[SATLUG] Drive ID Change (sdx vs hdx)
Daniel J. Givens
daniel at rugmonster.org
Wed Jun 18 19:46:10 CDT 2008
On Wed, 18 Jun 2008 18:17:11 -0500
"John Pappas" <j at jvpappas.net> wrote:
> On Wed, Jun 18, 2008 at 6:11 PM, scs at worldlinkisp.com
> <scs at worldlinkisp.com> wrote:
> > ------- Original Message -------
> > > From : John Pappas
> > >AFAIK only with SATA HDDs and IDE CDRW/DVDRWs. IDE drives are
> > >still hdx.
> > ----------------------------------------------
> > Sorry John, but " sdx " applies to the IDE hard drives too.
> I just noticed that my other Suse boxes that my "knowledge" was based
> on are 10.2, and my 10.3 boxes have SATA! Is there a udev rule that
> rewrites, or is it a code level kernel change?
> I will have to see what happens when I update those boxes to 11.0
libata is the new ATA driver that is supposed to phase out the
previous drivers/ide. libata apparently uses the scsi layer, which
causes the ATA devices to look like scsi devices. It looks like it was
officially added to the kernel starting with 2.6.19 and distributions
have been migrating to the new driver since. Ubuntu happened to be a
big name leading the charge with 7.04 Feisty.
I believe there were some issues in Feisty where you had systems get
installed using the legacy ide driver. Following a kernel update at
some point, some systems started using the libata driver without
properly updating the hd[a-z] references in critical places causing
some major breakage. I believe that was what happened on the server I
mentioned in my other message. Additionally, there have been cases
where the upgrade utilities failed to update the fstab and grub
configuration to refer to the file systems by UUID, which caused
problems to say the least.
At this point, the legacy ide driver is slowly being deprecated and
libata is supporting more and more chipsets. I'm sure there will be
older chipsets that never get support and will rely on the legacy ide
driver, but most of the work done now seems to be focused on libata.
An interesting comment made back in 2003 by J.A. Magallon about libata
not using the scsi layer in 2.7. Yeah, it's been a while and 2.7
doesn't exist yet, but this just shows that it's best to stay as driver
neutral as possible. Linus and crew have shown that they aren't afraid
of making changes rapidly.
More information about the SATLUG