Date: Wed, 20 Mar 1996 13:22:29 -0700 (MST) From: Terry Lambert <terry@lambert.org> To: andreas@knobel.gun.de (Andreas Klemm) Cc: dave@kachina.jetcafe.org, freebsd-hackers@freebsd.org Subject: Re: Adding a damn 2nd disk Message-ID: <199603202022.NAA27703@phaeton.artisoft.com> In-Reply-To: <Pine.BSF.3.91.960320190738.18629B-100000@knobel.gun.de> from "Andreas Klemm" at Mar 20, 96 07:25:37 pm
next in thread | previous in thread | raw e-mail | index | archive | help
> > [...] > > *Why* the hell is adding a 2nd disk to a currently running OS Guru > > level lore? > > [...] > > Yesterday evening I started to add my 2nd SCSI Harddisk to my > home System. Looks like I have 5 GB now ;-) > > Well, although I knew directly, that I have to fiddle around with > disklabel, I'd suggest to people like you, to first do the most > obviuos thing ... check the manpages. > > The command 'apropos disk' gives you lot's of ideas, where > to search for pieces of information. The most important here: > > ... > disklabel(5) - disk pack label > disklabel(8) - read and write disk pack label > disktab(5) - disk description file > ... Stop right there. the concept of "disktab" is fundamentally flawed. It assumes that it is not possible to determine the disk size from the controller, and it assumes uniform boundry recording for seek optimization. The seek optimization is, in reality, useless because ZBR media makes it very difficult (without SCSI II extended queries) to determine the real cylinder boundries to do the optimization. It is no longer a simple mathematical relationship... and if that were not enough, geometry translation pounds an oak stake into its heart. Finally, seek optimization is predicated on the idea that is no longer true: that seek times are many more times more expensive than rotations. This has changed by about a factor of 100 since the optimization first went in, but the code hasn't changed at all. This baically leaves default sector sparing settings, sector counts designed to be on (no longer applicable) cylinder boundries, and slice geometries that should be interactively determined instead of frozen in an obsolete data file. The disktab should go. I don't know if I speak for anyone else, but the two things that have kept me from jumping in, muzzles of my C compiler flashing, are: 1) Devfs is a necessary framework to really fix most of these problems in the most general and broadly applicable way possible. 2) Representational geometry (ie: a good user interface is a real bitch to design -- I want something that a GUI tool could open, but which is command line based, that won't be too hard to use in either mode). Help lobby to get rid of mknod, MAKEDEV, and specfs, and put in devfs by default, and you will be 90% of the way to "disk heaven". Terry Lambert terry@lambert.org --- Any opinions in this posting are my own and not those of my present or previous employers.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199603202022.NAA27703>