From: Trond Myklebust Subject: Re: regression: am-utils stopped working in 2.6.25-rc* Date: Tue, 01 Apr 2008 18:14:23 -0400 Message-ID: <1207088063.13443.22.camel@heimdal.trondhjem.org> References: <20080401210742.GE21343@fieldses.org> Mime-Version: 1.0 Content-Type: text/plain Cc: Meelis Roos , Linux Kernel list , linux-nfs@vger.kernel.org To: "J. Bruce Fields" Return-path: Received: from pat.uio.no ([129.240.10.15]:52432 "EHLO pat.uio.no" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754497AbYDAWO3 (ORCPT ); Tue, 1 Apr 2008 18:14:29 -0400 In-Reply-To: <20080401210742.GE21343@fieldses.org> Sender: linux-nfs-owner@vger.kernel.org List-ID: On Tue, 2008-04-01 at 17:07 -0400, J. Bruce Fields wrote: > On Fri, Mar 28, 2008 at 03:20:18PM +0200, Meelis Roos wrote: > > Current 2.6.25-rc* kernels have broken am-utils (NFS-based automounter) > > support. No error messages, just "no such file or directory" when trying > > to enter /net/anotherhostname. Just tried 2.6.24 and that worked fine. > > am-utils is the latest in Debian unstable, 6.1.5-9. > > > > Earliest 2.6.25-rc currently tested was 2.6.25-rc3+git as of 20080304. > > This version reports > > Invalid hostname "pid3477@koiott:/net" in NFS lock request > > into dmesg and am-utils does not work. > > > > 2.6.25-rc4+git as of 20080310 and later ones do not show this message > > but am-utils does still not work. > > > > Will try bisecting it further. > > Have you figured out anything more? Can strace show exactly which > system call is failing? See the bugzilla report: http://bugzilla.kernel.org/show_bug.cgi?id=10349 I'm still waiting to hear back from Meelis, but to me this looks like an am-utils bug. AFAICS, am-utils advertises a binary NFS mount structure version of '6', to the mount() syscall, but internally they only initialise the equivalent of a version '4' structure. When http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git&a=commitdiff&h=f9c3a3802119a2d30f3e4a69aef30a81e09d0209 was merged, then this failure to initialise the selinux 'context' field causes mount() to return an EINVAL. Cheers Trond