Re: /sys/.../enclosure_device:<something> ?

[James Bottomley <James.Bottomley@HansenPartnership.com>]

On Fri, 2016-01-15 at 14:32 -0500, Anne Mulhern wrote:
> 
> 
> 
> ----- Original Message -----
> > From: "James Bottomley" <James.Bottomley@HansenPartnership.com>
> > To: "Anne Mulhern" <amulhern@redhat.com>
> > Cc: linux-scsi@vger.kernel.org
> > Sent: Wednesday, January 13, 2016 3:30:07 PM
> > Subject: Re: /sys/.../enclosure_device:<something> ?
> > 
> > On Wed, 2016-01-13 at 15:15 -0500, Anne Mulhern wrote:
> > > 
> > > 
> > > 
> > > ----- Original Message -----
> > > > From: "James Bottomley" <James.Bottomley@HansenPartnership.com>
> > > > To: "Anne Mulhern" <amulhern@redhat.com>,
> > > > linux-scsi@vger.kernel.org
> > > > Sent: Wednesday, January 13, 2016 2:15:05 PM
> > > > Subject: Re: /sys/.../enclosure_device:<something> ?
> > > > 
> > > > On Wed, 2016-01-13 at 11:28 -0500, Anne Mulhern wrote:
> > > > > Hi!
> > > > > 
> > > > > I'm looking for the most precise information available about
> > > > > the
> > > > > meaning of <something>
> > > > > and from what it is derived. I'm also interested in what the
> > > > > values
> > > > > in the files in that directory
> > > > > may mean. Could somebody point me there?
> > > > 
> > > > It's in drivers/misc/enclosure.c:enclosure_link_name()
> > > > 
> > > 
> > > Thanks. I will do my best with that.
> > > 
> > > > > Also, how can there be more than one enclosure_device
> > > > > subdirectory
> > > > > for the same device,
> > > > > or, alternatively, why isn't <something> stored in a file in
> > > > > the
> > > > > more
> > > > > regularly named directory
> > > > > /sys/.../enclosure_device?
> > > > 
> > > > I'm not sure I parse the question, but if you're asking how can
> > > > a
> > > > single enclosure bay appear to have more than one device,
> > > > that's
> > > > when
> > > > there are multiple paths to the device.
> > > 
> > > I'm afraid the word path is so overloaded...I'm assuming that you
> > > are talking about a multipathed environment, where two device
> > > nodes,
> > > say /dev/sdb, /dev/sdc correspond to the same disk, and are hence
> > > multipathed. Of course, in that case, it makes sense that both
> > > devices should share a bay. And that's what I observe.
> > 
> > Yes, that's what I meant.
> > 
> > > If I understand you correctly, then my question is actually the
> > > opposite. Why does it look as if the directories are set up so
> > > that
> > > there can be multiple bays for a single device...i.e., the
> > > directory
> > > naming scheme allows me to specify an arbitrary number of bays
> > > for
> > > /dev/sdb. It seems like there should never be more than one?
> > 
> > There is only one bay per device since a physical disk can't be in
> > more
> > than one bay at once.  Why do you think it was set up to allow >1?
> > 
> 
> Purely because the naming scheme for the directory,
> 'enclosure_device:<identifier>' supports more than 1.
> 
> This directory pops up in the directory of the parent device
> representing
> a block device, and should represent just the bay for that device,
> AFAIU.
> 
> The most compelling reason for this naming scheme to support more
> than 1
> subdirectory is if it is necessary, i.e., there are more than one
> possible.
> 
> That might be true if my understanding of the block device's parent
> device's
> meaning is incorrect or if something else doesn't mean what I think 
> it means.

The enclosure_device: is a namespace prefix.  It's sysfs, so we can't
have two files of the same name and <identifier> is supplied by page 7
of the enclosure device ... it could be anything.  The only way to
ensure we don't have a clash is to use a namespace prefix.

James