From mboxrd@z Thu Jan 1 00:00:00 1970 From: "Aneesh Kumar K. V" Subject: Re: [PATCH -V14 04/11] vfs: Allow handle based open on symlinks Date: Wed, 07 Jul 2010 23:23:31 +0530 Message-ID: <874ogbfaas.fsf@linux.vnet.ibm.com> References: <1276621981-2774-1-git-send-email-aneesh.kumar@linux.vnet.ibm.com> <1276621981-2774-5-git-send-email-aneesh.kumar@linux.vnet.ibm.com> <20100707152303.GS11732@laptop> <87d3uzfeej.fsf@linux.vnet.ibm.com> <20100707165719.GC9263@laptop> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: hch@infradead.org, viro@zeniv.linux.org.uk, adilger@sun.com, corbet@lwn.net, serue@us.ibm.com, neilb@suse.de, hooanon05@yahoo.co.jp, bfields@fieldses.org, linux-fsdevel@vger.kernel.org, sfrench@us.ibm.com, philippe.deniel@CEA.FR, linux-kernel@vger.kernel.org To: Nick Piggin Return-path: In-Reply-To: <20100707165719.GC9263@laptop> Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Thu, 8 Jul 2010 02:57:19 +1000, Nick Piggin wrote: > On Wed, Jul 07, 2010 at 09:54:52PM +0530, Aneesh Kumar K. V wrote: > > On Thu, 8 Jul 2010 01:23:03 +1000, Nick Piggin wrote: > > > On Tue, Jun 15, 2010 at 10:42:54PM +0530, Aneesh Kumar K.V wrote: > > > > The patch update may_open to allow handle based open on symlinks. > > > > The file handle based API use file descritor returned from open_by_handle_at > > > > to do different file system operations. To find the link target name we > > > > need to get a file descriptor on symlinks. > > > > > > This is a pretty big change, isn't it? > > > > Yes > > > > > Have you looked through vfs to > > > ensure this is actually OK? I was just looking at remount,ro code, for > > > example, and it seems to assume only writable open files on ISREG > > > files. > > > > I verified that write and read returns an error on that descriptor. I > > will look at rest of code to ensure the it doesn't break assumptions in > > the rest of VFS. > > > > > > > > Is this restricted to RDONLY? Really, it should be O_NONE... > > > > > > > > > > we need the interface so that we can do fstat(lstat) and readlink using the > > handle. If we are against the idea of allowing an fd on symlink, how about > > adding stat_by_handle and readlink_by_handle syscall ? That way we can > > drop "vfs: Support null pathname in readlink" patch > > It would be nice to avoid introducing open files to symlinks, but > I don't know about really the *right* thing to do though. I think > avoiding adding any by-handle syscalls except for open by handle > may be a good idea... OTOH if this is really a special case for > handles... > > I haven't followed this thread closely, so can you just go back > and explain to me why you need this? Ie. that you can't achieve with > a path-to-handle follow/nofollow option. We need to be able to read the link target using file handle. The exact usecase i have is with respect to implementing READLINK operation on a userspace NFS server. The request contain the file handle and the response include target name. Similarly we need to able to get file attribute based on file handle. I was using readlinkat and fstat to obtain both the above information using the descriptor returned by open_by_handle on a file handle representing symlink -aneesh