[CALUG] Disk alignment and boundaries -- WAS: Fedora upgrade

Bryan J Smith b.j.smith at ieee.org
Mon Jun 20 13:25:40 EDT 2011


Did you mean plus (+) instead of star (*)?  The latter is used for active (not 
required with Linux, or a non-Windows MBR).

The problem is that 63/255 sectors/heads is _non-aligned_.  That's why one gets 
the plus (+) at times.  The problem is the legacy cylinder/heads/sectors (CHS) 
geometry.

Newer fdisk versions in Linux distribution releases no longer use CHS, but raw 
sectors -- be it 512b or 4KiB.  This makes things easier to deal with, and get 
perfect alignments.  As I mentioned, I use 1GiB myself.

As far as boot chain, that has nothing to do with fdisk.  Fdisk is not even 
related to such.  The boot chain is dynamic in GRUB, handled and generated at 
boot-time.  And Windows has an extremely dumb boot approach to complicate 
matters in dual-boot.

uEFI could have solved it better than it did.  But so far, it's just doing what 
ARC did prior.  You have a FAT slice where you locate boot loaders.  It is read 
dynamically, but the rest of the boot is handled as before.

I will say one thing.  At least with Linux, you can address boot issues.  With 
NT, you're f'd if it dorks up, which has been the case with some bare metal 
Windows systems of ours on uEFI.

________________________________
From: Walt Smith <waltechmail at yahoo.com>
Sent: Mon, June 20, 2011 10:32:06 AM

I've always been NOT happy with the little stars that show
in fdisk displaying partition sizes. (I'm not complaining about fdisk per se..
I think it was a good idea to display the info ).  

Windows seems to want more precise
boundaries, or used to..   Haven't tried anything lately though.

But Bryans suggestion is a good one: and I would add/amplify:

There should be a way to make ( create ) nice partition boundaries
to make the "star" go away ( rather than some user calculation on
a note pad ):  either an option in fdisk or 
a separate tool.  (you'll notice I'm not detailing the meaning
of "star").  I never got it down  quite perfect why other OS's
needed some boundary and others didn't.  ...  I believe there
was some discussion ( on another list ??)
regarding some defect in fdisk.   I'm not sure there was a defect so much as
it's design was to simply make the part size exactly as the user told 

it to.  Although perhaps round-down was a problem if the size entered
was "mega-something byte" rather than 123 sectors... sector boundary ???

Also, fdisk allowed more than one partition to be marked "bootable".
I don't have an explanation for that.  I'd like to see that fixed if
theres not a reason.  Perhaps a marker for each partition part of the boot chain 

??

An fdisk display also does not show the boot chain.
There have been times when I'd simply forget what the boot sequence
was  and have to trace it again..   I mean the path thru the partitions,
not the kernel boot vmlinuz.

If the boot is straight from MBR, there could be a simple text display
showing:
MBR =>  /dev/hda2

if the boot is chained, 

MBR => /dev/hda1 =>  /dev/hda5   ( primary partition )

MBR= > /dev/hda4(LBA) => /dev/hda6 

( chain is on partition of LBA, where LBA is encapsulation for logical hda6,7,8 
)



More information about the CALUG mailing list