From owner-freebsd-bugs Thu Apr 13 20:49:58 2000 Delivered-To: freebsd-bugs@freebsd.org Received: from mx1.ews.uiuc.edu (mx1.ews.uiuc.edu [130.126.161.237]) by hub.freebsd.org (Postfix) with ESMTP id D37A137B7D1 for ; Thu, 13 Apr 2000 20:49:55 -0700 (PDT) (envelope-from locke@mcs.net) Received: from pete.mcs.net (isr4186.urh.uiuc.edu [130.126.208.202]) by mx1.ews.uiuc.edu (8.9.3/8.9.3) with ESMTP id WAA16267; Thu, 13 Apr 2000 22:49:45 -0500 (CDT) Message-Id: <4.3.1.2.20000413224421.025bc9c0@popmail.mcs.net> X-Sender: locke@popmail.mcs.net X-Mailer: QUALCOMM Windows Eudora Version 4.3.1 Date: Thu, 13 Apr 2000 22:49:45 -0500 To: Bill Paul From: Peter Johnson Subject: Re: kern/17965: vr (MII-bus version in 4.0 ONLY) driver Cc: bugs@freebsd.org In-Reply-To: <200004140133.VAA00372@skynet.ctr.columbia.edu> References: <4.3.1.2.20000413185330.022f58f0@popmail.mcs.net> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed Sender: owner-freebsd-bugs@FreeBSD.ORG Precedence: bulk X-Loop: FreeBSD.org At 09:33 PM 4/13/2000 -0400, Bill Paul wrote: >Of all the gin joints in all the towns in all the world, Peter Johnson >had to walk into mine and say: > > > > > No, unfortunately, it didn't help. > > > > My test has been to download and then subsequently upload a 62 MB file > > using FTP from my FreeBSD server to my workstation, plugged into the same > > switch. With the non-MII-bus drivers this transfer completes with no > > problem. Even with the patch, the download locked up the link after about > > 40 meg had been transferred; it came back and finished after about 20 > secs, > > and the "watchdog timeout" message appeared in the system log. > >Hm. I'm running out of ideas. I still think it could be the polling, >but I'm not certain how to test this theory. One way to do it would be >to comment out all of the contents of the vr_tick() routine, and then >manually set the media with ifconfig. If the problem persists after >you do that, then it has to be something else, which would really tick >me off since I have no idea what it could be. Of course if it turns >out it is the polling, then I'll still be mad since I'm not sure how >to fix it. Yes, the problem persists after commenting out the entire contents of the vr_tick() routine. In fact, it made it worse :(. You didn't specify whether to use the patched version or the unpatched version, so I used the patched version. If you want me to try the unpatched version with vr_tick() commented out, I can do that as well. Interestingly, I did not have to manually set the media with ifconfig; autodetect seemed to work fine (I tested it both with autodetection and after setting it manually: both had the same problems we've been seeing). >Anyway, if you can try that, I would appreciate it. I'll have to see >if I can think up a better way to debug this. I'll try to help as much as I can. Peter Johnson locke@mcs.net To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-bugs" in the body of the message