From: "J. Bruce Fields" Subject: Re: [PATCH] AFS: Implement file locking [try #2] Date: Wed, 30 May 2007 14:03:09 -0400 Message-ID: <20070530180309.GG7736@fieldses.org> References: <20070529215736.GO6815@fieldses.org> <20070529145426.19366.35289.stgit@warthog.cambridge.redhat.com> <32087.1180514132@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: akpm@osdl.org, linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org To: David Howells Return-path: Received: from mail.fieldses.org ([66.93.2.214]:40779 "EHLO fieldses.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753794AbXE3SDO (ORCPT ); Wed, 30 May 2007 14:03:14 -0400 Content-Disposition: inline In-Reply-To: <32087.1180514132@redhat.com> Sender: linux-fsdevel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Wed, May 30, 2007 at 09:35:32AM +0100, David Howells wrote: > J. Bruce Fields wrote: > > > --without having tried to understand how they're actually used, these > > data structures (like the pending_locks and granted_locks lists) seem to > > duplicate stuff that's already kept in fs/locks.c. Is there a reason > > they're required? > > Yes. I need to get the server lock first, before going to the VFS locking > routines. That doesn't really answer the question. The NFS client has similar requirements, but it doesn't have to duplicate the per-inode lists of granted locks, for example. > The VFS routines as they stand aren't particularly useful, and they > permit things that AFS doesn't support (such as lock upgrading). Sure. But handling those things is just a matter of checking for a few special cases and making sure they're rejected before calling the posix routines. --b.