[SRP] March 1 Meeting Notes
cris.simpson at intel.com
Tue Mar 5 12:22:27 PST 2002
* From the T10 Reflector (t10 at t10.org), posted by:
* "Simpson, Cris" <cris.simpson at intel.com>
Chuck Gibson Adaptec
Bob Snively Brocade
Brian Forbes Brocade
Dave Peterson Cisco
Rob Elliott Compaq
Bob Griswold Crossroads
Bob Nixon Emulex
Ralph Weber ENDL
Santash Rao HP
Randy Haagens HP
Mallikarjun Chadalapaka HP
Giles Frazier IBM
George Penokie IBM
Cris Simpson Intel
Nathan Obr MSFT
Rob Haydt MSFT
Ed Gardner Ophidian
Gerald Maurer QLogic
Roger Cummings Veritas
BobS: Reservations are a poor tech for networked devices, why not obsolete?
CrisS: Anyone want to speak up for classic reservations?
DaveP: Time better spent on PRs...
RogerC: Let's obsolete for NEW stds, let legacy keep.
RobH: We don't have any problems with obsoleting classic reservations.
RobE: What about Mode Pages, Log Pages?
GOP: You should save them anyway.
Keep MPs and LPs separate from intconn.
(FC clears when last init logs out.)
<AR> DaveP - Proposal for obsoleting classic reservations
FCP clears MPs only when all sessions drop
GOP - don't even do it then
When we re-login, we assume that no state is valid, reset everything -
easier than looking.
GOP: Reasonable implementation - but std shouldn't require
RobE: Let's expand to Log Pages. LPs cleared on (some) resets.
Does FC Logout clear LPs?
RobE: There are other events that FC says cause clearing
Will send out list (Sent March 1)
RandyH: The real issue is the lifetime of an I_T(_L) nexus.
BobS: I've got SCSI-2 chapter and verse (r10L 3.1.28) defining duration of
ROW: Only meaningful for pSCSI
BobS: See also SAM-2 (r22 4.10, 3.1.68) crossref I_T nexus object
only instantiated (derived) in the MIB
RESOLVED: Classic reservations to be obsoleted.
DaveP will get on CAP agenda
MPs and LPs to wait for CAP.
BobS: New concept: "SAN domain"
Randy: Brings time into scope - if you can't talk to a device,
why keep the state? Sometimes, tossing state is a good thing to
EdG: Maybe we should prohibit mode pages per intiator
RandyH: Can support MPs/init through virtual targets
RobE: Port Control MP is per-init
RandyH: In networked world, targets should discard state at every
1. CPQ015 : MSB/LSB for Tag
Ed: What harm is there in having MSB/LSB?
GOP: For consistency, we need it, like other single-field values.
Ed: Need it for analyzers. (Argument carries - keep it)
2. Edit005 :
SPC-3 says "These [alias] associations shall be cleared under any event
that resets the logical unit and events designated by the SCSI protocol."
It appears that we need to have a list or a statement that there are no
such events. Where would it go, what would it say?
(RobE will include in list of issues for resolution at CAP)
3. intel041: may/shall send I_LOGOUT
Change to "SHOULD send"
4. intel042: should/shall send T_LOGOUT
Stay w/ SHOULD.
5. intel044: LOGIN_REJ reason code
Add new code:"RDMA Channel limit reached for this initiator see (5.1.3)"
6. intel084: replace 'count' with 'PMDL Length'
Ed liked it that way first, but wants keep it stable.
See if cleaner.
7. intel094: optional IUs?
All are Mandatory.
8. intel096: T_LOGOUT reason code for bad length?
Add new code
March 08 0900-1100 PST
1.888.316.5901 or +1.617.801.9781
Cris Simpson 503.712.4333
Intel EPG/ACD Architecture Hillsboro, OR
PGP Fingerprint:0DA0 418E A27B 0B76 5F02 3DD4 0546 6D13 F88A 1E60
* 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