SORT BY:

LIST ORDER
THREAD
AUTHOR
SUBJECT


SEARCH

IPS HOME


    [Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

    iSCSI: questions on tags, unsolicited data



    Julian,
    
    Comments on iSCSI-02b draft.  Please comment on the first 
    section (non-editorial).  Thanks!
    
    o The discussion of target tags in the draft is inconsistent. 
      Section 2.6.3 states that it must not be 0xffffffff.  The payload
      diagram in 2.6 for WRITE Data PDU shows that the value of 0 (zero)
      is reserved since zero is used for unsolicited data.  Seems like
      both are reserved for unsolicited data.  Why? 
    
      Sections 2.15.2, 1.2.2.1, 2.6.3 must be fixed to reflect this 
      reserved value. (some sections state that there's no reserved value...)
    
    o Section 2.1.5 allows target suggesting a tag width.  Why is this
      allowed?  Is the intention to support targets with 16-bit processors?
      I would argue that this provides an undue design burden on initiator
      implementations to cater to both 32 and <32 cases.  My recommendation
      is to drop this support.
    
    o More clarifications on unsolicited data are in order - 
    	- Are multiple unsolicited Data PDUs legal (total data size
              below the limit)?
    	- Can unsolicited data be in the Command PDU (immediate)
              AND continue into a separate unsolicited Data PDU?
    
      If either of this is true, it appears that the "F - Final" bit 
      of the SCSI Write PDU should also be supported in the SCSI Command
      PDU.
    
    EDITORIAL:
    
    o Suggest adding a statement to the discussion of CmdRNs in 
      section 1.2.2.1 to the effect of: "There should only be one
      outstanding task with a given CmdRN (including 0) in a given
      session."
    
    o Section 1.2.2.1, para 7 has a sentence "Incoming data is always 
      implicitly solicited.".  Suggest adding the phrase to the sentence
      "by virtue of the SCSI READ command issued earlier".
    
    o Since urgent pointer usage is out of consideration now, seems like
      we can relax the requirement about MSb's in bytes 0 & 1 of every 
      payload - thus using the bits for other purposes like retry.
    
    o Section 2.11, para 1: "still active and all it's components" 
      should be "still active and all its components".
    
    o Section 2.6, there is no discussion of bits "O" and "U" in the
      SCSI READ Data PDU.  I assume they are overflow and underflow,
      this needs to be added.
    
    o The last para in section 2.11.4 is a repetition of last para in 
      section 2.11.  Suggest dropping the second.
    
    o NOP-In payload diagram has no "P-bit" (bit 6, byte 1).  A typo
      of "0" needs to changed to "P".
    
    o Fix section 2.12.1 title from "Target Task Tag" to "Target Tag".
    
    o Section 2.14, first sentence: Change "The logout is used.." to
      "The logout response is used...".
    
    o Section 2.17, para 3 refers to map commands.  This is an obsolete
      reference, and should be removed.
    
    o Section 7, first sentence: Change "There will be a well-known port..."
      to "There will be a well-known TCP port...".
    
    --
    Mallikarjun 
    
    
    Mallikarjun Chadalapaka
    M/S 5601			
    Networked Storage Architecture
    Network Storage Solutions Organization
    Hewlett-Packard, Roseville.
    cbm@rose.hp.com
    


Home

Last updated: Tue Sep 04 01:06:03 2001
6315 messages in chronological order