Number of ECC bytes

Robert Snively rsnively at brocade.com
Thu Sep 18 11:28:56 PDT 2008


Formatted message: <A HREF="r0809183_f.htm">HTML-formatted message</A>

In the general case, there is no longer any way to know for sure what
the structure of the 
correction code is on the actual disk or which part of that structure
will be 
revealed to you by a READ LONG.  Vendors now use a huge combination of 
tools to create reliable data recovery.
So you have two choices:
    a)	Contact the vendor and see if he will tell you how to parse the
	READ LONG returned data.  He probably will do so only under NDA.
    b)	Do a WRITE to a block.	Then do a READ LONG to the block and
	see what comes back.  Note that it may have no relationship to
the
	write if there are internal block correction codes being used.
	Then to test error recovery, do a WRITE LONG to the block using
	the data just read with one or more bits changed.  A subsequent
	READ to the block should provide information about the error
	and its recoverability, though it may show up only in a log
page.
If I were a disk drive vendor, I would be pushing for making READ/WRITE
LONG obsolete.
Of course for RAID devices, this question is even more (and less)
interesting.
Bob
________________________________
From: owner-t10 at t10.org [mailto:owner-t10 at t10.org] On Behalf Of
Bill.Martin at emulex.com
Sent: Thursday, September 18, 2008 10:10 AM
To: t10 at t10.org
Subject: Number of ECC bytes
I have looked and not found a mechanism for determining the number of
ECC bytes per logical block on an attached device.  Is there any place
in any mode page that this information is available, or any other
mechanism to determine this information?  I know that READ LONG allows
reading of this information, but it is vendor specific where this
information is, and there is no information in the return about what
portion of the additional data is ECC and what portion is WRProtect etc.
Thanks for any input.
Bill Martin
Emulex
Office of Technology
Industry Standards
916 772-3658
916 765-6875 (Cell)
bill.martin at emulex.com



More information about the T10 mailing list