Date: Fri, 26 May 1995 14:16:16 -0400 (EDT) From: Peter Dufault <dufault@hda.com> To: rgrimes@gndrsh.aac.dev.com (Rodney W. Grimes) Cc: davidg@Root.COM, hsu@cs.hut.fi, hackers@FreeBSD.org Subject: Re: MAJOR problem with FreeBSD-2.0-RELEASE Message-ID: <199505261816.OAA06465@hda.com> In-Reply-To: <199505261716.KAA02456@gndrsh.aac.dev.com> from "Rodney W. Grimes" at May 26, 95 10:16:54 am
next in thread | previous in thread | raw e-mail | index | archive | help
(sent to -hackers instead of bugs) Rodney W. Grimes writes: > > > Properly resetting the SCSI bus, the host adapter, renegotiating > > sync transfers, waiting for all devices to come ready again and > > getting their "bus device reset occurred" message, reaping all > > outstanding I/O transactions, and then retrying those outstanding > > transactions is an effort that includes modifying all the host > > adapter drivers (and looking for a common interface to pull up out of > > them) and so will be a tough job to adequately test. It should > > also be done in conjunction with better I/O transaction scheduling > > to cleanly support tag queuing. This is a 2.1 adventure. > > This is a 2.2 adventure, we are not going to do that kind of massive > work in any part of the system until after 2.1 ships. We simply > can not afford the risk factor at this time. This is of cource, IMHO. Some pieces of this has to be done.. Something that lets us live through devices grabbing the SCSI bus without splattering things should be done. I started I/O transaction scheduling a while ago and put it on hold with the code freeze. The above sketch is my overall take on the best way to do it. Based on when 2.1 is to be released and what the definition I can look for something lower risk. -- Peter Dufault Real Time Machine Control and Simulation HD Associates, Inc. Voice: 508 433 6936 dufault@hda.com Fax: 508 433 5267
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199505261816.OAA06465>