Return-Path: Received: from mail-ob0-f180.google.com ([209.85.214.180]:36555 "EHLO mail-ob0-f180.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752843AbbJOTt7 (ORCPT ); Thu, 15 Oct 2015 15:49:59 -0400 Received: by obbrx8 with SMTP id rx8so73893930obb.3 for ; Thu, 15 Oct 2015 12:49:59 -0700 (PDT) MIME-Version: 1.0 In-Reply-To: <20151015190945.GH49747@tonberry.usersys.redhat.com> References: <20151015190945.GH49747@tonberry.usersys.redhat.com> Date: Thu, 15 Oct 2015 15:49:58 -0400 Message-ID: Subject: Re: Diskless client with NFSv4 root? From: Trond Myklebust To: Scott Mayhew Cc: Linux NFS Mailing List Content-Type: text/plain; charset=UTF-8 Sender: linux-nfs-owner@vger.kernel.org List-ID: On Thu, Oct 15, 2015 at 3:09 PM, Scott Mayhew wrote: > > Is this something that is expected to work? It pretty much deadlocks > during boot. I took a vmcore... the oldest RPC task (and the one that > had the RPC transport locked) was a SETATTR that was waiting on an id > mapper upcall, and the first task in the sending RPC wait queue was the > OPEN of the /sbin/request-key program. > > If I disable id mapping on the NFS server then it looks like it works... > Yes, that is 100% expected behaviour. You get circular behaviour if you try to use idmapped identities when starting the idmapper. :-) In fact, if you look in the commit logs, you'll find that enabling rootless NFSv4 was one of the justifications for adding the AUTH_SYS non-idmapped mode on the client. Cheers Trond