Question: Deferred Sense reporting priority

Joseph C. Nemeth jnemeth at concentric.net
Tue May 5 14:32:05 PDT 1998


* From the T10 (formerly SCSI) Reflector (t10 at symbios.com), posted by:
* "Joseph C. Nemeth" <jnemeth at concentric.net>
*

-----Original Message-----
From: Rodney Van Meter <Rodney.Vanmeter at qntm.com>
To: 'Jeff Stai' <stai at Brocade.COM>; 'Joseph C. Nemeth'
<jnemeth at concentric.net>; t10 at Symbios.COM <t10 at Symbios.COM>
Date: Monday, May 04, 1998 4:22 PM
Subject: RE: Question: Deferred Sense reporting priority


> Okay, I'll follow up with a related question.  Forgive me, too,
>if I'm rusty on target implementations, it's been several years since I
>did one.
>
> How many Unit Attention conditions should the target keep?
>


Rodney, this whole issue of trying to prioritize and age these errors
through a single sense data block for the LUN has always made my head hurt
so much that I've thrown it out. I keep ALL the UA conditions for ALL the
initiators, for as long as it takes them to show up and claim the UA
condition.

This isn't as expensive as it sounds: I need one bit for each UA condition
on each initiator. For a private loop, that's 126 bits, times the number of
UA conditions I want to keep: PON, media removed, mode page changed, etc.

Part of my command pre-check simply tests the bit for each UA condition (in
some order) for that initiator. If the bit is set, I *generate* the UA
condition in the sense data right on the spot, then clear the bit (UA
condition cleared). The bit-mask sits around forever, so any new initiator
will see every UA condition that has ever accumulated for the device.

This obviously doesn't scale well for fabrics, with 2^24 (give or take)
potential initiators, which is why I will eventually need to enumerate the
*actual* initiators that log in, and build my bit mask from there.

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





More information about the T10 mailing list