seek obsoletable yet

Pat LaVarre p.lavarre at IEEE.org
Thu Oct 9 08:27:35 PDT 2003


* From the T10 Reflector (t10 at t10.org), posted by:
* Pat LaVarre <p.lavarre at ieee.org>
*
> Re: LOCK UNLOCK CACHE, PRE-FETCH, and SEEK commands
> * From the T10 Reflector (t10 at t10.org), posted by:
> * "Elliott, Robert (Server Storage)" <elliott at hp.com>
> *
> Does anyone use these block commands (on block devices)?
> 
...
> SEEK - advance notification that an LBA will be accessed
> 
...
> 
> c) obsolete SEEK (10).  An 8-byte LBA version was never proposed.
> MMC-4 has its own definition of SEEK (10) for the multimedia
> device type - SBC's definition isn't necessary for multimedia.
> SEEK(6) was defined in SCSI-2 but obsolete in [SCSI-3] SBC[-1].

Wish I knew if the binary code only of Win XP used seek ops.

I remember seeing some flavours of Microsoft Windows send op x2B SEEK
(10) to SBC devices as if they were MMC devices, choking over anything
except the MMC SEEK IMMEDIATE behaviour.

I remember the pain of the host choking if the device did not volunteer
to queue the following command.

I remember the pain of having the host respond errativally when the host
had happened to mix op x2B together with the ATA op x EF 95 etc. MSN
failure-by-design-when-write-protected.

I remember these supposedly immediate seeks appeared when the SBC device
was an ATAPI Slave to an ATA Master.  To my eye, Microsoft's design
intent appeared to be to kill the Slave bandwidth for the sake of the
Master.  I especially remember seeing, mixed into the protocol, a
snail-paced asynchronous timer tick poll i.e. about 0.000019 MHz.

> ... obsolete seek ...

I also know people now feeling the pain of the omission of an 8 byte lba
seek.

Given the t10 legacy of such patterns as op x 0A/ 2A/ AA variations on
parameter bit width in write and op x 08/ 28/ 28 variations on read, I
can easily imagine someone helpfully extending the op x0B/ 2B pattern of
variations on seek.

But I don't know enough people yet feeling this pain deeply enough for
me to have memorised the 64-bit lba opcodes, for example.  The limit of
32 bit lba = 2 TiB / 0.5 KiB/block is as yet too small for that.  More
ppeople I know are still working thru places where the 32-bit byte/disk
= 2 to 4 GiB/disk limit appears e.g. seemingly I had to write my own
pldd to serve as a dd for Windows up past 2 GiB.

Pat LaVarre


*
* For T10 Reflector information, send a message with
* 'info t10' (no quotes) in the message body to majordomo at t10.org




More information about the T10 mailing list