From owner-freebsd-current  Thu Jun 22 22:15:49 1995
Return-Path: current-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id WAA21323
          for current-outgoing; Thu, 22 Jun 1995 22:15:49 -0700
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.34])
          by freefall.cdrom.com (8.6.10/8.6.6) with ESMTP id WAA21317
          for <current@freebsd.org>; Thu, 22 Jun 1995 22:15:44 -0700
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id PAA26155; Fri, 23 Jun 1995 15:13:12 +1000
Date: Fri, 23 Jun 1995 15:13:12 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199506230513.PAA26155@godzilla.zeta.org.au>
To: current@freebsd.org, jhs@vector.eikon.e-technik.tu-muenchen.de
Subject: Re: kernel response
Sender: current-owner@freebsd.org
Precedence: bulk

>When I print via the standard spooler (lpr -#2), a 300K file, going to an hp3p
>laser via a parallel port, all screen updating goes dead for several seconds,
>( the -#2 doubles the data, doesnt use the PCL5 print twice escape sequence).

Interrupt driven lpt output can't go faster than about 150K/sec to an
ISA port (each byte takes at least 4 i/o's in the driver and 2 in the
interrupt handler (unless AUTO_EOI_1 is defined - it should be the
default but isn't) and each i/o takes > 1usec).  If the printer can
keep up with 150K/sec, then it will generate a solid block of interrupts
and timeout routines, including the one that updates the screen, will
not run for several seconds.  20 seconds for 3MB of output :-(.

>So it seems the parallel port output flush is running at too high a priority
>for too long.

It isn't all that high, but it's higher than none, so it blocks timeouts,
and it's equal to tty priority, so it may block echoing of type-ahead
even for the console.  sio uses timeouts so lpt interrupts certainly
block everything except the interrupt handler.

Bruce