From owner-freebsd-hackers  Sun Sep 22 21:01:41 1996
Return-Path: owner-hackers
Received: (from root@localhost)
          by freefall.freebsd.org (8.7.5/8.7.3) id VAA21722
          for hackers-outgoing; Sun, 22 Sep 1996 21:01:41 -0700 (PDT)
Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19])
          by freefall.freebsd.org (8.7.5/8.7.3) with ESMTP id VAA21655
          for <freebsd-hackers@FreeBSD.org>; Sun, 22 Sep 1996 21:01:32 -0700 (PDT)
Received: (from bde@localhost) by godzilla.zeta.org.au (8.7.6/8.6.9) id NAA03404; Mon, 23 Sep 1996 13:54:40 +1000
Date: Mon, 23 Sep 1996 13:54:40 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199609230354.NAA03404@godzilla.zeta.org.au>
To: msmith@atrad.adelaide.edu.au, terry@lambert.org
Subject: Re: splash-page on bootup..
Cc: brandon@glacier.cold.org, freebsd-hackers@FreeBSD.org,
        jgreco@brasil.moneng.mei.com
Sender: owner-hackers@FreeBSD.org
X-Loop: FreeBSD.org
Precedence: bulk

>> This is only because we are stupid and use DELAY() instead of a
>> calibrated timer list of one-shot outcall functions.  Using a
>> spinloop is just inherently stupid.
>
>Uh?  You are still thinking like a CS guy, not a hardware programmer.
>
>When I am looking for some hardware in a probe routine, I want _nothing_
>_nada_ _zip_ happening behind my back.  No interrupts, no "strategically
>placed callbacks", nothing.
>
>This has nothing to do with how DELAY() works, it's basic 'least surprise'
>stuff.

Yes, DELAY() works like it does because it must work at probe time before
timer interrupts are enabled.  DELAY() shouldn't be used at other times
except for very short delays (short = about the same time as it takes
to initialize a one-shot timer and service a one-shot timer interrupt
(10-50 usec).

OTOH, probes in lkm'ed drivers can timer expect interrupts to work, and
shouldn't use DELAY() or otherwise hog the cpu any more than the driver
would for normal operations.

Bruce