<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>
Date   : Tue, 12 May 2009 18:30:03 +0100
From   : philb@... (Phil Blundell)
Subject: BBC B+ Econet weirdness

On Tue, 2009-05-12 at 09:50 +0100, michael.firth@... wrote:
> Having had a look at the B+ service manual, there is a 4.5s timeout on
> attempting to send a packet, so I think if it doesn't error out before
> this then the hardware will lock up.
> I've certainly left it 60s+ and it doesn't return. Escape doesn't work
> either, Break is needed to release the lockup.

That does sound rather like the NMIs from the 68B54 are getting lost
somewhere between the ADLC and the CPU.  It'd be interesting to know
what you see on a network monitor (running on some other machine
obviously) when you try to do something on this one.

> The machine in question is a B+, not a B. The B+ has space on the
> PCB for collision detection, but, quoting the service manual "the
> collision detect circuitry is not fitted because the software protocols
> should prevent any collision"

If you're on a smallish network (just a few active stations) then you
won't be having many collisions in the first place and it probably isn't
worth worrying too much about the extra hardware.  Although it isn't
true to say that the software protocols will "prevent any collision",
they do go some way towards avoiding them in the first place and
mitigating their effects when they do occur.  (There isn't any collision
detect across a bridge, for example, so in general it wouldn't be
practical to place all your faith in the hardware even if all your
stations were properly equipped.)

Anyway, excessive undetected collisions wouldn't cause the kinds of
lockup that you're seeing: they'd be more likely to manifest as
intermittent "Not listening", "Net Error" or "No reply" kind of
messages.

p.
<< Previous Message Main Index Next Message >>
<< Previous Message in Thread This Month Next Message in Thread >>