From owner-freebsd-hackers Sat Nov 18 05:28:59 1995 Return-Path: owner-hackers Received: (from root@localhost) by freefall.freebsd.org (8.6.12/8.6.6) id FAA23936 for hackers-outgoing; Sat, 18 Nov 1995 05:28:59 -0800 Received: from godzilla.zeta.org.au (godzilla.zeta.org.au [203.2.228.19]) by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id FAA23923 for ; Sat, 18 Nov 1995 05:28:54 -0800 Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id AAA25738; Sun, 19 Nov 1995 00:24:23 +1100 Date: Sun, 19 Nov 1995 00:24:23 +1100 From: Bruce Evans Message-Id: <199511181324.AAA25738@godzilla.zeta.org.au> To: bde@zeta.org.au, hm@altona.hamburg.com Subject: Re: DELAY's in syscons Cc: freebsd-hackers@freebsd.org, hosokawa@mt.cs.keio.ac.jp Sender: owner-hackers@freebsd.org Precedence: bulk >> DELAY(n) only delays n-20 usec on an infinitely fast machine so it >If not told otherwise, pcvt does NOT use DELAY() for exactly the reason >you just describe. >Instead it uses dummy reads from port 0x84 which execute in a defined time >of ~= 1,25us. (IF they ever do execute - i was told that some implementations >do know that there is nothing to read from port 0x84 and somehow don't let >this read though to the bus, the result is, that these reads do execute very >fast). I think this only works reliably for 8MHz ISA buses. Otherwise I would have used it in DELAY(). The only way it can be reliable is if all buses know it is special and put something there that inserts wait states to extend the i/o time to 1.25us. Bruce