Date: Tue, 26 Nov 1996 13:53:14 +1100 From: Bruce Evans <bde@zeta.org.au> To: j@uriah.heep.sax.de, julian@whistle.com Cc: freebsd-current@freebsd.org Subject: Re: 2.2-ALPHA install failure Message-ID: <199611260253.NAA21607@godzilla.zeta.org.au>
next in thread | raw e-mail | index | archive | help
>ENOUGH ALREADY! Sorry, more last words... >When I implimented the dummy fdisk table in the boot loader, I did >it simply to allow people who had loaded the previous style of >bootblocks, and who had used IT's style of "dangerously dedicated" >to be able to replace the old bootblocks with the new ones and >still be able to boot. When I implemented the slice code for 2.0.5 I had to keep supporting these old boot blocks, so they still haven't gone away. >the end sect/cyl/head numbers are NEVER USED except by FreeBSD Not true. At least the ncr SCSI BIOS uses them to get some hints about the geometry. H=255/S=63 gives hints that are inconsistent with the size of 50000, so the hints are ignored. >Ehead = Bios #heads -1 >ECYL = 2 (or maybe 1?) >Esectors = Bios# sectors >note that the slice SIZE does not come into this calculation at all.... >to try get sme hints about the geometry, and I tend to believe in >setting them to the following would be best. I think you mean the preceding. This has always been a requirement (for FreeBSD on drives that you want to put a FreeBSD partition on). >this would ensure that we could derive the geometry from it.. No, it would ensure that we couldn't tell if the geometry derived from it is the correct one (unless the size is adjusted to match). >but the NCR bios might not like it too much Yes, it would also ensure that the NCR BIOS couldn't tell if the geometry derived from it is the correct one. >but then, who cares.. it'd still work I believe. It depends. It gives the NCR default, which is OK. However, the geometry may change when another OS writes a partition table entry that follows the rules (this shouldn't be a problem unless the other is confused about the geometry), and the geometry may not work with other controllers, so it is better to force it to 255/63 or 64/32 if you can. >this would require altering disklabel to fiddle these bits >(which it presently does not do), of setting sysinstall to do >it in a compatible manner. Disklabel is a BSD utility for handling labels. It should not know anything about partition tables of foreign OS's. >personally I NEVER EVER EVER use dedicated mode. >I find it too useful to have the 'dead area' before the >first partition, and hell it's only 32K usually! Same here (except on my newest SCSI disk I tried using dedicated mode to test it. It isn't usable since I need a partition selector to boot non-BSD partitions on other drives. This should be supported in our bootblocks). Bruce
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199611260253.NAA21607>