Re: Problems with SMC Ultra

Zoltan Hidvegi (hzoli@cs.elte.hu)
Tue, 5 Sep 1995 20:11:21 +0200 (MET DST)


> From: andrew@kryten.it.com.au (Andrew Howell)
> Date: Fri, 1 Sep 1995 18:15:08 +0800 (WST)
> Subject: Problems with SMC Ultra
>
> Hi, well we fixed our earlier UDP problems with the ultra by jumpering
> the IRQ etc instead of using the software settings. But we still get
> these kernel errors often and they impact greatly on performance,
> sometimes transfers are as slow at 10K a second.
>
> Sep 1 18:00:03 beldin kernel: eth0: Interrupted while interrupts are
> masked! isr=0x2 imr=0x0.
> ...

I had the same problem here, but it usually doesn't slow down the system. It
happens only occasionally. I have an AMD 486DX2/100, SIDE SCSI-2-IDE-Multi-IO
VLB card, 32MB RAM and an SMC EtherEZ card, but I replaced it temporarily with
an SMC-ultra and this messages appears with both cards. I run linux-1.2.13
with libc-5.0.9 (slackware-elf-beta). I have a quite complicated algorithm to
reproduce this bug.

My home machine is A. I call B with a modem, then rlogin from B to C, which
it the machine with the SMC card. Then I start pppd on C (there is no ppp on
B that's why I have to rlogin). Then I rlogin from A to B via the ppp link,
and cat a long file from zsh. It starts to print out the file then stops for
a few seconds and gives me back the shell prompt without printing the whole
file. That only happens it I use cat from zsh (that is the most mystic
part). When this happens, the eth0: Interrupted message appears in the syslog
of machine C. It may be relevant that the MTU of the ppp link was set to 296
bytes while the MTU of the ethernet between B and C is 1500. I didn't test
this too much. Note that the modem is on machine B, on the problematic
machine there is only the ethernet card. C only gets the packets from the
ethernet from B then transfers this to the ppp which goes back to machine B
via the rlogin than to me via the modem. B is a Sparc with Solaris 2.3.

This message appears randomly at other times as well but I usually discover it
later looking at the syslog.

I know that it is still possible that the problem is in the hardware
(e.g. faulty motherboard). I'll try to reconfigure the interrupt and see if
it helps. Then increase ISA IO wait-states etc.

Please, if you reply, send a CC to my mail address as I only read the digest.

Zoltan Hidvegi
hzoli@cs.elte.hu