Subject: Re: Silo overflows (was: Communications package
To: Multiple recipients of list <port-arm32@NetBSD.ORG>
From: Robert Black <r.black@ic.ac.uk>
List: port-arm32
Date: 05/19/1997 14:42:15
On May 17,  1:20pm, Rich 'Wolflord' Bradshaw wrote:
> Subject: Silo overflows (was: Communications package
> On Fri, 25 Apr 1997, Rich 'Wolflord' Bradshaw wrote:
>
> > Anyone know what silo overflows are? Are they serious?

Incoming serial data is stored in a buffer on the serial chip. This buffer is
called a silo. When the silo reaches a threshold a serial interrupt is
triggered. This interrupt empties the silo into a normal memory buffer. The
interrupt handler takes a while to figure out what the interrupt is and empty
the silo. If the interrupt is flagged whilst serial interrupts are disabled
then the interrupt will not occur until they are reenabled. During this delay
(the interrupt latency)  serial data may continue to arrive and fill up the
silo. A silo overflow occurs when the computer takes too long to empty the
silo.

> > On a possibly related note, running an X session using fvwm as the window
> > manager occasionaly goes a bit screwy. I notice a large amount of silo
> > overflows, no ibuf overflows and no uart failure.
> >
> > What happens is, either the windows stop autofocussing, it remains on a
> > single window. Or it stops accepting keyboard input in X only. If I switch
> > back to a virtual terminal I can log in OK, and the X session is still
> > running.
> >
> > Switching back to the X session, the problem still persists although I've
> > found it can be cleared by restarting fvwm. It doesn't need to be exitted,
> > just restarted.

This might be an fvwm problem. There will be a new release of X at around the
same time as the new console RSN (initially the console will not be any use for
doing anything other than starting X).

Cheers

Rob