svn commit: r195817 - head/usr.sbin/sysinstall

John Baldwin jhb at freebsd.org
Mon Jul 27 13:34:25 UTC 2009


On Saturday 25 July 2009 12:09:00 am Scott Long wrote:
> M. Warner Losh wrote:
> > In message: <4A6A7F2D.90807 at samsco.org>
> >             Scott Long <scottl at samsco.org> writes:
> > : M. Warner Losh wrote:
> > : > In message: <4A68DA56.60301 at samsco.org>
> > : >             Scott Long <scottl at samsco.org> writes:
> > : > : M. Warner Losh wrote:
> > : > : > In message: <4A66D0F4.4030108 at FreeBSD.org>
> > : > : >             Alexander Motin <mav at FreeBSD.org> writes:
> > : > : > : Colin Percival wrote:
> > : > : > : > Author: cperciva
> > : > : > : > Date: Wed Jul 22 03:50:54 2009
> > : > : > : > New Revision: 195817
> > : > : > : > URL: http://svn.freebsd.org/changeset/base/195817
> > : > : > : > 
> > : > : > : > Log:
> > : > : > : >   Remove the "dedicated disk mode" partitioning option from 
sysinstall, in
> > : > : > : >   both the disk partitioning screen (the 'F' key) and via 
install.cfg (the
> > : > : > : >   VAR_DEDICATED_DISK option).  This functionality is currently 
broken in 8.x
> > : > : > : >   due to libdisk and geom generating different partition 
names; this commit
> > : > : > : >   merely acts to help steer users away from the breakage.
> > : > : > : >   
> > : > : > : >   Submitted by:	randi
> > : > : > : >   Approved by:	re (kensmith)
> > : > : > : > 
> > : > : > : > Modified:
> > : > : > : >   head/usr.sbin/sysinstall/disks.c
> > : > : > : 
> > : > : > : Is there any other way to not align FS block to the ugly legacy 
63
> > : > : > : sectors per track boundary with sysinstall now? I think RAIDs 
won't be
> > : > : > : happy. May be it would be better to fix it?
> > : > : > 
> > : > : > The reason we do it like this is because the boot blocks occupy 
the
> > : > : > first cylinder group.  Which with today's fake goemetry is 63
> > : > : > blocks...
> > : > : > 
> > : > : 
> > : > : I already covered this.  Unfortunately, Randi didn't take my bait. 
> > : > : Anyways, '63' is a minimum, it's not a singular value.  It's also
> > : > : an incredibly bad default.
> > : > 
> > : > Yes.  256 is just as good as 63, unless the geometry reports more than
> > : > 256 heads, which I don't think is possible...
> > : > 
> > : > Warner
> > : 
> > : If there is a BIOS in existence that can't find sector 256, then it 
> > : likely can't boot off of anything other than partition 1 in the fdisk
> > : table.  I bet that bugs like that were likely caught 15-20 years ago.
> > 
> > I think you were missing the point I was trying to make, I must have
> > been unclear.
> > 
> > If there were more than 256 sectors in the first cylinder, then
> > starting at 256 would overlap.  This is largely a theoretical
> > problem, I think, since I don't think that the BIOS can report a
> > geometry > 64.
> > 
> > The only issue that I can think of is that FreeBSD's fdisk wants to
> > have things on cylinder boundaries, which 256 isn't, so fdisk will
> > bitch and moan...
> > 
> > Warner
> 
> Ah, we will never be freed of the CHS bonds, will we.  Oh well.

GPT doesn't use C/H/S and thus doesn't have the weird limitations of the MBR + 
BSD label arrangement.  As soon as sysinstall can install to a GPT-labelled 
disk you will be free of the bonds in sysinstall.  I think the easiest way to 
allow for that w/o dealing with the headaches of fdisk.c and label.c and 
libgeom support in sysinstall, etc., etc. is to add a sort of "cheat" mode to 
sysinstall where you can say "look, I've already partitioned my drives and 
newfs'd my filesystems and they are mounted at '/foo', go install all the 
bits into there".  We could then let people setup their disks using command 
line tools with the livefs or what is in the install rescue build using GPT 
or ZFS or gmirror, etc.  Getting /etc/fstab correct in that instance becomes 
a bit more tricky, but not impossible.  (If nothing else one could take the 
output of mount -p, find all the paths starting with /foo, chop off the 
leading /foo, and write that out as an initial /etc/fstab.)  That would at 
least allow people to do more creative things with disks with sysinstall 
until the disk labeller can be overhauled.

-- 
John Baldwin


More information about the svn-src-head mailing list