From owner-freebsd-stable Fri Sep 1 12:12:54 2000 Delivered-To: freebsd-stable@freebsd.org Received: from cgi.sstar.com (cgi.sstar.com [209.205.176.12]) by hub.freebsd.org (Postfix) with ESMTP id 0F73837B423 for ; Fri, 1 Sep 2000 12:12:52 -0700 (PDT) Received: from bluto.jimking.net (bluto.jimking.net [216.54.255.8]) by cgi.sstar.com (8.9.3/8.9.3) with ESMTP id OAA83941; Fri, 1 Sep 2000 14:12:50 -0500 (CDT) (envelope-from jim@jimking.net) Received: from jking (jking.lgc.com [134.132.75.164]) by bluto.jimking.net (8.9.3/8.9.3) with SMTP id OAA69520; Fri, 1 Sep 2000 14:12:48 -0500 (CDT) (envelope-from jim@jimking.net) Message-ID: <004d01c01448$9dfabcf0$a44b8486@jking> From: "Jim King" To: "Warner Losh" Cc: "Gary Kline" , References: <002d01c01444$350c8b00$a44b8486@jking> <200009010501.WAA54972@tao.thought.org> <200009011832.MAA37168@harmony.village.org> <200009011853.MAA37372@harmony.village.org> Subject: Re: bad 16550A maybe? Date: Fri, 1 Sep 2000 14:12:48 -0500 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.00.3018.1300 X-MimeOLE: Produced By Microsoft MimeOLE V5.00.3018.1300 Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.ORG Warner Losh wrote: > In message <002d01c01444$350c8b00$a44b8486@jking> "Jim King" writes: > : I think we'd see a lot fewer of these messages on the mailing lists if sio.c > : set the 16550A trigger level to 8 bytes instead of 14 bytes. That's the > : first thing I do when I see this problem, and that usually fixes it; much > : easier than replacing the misbehaving hardware. :-) > > For fast machines, this is a lose since it double the interrupt rate, > but does give a *MUCH* better interrupt latency tolerance. I had to > do this when i had a 16550A attached to a custom card via PCI. I > couldn't make it have fast interrupts, so I did the next best thing by > increasing the number of interrupts and checking more often (the card > itself had about half a dozen interrupts going all the time, so long > as I could afford a 16ms latency I was fine). But it's a pretty small lose in the grand scheme of things, and it would prevent a lot of problems in the field (judging from the amount of times this comes up on -stable and -questions). FreeBSD and 16-bit Windows are the only systems I've used that defaulted to setting the trigger to 14. (They also share the unfortunate characteristic that you have to re-compile something to change the default; Windows was more unfortunate in that you had to own the Windows DDK to do this.) All the DOS-based communications software I used when the 16550A first hit the scene used 8 (usually easily changed in a configuration file), and all the 32-bit versions of Windows default to 8 (easily changed in a fancy GUI configuration dialog). If we're going to hard-code this we should hard-code a middle-of-the-road setting, not something that's wonderful on some boxes but awful on others. Jim To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-stable" in the body of the message