[nfsv4] CLIENTID race condition during server GRACE.

New Message Reply About this list Date view Thread view Subject view Author view Attachment view

From: Stevan Steve Allen (scallen@us.ibm.com)
Date: 01/17/05-03:02:50 PM Z


Message-ID: <OF361981E1.41899D63-ON88256F8C.00714B10-88256F8C.00739DBF@us.ibm.com>
From: Stevan Steve Allen <scallen@us.ibm.com>
Date: Mon, 17 Jan 2005 13:02:50 -0800
Subject: [nfsv4] CLIENTID race condition during server GRACE.








Comments?

RFC section 8.1.1 discusses ways to generate a unique CLIENT ID.
SETCLIENTID discusses the use of credentials to detect a conflicting CLIENT
ID when a client lease is active.  If the NFS server restarts, the
credential relationship is lost.  The resulting server GRACE period looks
to provide a race condition between clients using the same CLIENT ID.

For added data integrity, we were discussing adding additional checks in
the server to limit a SETCLIENTID into the previously active client.  i.e.
harden domain names of stateful hosts and have SETCLIENTID compare HOST
names during grace.  Return NFSERR_CLID_INUSE for host name mismatch.  This
does not address a CLIENT ID race between clients on the same host (which
seems to be a buggy client).


Thanks,
Stevan C. Allen

z/NFS Development: 5600 Cottle Rd. San Jose CA. 95193


Add the following important questions when deciding between designs:
   1. Which design can be clearly documented and understood?
   2. Which design would you prefer to debug?




C5194180.jpg

_______________________________________________
nfsv4 mailing list
nfsv4@ietf.org
https://www1.ietf.org/mailman/listinfo/nfsv4


New Message Reply About this list Date view Thread view Subject view Author view Attachment view

This archive was generated by hypermail 2.1.2 : 03/04/05-02:13:51 AM Z CST