Subject: Re: Still hangs with i386 ethernet?
To: William O Ferry <WOFerry+@CMU.EDU>
From: Mason Loring Bliss <>
List: current-users
Date: 09/07/1998 18:39:03
On Mon, Sep 07, 1998 at 12:53:19PM -0400, William O Ferry wrote:

> A few months ago I was having a problem with NFS occasionally
> hanging between two NetBSD/i386 boxes.  I recall other people seeing the
> problem, and thought it had been fixed (unfortunately I was away from my
> machines since about that time).

Hm... This matches behaviour I'm seeing. Running 1.3 beta on one box and
1.3.2 on my main box worked fine, but now that I'm running -current on my
main box, I'm seeing hangs. The machines don't go down, but the affected
processes refuse to die. Killing them doesn't work... kill -9 doesn't
work... A reboot works, though. <grin> But that's a bad solution.

A symptom I see is receive ring buffer overruns (?) on the slower of the
two machines.

I'm moving my 1.3 beta box up to 1.3.2, as I'd hoped that would correct the
problem, but it looks like the trouble might be on the -current end, from
what you're saying.

FWIW, I build -current fairly frequently on my main box, so I imagine that
I'd be a reasonable tester for fixes...

My two systems:

486DX/40, eight megs of RAM, we-based ethernet. (I can't see what sort it
is at the moment, given the excessive numbers of ring buffer overruns on
dmesg shows me... Heh!) This was running 1.3 beta, and by tonight will be
running 1.3.2.

AMD K6 233mmx, sixty four megs of RAM, PCI Realtek 8029 Ethernet. This is
running -current, compiled the last time compiles were still working. (I'm
seeing the ld problem that others have mentioned in here.)


Mason Loring
"In the drowsy dark cave of the mind dreams build their nest with fragments
  dropped from day's caravan."--Rabindranath Tagore..awake ? sleep : dream;