FreeBSD multi Lun detection query

[Kenneth D. Merry]

On Wed, Apr 04, 2012 at 23:11:22 +0530, Desai, Kashyap wrote:
> 
> 
> > -----Original Message-----
> > From: Kenneth D. Merry [mailto:ken at freebsd.org]
> > Sent: Wednesday, April 04, 2012 8:58 PM
> > To: Desai, Kashyap
> > Cc: freebsd-scsi at freebsd.org; McConnell, Stephen; gibbs at freebsd.org
> > Subject: Re: FreeBSD multi Lun detection query
> > 
> > On Wed, Apr 04, 2012 at 11:03:04 +0530, Desai, Kashyap wrote:
> > > Hi,
> > >
> > > Our test team has one test case, where they need to verify multi lun
> > detection by FreeBS8.2 OS.
> > > Here are the steps to reproduce.
> > >
> > > Step 1: Configured Multi-Luns in RBOD (DOTHILL) and map to controller
> > which is booted with FreeBSD
> > > Step 2: Issue the command "camcontrol rescan all" and "camcontrol dev"
> > >
> > > Expected Result: the command should display all the configured luns
> > >
> > > Actual Result: Only the first Lun is discovered
> > >
> > > Note: When we issue explicit scan using "camcontrol rescan
> > bus:target:lun", it detects that specific lun.
> > >
> > > Query: How does FreeBSD detect multi luns ? Driver issue
> > XPT_BUS_RESCAN and expect OS should detect both the luns.
> > > Is there any limitation associated with multi lun in FreeBSD ?
> > 
> > A couple of questions:
> > 
> >  - Do you have change 231716 in your version of the driver?  The max_lun
> >    field of the path inquiry CCB was set to 0, which disabled multiple
> > LUN
> >    scanning.  I set it to 8, but in reality it needs to be set to the
> >    maximum LUN supported by the chip.
> 
> Hi Ken, This is very good information. I was not aware of this max_lun field.
> We used FreeBSD-8.2-RELEASE. I can ask test team to verify with latest upstream.(rev 231716)
> 

If you were using 8.2, were you using your own mpslsi driver with it?  I'm
guessing that your version of the mps driver didn't have the max_lun change.

> What about FreeBSD-9-RELEASE ? Is max_luns are set to non-zero. ?

The max_lun field needs to be set by the driver.  The version of the old
mps driver in 9.0 release did set the max_lun field to a non-zero value.

But if you were using the mpslsi driver, it probably didn't have that
change.

> >  - What LUN numbers do you have configured on the array?
> 
> Not sure what is correct answer for this. _but_ we have created two luns. And expected to see lun 0 and lun 1.
> 

There is no correct or incorrect answer, but both LUNs should be probed if
the max_lun field in the path inquiry CCB returned by the driver is
non-zero.

What is the maximum LUN number supported by the hardware?  You should set
the max_lun field to whatever that number is.

Ken
-- 
Kenneth Merry
ken at FreeBSD.ORG