[OpenIndiana-discuss] The 4k transition & the 512bit emulation in hdds

Reginald Beardsley pulaskite at yahoo.com
Thu Mar 10 03:15:10 UTC 2011

Long ago, drives allowed jumper selecting any of several sector sizes (e.g. 512 to 4096 or 8192 in powers of two).  Then the PC came along w/ a large market and things settled on 512 byte hardware sectors.  Over the years we've told the software an assortment of lies to get around  our lack of prescience.  Now we're looking for a new set of lies to tell the software.  Same song different verse :-(

--- On Wed, 3/9/11, c d <cdlist at hotmail.no> wrote:

From: c d <cdlist at hotmail.no>
Subject: [OpenIndiana-discuss] The 4k transition & the 512bit emulation in hdds
To: OpenIndiana-discuss at openindiana.org
Date: Wednesday, March 9, 2011, 3:40 PM


So the hdd makers are pushing out 4k drives, but are failing miserably by emulating 512bit sectors to the OS.

My understanding is that OI supports 4k sectors – as long as the drive reports this natively. But since all new 4k drives – as far as I'm aware – does an 512bit emulation, there is no way the OS would know about it.
This means that you need an external tool to properly align the partitions. Since there's no such tool for ZFS, it therefore rely on manually editing and compiling the source of zpool, to force an alignment.

The half-hearted way this transition has been done by the hdd makers, simply speaks for itself.
Since the emulation takes place in the PCB of the drive, it baffles me that they aren't release two versions of their drives. Why not make it optional? Even if they were slightly more expensive, it would make the transition flawless for a lot of people, by simply, well not having a transition!
Instead they are pushing the problem onto the consumers. And it seems that it can take years before they actually make the drive speak 4k natively.

So, as this problem concerns all of us, perhaps we should try to find a solution?
If all it takes is a modified binary, how about supplementing it with OpenIndiana?

Here's a link to tests and how-to for compiling zpool with ashift=12:

I'm also very interested in hearing your experiences with the following drives, and how they perform with / without the modified binary:
WD20EARS        (Western Digital Green) 
HD204UI         (Samsung F4EG)
ST32000542AS     (Seagate Barracuda LP 2)

- Frederic
OpenIndiana-discuss mailing list
OpenIndiana-discuss at openindiana.org


More information about the OpenIndiana-discuss mailing list