Re: [nfsv4] CREATE_SESSION backchannel question

"J. Bruce Fields" <> Thu, 14 October 2010 16:24 UTC

Return-Path: <>
Received: from localhost (localhost []) by (Postfix) with ESMTP id D0FBD3A6A8B for <>; Thu, 14 Oct 2010 09:24:43 -0700 (PDT)
X-Virus-Scanned: amavisd-new at
X-Spam-Flag: NO
X-Spam-Score: -2.581
X-Spam-Status: No, score=-2.581 tagged_above=-999 required=5 tests=[AWL=0.018, BAYES_00=-2.599]
Received: from ([]) by localhost ( []) (amavisd-new, port 10024) with ESMTP id TN38JM+rn2II for <>; Thu, 14 Oct 2010 09:24:43 -0700 (PDT)
Received: from ( []) by (Postfix) with ESMTP id 07A833A6A6E for <>; Thu, 14 Oct 2010 09:24:43 -0700 (PDT)
Received: from bfields by with local (Exim 4.71) (envelope-from <>) id 1P6Qcz-0008Gd-UX; Thu, 14 Oct 2010 12:26:02 -0400
Date: Thu, 14 Oct 2010 12:26:01 -0400
To: "Matt W. Benjamin" <>
Message-ID: <>
References: <> <>
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Disposition: inline
In-Reply-To: <>
User-Agent: Mutt/1.5.20 (2009-06-14)
From: "J. Bruce Fields" <>
Cc: nfsv4 <>
Subject: Re: [nfsv4] CREATE_SESSION backchannel question
X-Mailman-Version: 2.1.9
Precedence: list
List-Id: NFSv4 Working Group <>
List-Unsubscribe: <>, <>
List-Archive: <>
List-Post: <>
List-Help: <>
List-Subscribe: <>, <>
X-List-Received-Date: Thu, 14 Oct 2010 16:24:43 -0000

On Thu, Oct 14, 2010 at 11:07:33AM -0400, Matt W. Benjamin wrote:
> I wondered a bit about the server's backchannel preferences in
> CREATE_SESSION, for example, if the server implementation would prefer
> (or even require) a dedicated backchannel.  The server can accept or
> reject a client request to use the current connection as a forechannel
> and backchannel.

I'd never actually noticed that was allowed!

> If the server clears the backchannel flag, or the
> client did not set it, can the server expect that the client will
> offer a backchannel in a subsequent BIND_CONN_TO_SESSION (under its
> obligations under 2.9.2, item 3).

I don't think the existing Linux client is prepared to handle that; more
likely telling it that the backchannel wasn't set up will cause it to
destroy the client and try to create a new one, resulting in an infinite
loop....  (But I haven't tested that, much less other clients.)

It seems like mildly antisocial behavior on the part of the server, but
I can believe it would be convenient to not have to handle the
shared-connection case, and probably wouldn't be difficult to handle on
the client.