From mboxrd@z Thu Jan 1 00:00:00 1970 From: "J. Bruce Fields" Subject: Re: [PATCH] NFS: Restore missing hunk in NFS mount option parser Date: Fri, 5 Sep 2008 17:34:25 -0400 Message-ID: <20080905213425.GH12947@fieldses.org> References: <20080905181339.17103.36231.stgit@ingres.1015granger.net> <20080905211110.GE12947@fieldses.org> <76bd70e30809051425y21d1ce1kb06c8d72a83bdb4e@mail.gmail.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: neilb@suse.de, linux-nfs@vger.kernel.org, Trond Myklebust To: chucklever@gmail.com Return-path: Received: from mail.fieldses.org ([66.93.2.214]:59171 "EHLO fieldses.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750955AbYIEVe3 (ORCPT ); Fri, 5 Sep 2008 17:34:29 -0400 In-Reply-To: <76bd70e30809051425y21d1ce1kb06c8d72a83bdb4e-JsoAwUIsXosN+BqQ9rBEUg@public.gmane.org> Sender: linux-nfs-owner@vger.kernel.org List-ID: On Fri, Sep 05, 2008 at 05:25:31PM -0400, Chuck Lever wrote: > On Fri, Sep 5, 2008 at 5:11 PM, J. Bruce Fields wrote: > > On Fri, Sep 05, 2008 at 02:16:07PM -0400, Chuck Lever wrote: > >> Commit f45663ce5fb30f76a3414ab3ac69f4dd320e760a was missing a hunk that > >> prevented the new "sloppy" mount option from having any effect. > > > > I don't think that on its own would justify sending it in for 2.6.27.[1] > > The original patch for 27 was supposed to fix a regression (ie > automounter stopped working in heterogenous environments). This patch > does fix the full regression. > > There is already logic in nfs-utils-1.1.3 that maps the "-s" option > (which has been around for EVAR) to "-o sloppy". This logic is > enabled for 2.6.27 kernels and later. So this does need to go in 27. > > Would it help if I rewrote the description? Could be. Assume I'm tired and stupid.... > >> Tested against 2.6.27-rc. 2.6.26 is not affected. > > > > But if I understand right, the effect of leaving out this chunk was to > > make the *default* behavior "sloppy"? Which seems a drastic change from > > the previous behavior. And it's a simple enough patch. > > No, the default behavior is as before. The behavior without this > patch is that the kernel recognizes "sloppy" but it doesn't do > anything about it. That can't be right: > >> + if (errors > 0) { > >> + dfprintk(MOUNT, "NFS: parsing encountered %d error%s\n", > >> + errors, (errors == 1 ? "" : "s")); > >> + if (!sloppy) > >> + return 0; > >> + } > >> return 1; Ignoring the printk, the *only* change in behavior here happens when sloppy is *not* set. Right? --b. (not a big fan of not's)