Subject: Re: CVS commit: syssrc
To: Jason R Thorpe <devnull@autogen.freeservers.com>
From: Bruce Korb <bkorb@pacbell.net>
List: port-macppc
Date: 05/17/2001 18:15:32
Last 4/29 you wrote:
> Anyway, is it my imagination, or did Tsubai's commit revert this
> change??

It was not your imagination.  Tsubai's commit reverted the change
and did paper over the GMAC problem.  Unfortunately, he has never
bothered to talk to anyone about his patch and his patch works by
throwing messages away until the hardware starts putting messages
into the slot the driver thinks is correct.  It probably won't take
too long for 32 messages to get thrown away, so the hiccup doesn't
last very long.  See:

  http://www.NetBSD.org/cgi-bin/query-pr-single.pl?number=11529

I have actually been trying to communicate with Tsubai since before
I started making public postings on the issue.  No luck.  In any
event, my experience with the patch above was that messages were
not dropped.  I did do a good deal of research and "EWAG" before
I wrote my widely dispersed posting.  In fact, that posting did
draw some responses that indicate that others have resolved the
problem by resetting the device when it hiccups.  Even Tsubai's
fix is better than that.

So, WRT your last posting on the subject, you should not see the
system "seem to" hang for long anymore, but you will loose 30 or
so messages every time it happens, unless you apply my patch.

Regards,
	Bruce