Date: Fri, 28 Nov 1997 10:23:00 -0700 From: "Justin T. Gibbs" <gibbs@plutotech.com> To: dg@root.com Cc: isp@FreeBSD.org, gibbs@FreeBSD.org Subject: Re: Adaptec 2940 and FreeBSD 2.2.2-RELEASE Message-ID: <199711281724.KAA21915@pluto.plutotech.com>
next in thread | raw e-mail | index | archive | help
>> Michael Slater wrote: >> >> > Hello, >> > I have heard that the Adaptec 2940 driver has had problems in the pas >t. >> > Is this sorted out in 2.2.2-RELEASE ? > >We have had ongoing problems with our Adaptec 2940UW servers since the >2.1.7 debacle. Our servers were gloriously stable up through 2.1.5, >unreliable since. 2.2, 2.2.1, and 2.2.2 have shown no improvement, same >old story that others have mentioned, especially the spontaneous lockups >with "SCSI bus resets." Haven't tried with 2.2.5, we're probably >converting to another OS since it looks like it's not going to get fixed. >Personally, I'm tired of getting up at 2am to power-cycle a server, if I >enjoyed that we'd be running NT :) > It's actually already fixed, but not in any stable or current release. The main problem you are facing is that the SCSI code FreeBSD has used for some time simply wasn't designed to perform adequate error recovery. So, I started rewriting the SCSI layer a few months ago to deal with these kinds of problems. That work is available for running under FreeBSD-current, but it currently only supports the aic7xxx and advansys controllers with disk and cdrom devices. Tape support is close to being ready. Worm device and other controller support will follow after that. You may want to talk to Satoshi Asami who has been running the CAM code for some time on one of his servers. This server would crash readily under the standard SCSI code, but now handels bus stuffups without a hichup (His IBM drives look like they have a firmware bug...). I'm hoping that CAM will be integrated into current early next year, but I'll need some help from Stephan (to port the NCR driver) to make that the case. Even if you don't choose to try this solution to your problem, at least now you know that there is (and has been for some time) work in progress to make the SCSI subsystem much more robust. -- Justin
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199711281724.KAA21915>