[Bug 191348] [mps] LSI2308 with WD3000FYYZ drives disappears after hotswapping

[bugzilla-noreply at freebsd.org]

https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=191348

Stephen McConnell <slm at freebsd.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |slm at freebsd.org

--- Comment #4 from Stephen McConnell <slm at freebsd.org> ---
Hi Michal, this is Stephen McConnell from Avago Tech (formally LSI Corp.).  I
looked at this, but I'm not sure what I can do next.  I was not able to
reproduce the problem on my system.

I used the 16.00.00.00-fbsd driver and v10 of FreeBSD and I was able to find a
WD 3TB SATA 6GB/S drive to test with and everything was OK.  I also tested with
my latest driver.  I used a 2208 instead of a 2308.  Honestly, I'm not sure
offhand what the difference is, but I bet that doesn't matter.

The IOCStatus that is coming back is a 0x47, which is a SCSI Protocol Error. 
According to the MPI spec, this means "An unrecoverable protocol error has
caused termination of the SCSI I/O."  I'm not exactly sure why that would
happen, but it sounds like a disk problem.  The driver will try 5 times and
then fail.  You can see in the 9.2 driver, the same error happens but only
once.  The next retry works.

It seems to be a timing issue or something.  Maybe the enclosure is causing a
problem with this particular drive due to the time it takes to get ready or
spin up?  If that's the case we could try putting a delay between the retries
of a few seconds.  That shouldn't cause any problems since retries should be
rare.

Does this sound reasonable?  If so, can you make the change and test it or do
you need my help?

Stephen McConnell

-- 
You are receiving this mail because:
You are the assignee for the bug.