Does FreeBSD start slices at head boundaries?
Rick Miller
vmiller at hostileadmin.com
Fri Jul 6 21:13:05 UTC 2012
I went through this exercise to determine if there were boundary
issues installing FreeBSD on disks. I concluded that FreeBSD was
indeed installing at head boundaries. A colleague then pointed me to
http://ivoras.net/blog/tree/2011-01-01.freebsd-on-4k-sector-drives.html
which calls into question whether sysinstall and fdisk really are
installing FreeBSD's slice at the 64th cylinder. Should I be
concerned with this?
This came about due to a scenario where Linux would start its
filesystem at sector 63, right before the head boundary. On I/O
intensive applications, it was common for reads/write to cross the
head boundary resulting in unnecessary disk thrashing and long I/O
wait times. The issue was corrected in Linux by changing the start
cylinder to 2048. Some theorized that FreeBSD was vulnerable to this
scenario.
Thoughts/feedback?
On Fri, Jul 6, 2012 at 11:09 AM, Rick Miller <vmiller at hostileadmin.com> wrote:
> Hi All,
>
> Installing FreeBSD 8.x I select "A" at the fdisk partition editor to
> use the entire disk. It creates an unused slice with offset 0 and 63
> sectors in size. Then partition 1 starts at sector 63 and utilizes
> the remaining disk space. Does sysinstall's diskPartitonEditor macro
> automatically start partitions at head boundaries? The reason I ask
> is because I am most familiar with sector 64 being the start of a head
> boundary as opposed to 63. Is my understanding incorrect?
>
> --
> Take care
> Rick Miller
--
Take care
Rick Miller
More information about the freebsd-questions
mailing list