From mboxrd@z Thu Jan 1 00:00:00 1970 From: Trond Myklebust Subject: Re: bug in nfs in 2.6.18-rc5? Date: Thu, 31 Aug 2006 12:43:12 -0400 Message-ID: <1157042592.11347.70.camel@localhost> References: <44F6F80F.1000202@cs.columbia.edu> <1157040230.11347.31.camel@localhost> <44F70D22.2030703@cs.columbia.edu> Mime-Version: 1.0 Content-Type: text/plain Content-Transfer-Encoding: 7bit Cc: linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org, unionfs@fsl.cs.sunysb.edu Return-path: Received: from pat.uio.no ([129.240.10.4]:43449 "EHLO pat.uio.no") by vger.kernel.org with ESMTP id S932195AbWHaQnY (ORCPT ); Thu, 31 Aug 2006 12:43:24 -0400 To: Shaya Potter In-Reply-To: <44F70D22.2030703@cs.columbia.edu> Sender: linux-fsdevel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Thu, 2006-08-31 at 12:24 -0400, Shaya Potter wrote: > why does the client care about server mounted file systems? It wants to allow POSIX applications to work correctly even in the case where the nfsd administrator is using 'nohide'. It wants those same applications to work correctly in the case where the nfsd administrator is exporting more than one filesystem over NFSv4. > The > server's nfsd has to tell them apart, otherwise shouldn't give them to > the client. Otherwise it seems like the nfsd and the nfs client have to > have innate knowledge of each other. Of course the server knows that it is crossing a mountpoint. The client figures it out by looking at the 'fsid' attribute (which uniquely labels the filesystem on that server) in order to figure out which filesystem that the file/directory it just looked up belongs to. Whenever the fileid changes between parent directory and child, that means that a mountpoint was crossed on the server. Trond