robert@hslrswi.UUCP (01/21/86)
Hanging Interlan NI-1010A board on 4.2bsd Vax-11/750
====================================================
Has anyone had experience of an Interlan NI-1010A Ethernet board
repeatedly hanging ?
This problem seems to happen at random. It has happened a few times out
of the blue for no apparent reason but the easiest way to make it
repeat itself is to do an Rdump down the Ethernet from a remote machine.
This isn't a sure-fire way of forcing it to hang but will do the job
nicely 50% of the time.
When it does happen, all current connections through that board are
lost. The output error count as reported by `netstat -i' climbs rapidly.
Unfortunately, a reboot is the only known solution.
Very rarely, the console also prints the message
il0:stray xmit interrupt, csr=ffffffd0<CDONE,CIE,RIE>
One further point worth mentioning is that we have installed the fixes
to /sys/netinet/tcp_input.c (circa lines 308 and 390) as recommended by
Andrew Russel of Sun UK (sunuk!ar) and which are supposed to solve a
similar problem (losing connections when rdumping Sun to Vax). In
practice, though, these fixes don't seem to make any difference.
The system is a (almost) vanilla 4.2bsd Vax-11/750 with 3MB, UDA-50,
RA-80, RA-81, TU-78, DZ-11, VMZ-32 and NI-1010A. The remote machine
(being rdumped) is a Sun-2/120.
If anyone has any idea or knows of the cure for this problem, please
let me know. (If anyone also knows a way of restarting the Ethernet
board without resorting to a reboot, that would also be extremely
useful information). All help will be very gratefully received.
Many thanks in advance,
Cheers,
Robert.
******************************************************************************
Robert Ward, ___________
Hasler AG, Murtenstrasse 137a, CH-3008 Bern, Switzerland | _ |
| _| |_ |
Tel.: (+41) (31) 65 23 19 | |_ _| |
Bitnet: hslrswi!robert@cernvax.bitnet | |_| |
Arpa: hslrswi!robert%cernvax.bitnet@WISCVM.ARPA |_________|
Edunet: hslrswi!robert%cernvax.bitnet@UCBJADE.Berkeley.EDU
Uucp: ... {seismo,decvax,ukc, ... }!mcvax!cernvax!hslrswi!robert
******************************************************************************chris@umcp-cs.UUCP (Chris Torek) (01/26/86)
This is a hardware problem in certain Interlan boards. Some do it, some do not. You just got a bad one. I think it is possible to reset the board without powering off, but we have only had it happen two or three times on one 750, and I forget whether a Unibus init fixed things, or whether we power cycled the UBA. -- In-Real-Life: Chris Torek, Univ of MD Comp Sci Dept (+1 301 454 4251) UUCP: seismo!umcp-cs!chris CSNet: chris@umcp-cs ARPA: chris@mimsy.umd.edu
ks@pur-ee.UUCP (Kirk Smith) (01/27/86)
In article <146@hslrswi.UUCP> robert@hslrswi.UUCP writes: > > Hanging Interlan NI-1010A board on 4.2bsd Vax-11/750 > ==================================================== > >Has anyone had experience of an Interlan NI-1010A Ethernet board >repeatedly hanging ? All you need to do is install a timeout in the driver. Start the timer when you transmit a packet. If 2 seconds go by without a transmit complete interrupt, re-init the board, (drop the packet), and take off running again. We had to do this within 1 week of installing 4.2BSD networking. With us, this condition happens under normal use about 10 times a day. Kirk Smith Purdue Engineering Computer Network
mts@cosivax.UUCP (Michael Stolarchuk) (01/27/86)
I think its hardware. The first set of interlan tranceivers we got all had the same problem. It was a bit tough trying to install the first network. Some of the early transceivers had (what interlan called) a jabber jumper. It determined how much garbage the transceiver would put up with. You may want to call interlan and ask them about the specific transceiver you have. Since we remove the jabber jumper, no serious problems. -------------------- Empty-S in Ann Arbor, Michigan. --------------------