Dean Long (dlong@oak.ucsc.edu) for
the original fix.
=================
It was also suggested to modify the gettytab file as follows:
From: Kevin Cosgrove 642-2676 <qiclab!solomon!kevinc>
I forgot to mention that I added "to#60" to /etc/gettytab
to restart the getty after a 60 second timeout. If you've
been having to kill your getty's to get them to restart,
this could really help.
=================
Thanks to all who responded!
-- Bert Medley medley@synercom.hounix.org [work] bert@medley.ssdl.com [home]
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:07:24 CDT