FCP-4:

Banther, Michael michael.banther at hp.com
Thu Aug 31 02:42:44 PDT 2006


Formatted message: <A HREF="r0608311_f.htm">HTML-formatted message</A>
Attachment #1: <A HREF="r0608311_banther_michael.vcf">banther_michael.vcf</A>

Hi Dave,
HP tape drives do not support retrying a sequence of writes from
Relative Offset=0.  We expect the Relative Offset to match what we
reported in a preceding REC.  In the REC we report the Relative Offset
of the previous XFER_RDY.
Regards,
Michael Banther
Hewlett-Packard Ltd.
+44 117 312 9503
________________________________
	From: David Peterson 
	Sent: Friday, August 25, 2006 1:38 PM
	To: 'BANTHER,MICHAEL (HP-UnitedKingdom,ex2)'; 'Erich Oetting';
'Greg Wheeless'; 'kdbutt at us.ibm.com'; 'paul.entzel at quantum.com';
'paul.suhler at quantum.com'
	Subject: FCP-4: 
	Howdy,
	I took an action item to contact the tape vendors regarding the
Relative Offset value on a retry. From the minutes:
	6.1 Relative Offset on retry Peterson/McDATA
	In the event of Sequence retry, is the initiator allowed to
restart from Relative Offset=0, which forces the target to hold the
application client buffer until confirmed completion (may be megabytes)?
	Tape vendors say this is ok for reads, but a problem for writes.
They are able to back up only to the last XFER_RDY point. This seems to
work with all HBAs in use. It is requested to remove the allowance to go
back to zero.
	ACTION: HBA vendors to determine whether they need to be able to
do Sequence retry from relative offset 0.
	060711-1 Dave Peterson to contact tape vendors to see whether
they are able to force the Sequence Count to zero on Sequence retries.
	So what say ye? Note this "feature" has been in place since
FCP-2.
	The text in question:
	12.4.1.7 FCP_DATA IU recovery - read operations
	The initiator FCP_Port shall set the RELATIVE OFFSET field in
the SRR to that of the next data requested. If the initiator FCP_Port is
unable to determine the Relative Offset of the next data requested, the
initiator FCP_Port shall set the RELATIVE OFFSET field to zero.
	...Dave
	(no disclaimer)



More information about the T10 mailing list