Return to BSD News archive
Path: sserve!newshost.anu.edu.au!munnari.oz.au!news.Hawaii.Edu!ames!elroy.jpl.nasa.gov!swrinde!cs.utexas.edu!uunet!pipex!sunic!isgate!veda.is!adam From: adam@veda.is (Adam David) Newsgroups: comp.os.386bsd.bugs Subject: Re: Problem with Julian's SCSI drivers Message-ID: <CDEzuF.519@veda.is> Date: 15 Sep 93 21:32:24 GMT References: <26reei$e4o@news.ysu.edu> <CD7EyC.4Ax@nas.nasa.gov> <CDErsF.1Cx@cbnewsj.cb.att.com> Organization: Veda Systems, Iceland Lines: 33 dwex@mtgzfs3.att.com (David E. Wexelblat) writes: [about dynamically configured device numbers] >Which is a horribly bad way to do things. SVR4 does this, too. I hate it. >What happens when you remove a device, or one dies? If I have disks at >SCSI IDs 0, 1, and 2, and disk 1 dies, then the system will see disk 2 >as disk 1 at boot time, and all of my filesystems will fail to get mounted >or get mounted from the wrong disk. Disks don't die :-) Besides, if a disk early in fstab fails, the system wouldn't be much use anyway. >Who could possibly think that this is a good idea? What is wrong with >always having the SCSI ID in the name/minor-number? It seems the ONLY >correct way to do things. It saved me having to dismantle and fiddle with the hardware once when I had a working kernel configured for 2 disks and /usr on the third disk (by SCSI id). All I had to do to get a more or less normal base system up again was to unplug the second disk, then I could recompile a working 3-disk kernel. Normally, however I would agree with the original complaint that assigning device ids in the order that disks are found is often a silly idea. What I would like to see is a file that maps SCSI ids to device ids and back. It could either be compiled into the kernel or read from /etc/scsitab. Editing /etc/scsitab would have fixed my emergency more simply and perfectly, if that could have somehow be read at boot time for the initial device configuration. -- adam@veda.is