From owner-freebsd-hackers  Wed Apr  5 07:20:33 1995
Return-Path: hackers-owner
Received: (from majordom@localhost)
          by freefall.cdrom.com (8.6.10/8.6.6) id HAA21181
          for hackers-outgoing; Wed, 5 Apr 1995 07:20:33 -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 HAA21151
          for <freebsd-hackers@FreeBSD.org>; Wed, 5 Apr 1995 07:20:05 -0700
Received: (from bde@localhost) by godzilla.zeta.org.au (8.6.9/8.6.9) id AAA21676; Thu, 6 Apr 1995 00:16:45 +1000
Date: Thu, 6 Apr 1995 00:16:45 +1000
From: Bruce Evans <bde@zeta.org.au>
Message-Id: <199504051416.AAA21676@godzilla.zeta.org.au>
To: freebsd-hackers@FreeBSD.org, steve2@freefall.cdrom.com
Subject: Re: Colorado Jumbo 250MB ft, and FreeBSD 2.0R
Sender: hackers-owner@FreeBSD.org
Precedence: bulk

>> Bruce, do we need to something with the DELAY code?  I have 4 different
>> Pentiums here, 3 w/PCI if you need someone to run some test code.

>This turned out to be the culprit.  If you look at the code around line
>1473 of the driver in ftintr_wait() it is doing:

>	for (retries = 0; retries < 10000; retries++) {
>		DELAY(100);
>		...
>	}

>This used to allow up to 1 second for a recalibrate or seek to
>succeed before giving up but now it is a bit too fast for some systems.
                                          ^^^
It allows at least 0.8 seconds, probably closer to 1.0 than 0.8.

>Someone had suggested a while back that the count be bumped up to
>100000 and I thought that was a committed fix.  I thought wrong. :-(

I think a factor of 10 larger can't be right (unless 1 second was never
right :-).

There should be an interface to a global counter.  E.g., DELAY()
could simply increment a counter for each tick that it notices (it
can be changed a little so that it is guaranteed to notice all
ticks provided it is called at least once every 10msec).  Then the
above loop could be better written as;

	extern volatile u_long proposed_counter;
	u_long start_count, end_count;

	reset_proposed_counter();	/* avoid overflow */
	proposed_counter = start_count;
	end_count = start_count + 1000000 * proposed_counter_freq;
	do {
		DELAY(100);	/* maybe less */
		...
	} while (proposed_counter < end_count);

Bruce