From owner-freebsd-questions Fri Sep 15 09:22:47 1995 Return-Path: owner-questions Received: (from root@localhost) by freefall.freebsd.org (8.6.12/8.6.6) id JAA18894 for questions-outgoing; Fri, 15 Sep 1995 09:22:47 -0700 Received: from phaeton.artisoft.com (phaeton.Artisoft.COM [198.17.250.211]) by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id JAA18872 for ; Fri, 15 Sep 1995 09:22:40 -0700 Received: (from terry@localhost) by phaeton.artisoft.com (8.6.11/8.6.9) id JAA01252; Fri, 15 Sep 1995 09:19:54 -0700 From: Terry Lambert Message-Id: <199509151619.JAA01252@phaeton.artisoft.com> Subject: Re: Cache test software? To: Nik.Clayton@brunel.ac.uk (Nik Clayton) Date: Fri, 15 Sep 1995 09:19:53 -0700 (MST) Cc: questions@freebsd.org In-Reply-To: <17690.9509150936@molnir.brunel.ac.uk> from "Nik Clayton" at Sep 15, 95 10:36:35 am X-Mailer: ELM [version 2.4 PL24] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Content-Length: 1643 Sender: owner-questions@freebsd.org Precedence: bulk > Sort of off-topic for a FreeBSD mailing list, but the problem is > affecting FreeBSD, so: > > Can anyone on this list point me towards software that can test memory > caches? Specifically, software that can diagnose problems involving bus > mastering DMA xfers to memory that's been cached? I have detailed algorithm descriptions, which should be obvious to anyone with a working knowledge of DMA/stale-cache problems in any case, so I don't know how useful that would be. The problem with turning these into software is that you *must* have a bus mastering DMA card to do the tests, and that means a driver for the card, and that means making the test very card specific or part of the OS. Obviously, if you have the problems, it would be difficult or impossible to do this without another driver that uses PIO to get the system to the point it can run the tests (basically, my favorite hobby-horse of a fallback driver). I considered building software into a boot ROM and putting it on a DMA capable card for a plug in tester. This would be pretty straight forward, but probably not very economical unless you rolled it into a lot of other diagnostic capabilities, and even then your market would be limited to VARs/VADs/OEMs who care that the systems they ship aren't buggered (which is not many of them; maybe Rod would buy one, but he'd probably build his own first using BSD and a particular SCSI card). This is basically one-off software, and you will probably have to write it yourself. Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.