[T10] delayed review comment on 18-117r2 (Configure port mode)
Ballard, Curtis C (HPE Storage)
curtis.ballard at hpe.com
Thu Jan 17 12:19:13 PST 2019
Thanks for the feedback Gerry. The text you reference got added live during the meeting in the middle of another discussion and I meant to tag that to go back later but missed getting that tag on it.
I will submit a new proposal for corrections to 18-117.
Regarding the behavior if SP=0, I’ll have to check with my team. There may be use cases where the phy configuration changing on a reset makes sense. There are other mode parameters where having a different saved value become active after a reset creates complications and typically the model has been that the host is expected to save the setting in that case so as you suggest we might take the ‘hosts shouldn’t do stupid things’ approach.
Hewlett Packard Enterprise
HPE Storage R&D
Fort Collins, CO
From: t10-bounces at t10.org [mailto:t10-bounces at t10.org] On Behalf Of Gerry Houlder
Sent: Thursday, January 17, 2019 9:03 AM
To: T10 Reflector <t10 at t10.org>
Subject: [T10] delayed review comment on 18-117r2 (Configure port mode)
I have observed an issue with this wording in the proposal:
For a MODE SELECT command with the PC field set to 00b (i.e, current values), if any bit in the port mode bitmap in the parameter data is set to one and is different than the current value for that bit, then upon successful completion...
First, the PC field only applies to a MODE SENSE command not MODE SELECT. This is a serious mis-statement.
Second, If a MODE SELECT command only changes the current value of a mode page bit (i.e., because SP=0), then the saved value is likely to be different. Then when a reset occurs, the current value is updated with the saved value. This will cause the port configuration to revert to the previous phy configuration. I suspect this would be bad.
The best way for an initiator to make this change is with a MODE SELECT command with SP field set to one. i request that this besubmitted in a new proposal.
This forces the current and saved value of the mode page bits to be the same. I don't know if it is reasonable to require a target to reject a MODE SELECT command that changes any of the port mode bitmap bits if SP=0 -- I wonder if we can specify that any MODE SELECT command that changes port mode bitmap shall update both current and saved bits even if SP=0. Alternatively, we can just let the phys fall where they may if the initiator does something this stupid.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the T10