using Request Sense

Pat LaVarre p.lavarre at IEEE.org
Sat Jun 28 12:38:04 PDT 2008


* From the T10 Reflector (t10 at t10.org), posted by:
* Pat LaVarre <p.lavarre at ieee.org>
*
Four thoughts ...
1.
>> Since some protocols use autosense then the sense data will be  
>> cleared when a command creates the condition and request sense  
>> will just get back "no sense". Is there a case where request sense  
>> is still useful?
>
> protocols that do not have autosense ...
> progress indication ...
> unit attention interlock ...
Also I remember seeing unsolicited REQUEST SENSE read-and-clear  
queued unit attentions without triggering a CHECK CONDITION.
I don't immediately remember how the standards speak to that behavior.
2.
> unit attention interlock ...
> iSCSI ... REQUEST SENSE command clears a unit attention interlock  
> the same way a CLEAR ACA task management function clears an ACA  
> condition.
Ah. The CBI transport of SCSI thru full-speed USB to FDD worked/works  
that way. Ditto one of the never-published formally-obsolete simple- 
queueing feature bits of the BBB transport of high/ full/ super speed  
SCSI thru USB. Read-and-clear is evil, aye.
3.
>> progress indication ...
>
>>
> From: http://www.t10.org/ftp/t10/document.08/08-263r0.pdf
> ...
>>
> a suitably inclined device server *may* report multiple progress  
> counters in response to a single REQUEST SENSE command that is  
> sagacious enough to have the DESC bit set to one in the CDB ...
Possibly we have consensus that autosensed Test Unit Ready and	
explicit unsolicited Request Sense can each be found to work in  
places where the other does not.
Quite apart from the device, hosts so commonly manage the Data In  
memory of Request Sense differently than the Sense In memory of Test  
Unit Ready that any number of complexities in the host can favour the  
one status-gathering technique or the other.
4.
>> Since some protocols use autosense then the sense data will be  
>> cleared when a command creates the condition and request sense  
>> will just get back "no sense". Is there a case where request sense  
>> is still useful?
>>
I see that question asked and answered at:
""" 5.2 Important commands for all SCSI device servers
""" 5.2.6 Using the REQUEST SENSE command
The REQUEST SENSE command ... may be used by an application client to  
poll the status of some background operations and to clear  
interlocked unit attention conditions ...
""" http://t10.org/ftp/t10/drafts/spc4/spc4r15.pdf
I'm not yet persuaded that text should change.
Happy Saturday,
*
* 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