Discussion about 11-376 -- add extension for defect descriptors

Jon Haswell - SISA jon.haswell at sisa.samsung.com
Tue Sep 20 17:59:45 PDT 2011


Formatted message: <a href="http://www.t10.org/cgi-bin/ac.pl?t=r&f=r1109202_f.htm">HTML-formatted message</a>

Speaking as an SSD supplier ...
We do not actually use the vendor unique field but map our defects into
the existing formats such as Byte from Index. We do this at our
customers request as their systems expect to be able to read back or
send defect lists in these formats across HDDs and SSDs.
Our mapping into a format such as Bytes From Index is not truly
meaningful if you try and interpret it, but we and our customers
effectively just use it as an opaque container with other higher level
software that is able to interpret it used during failure analysis or
the like.
Personally I don't see a need/desire for new formats for SSDs. I believe
every SSD will effectively use a proprietary format for reporting
defects and it would be very difficult and of little value to try and
standardize it.
Thanks
Jon Haswell
Dir SSD Development
Office	408 544 5869
Cell	 408 472 2495
From: owner-t10 at t10.org [mailto:owner-t10 at t10.org] On Behalf Of Gerry
Houlder
Sent: Tuesday, September 20, 2011 2:41 PM
To: T10 Reflector
Subject: Discussion about 11-376 -- add extension for defect descriptors
Hi everyone,
During discussion of 11-376r0 at the September CAP meeting, I was given
direction to use two new address descriptor values for the new Bytes
>From Index and Cylinder Head Physical Block formats. If I do this, this
will consume two of the three reserved codes.
Before I do this, I'd like to point out that the currently defined
address formats are not very applicable to Solid State Drives. I suspect
these drives use the vendor specific option today, if they support any
of the options at all. Is there any thought that there needs to be
standard defect descriptor formats that are designed for SSD use? If so,
then it might not be a good idea to consume two of the three remaining
defect list formats for HDD use.
This direction was instigated by the head field not being able to give
up one of its bits for the MADS function. If the bit was carved from the
most significant bit of the four byte BYTES FROM INDEX or SECTOR NUMBER
fields, would this make it more acceptable to extend the current address
descriptor formats instead of creating new ones?
I'd like to see more discussion on whether new address formats are
needed for SSD or other new technology products before committing to
using up two more address descriptor formats. Note that we are
constrained by the 3 bit field in the FORMAT command to a maximum of 8
address formats. However, I suppose this could be expanded if SBC-3
reclaimed the two obsolete bytes in the FORMAT CDB, the 111b code could
be a trigger to get the actual address descriptor format from of of
these bytes ...



More information about the T10 mailing list