From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757628AbZKSKMl (ORCPT ); Thu, 19 Nov 2009 05:12:41 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S1755348AbZKSKMk (ORCPT ); Thu, 19 Nov 2009 05:12:40 -0500 Received: from earthlight.etchedpixels.co.uk ([81.2.110.250]:41745 "EHLO www.etchedpixels.co.uk" rhost-flags-OK-OK-OK-FAIL) by vger.kernel.org with ESMTP id S1755337AbZKSKMj (ORCPT ); Thu, 19 Nov 2009 05:12:39 -0500 Date: Thu, 19 Nov 2009 10:13:30 +0000 From: Alan Cox To: Jamie Lokier Cc: Oliver Neukum , Jan Blunck , linux-fsdevel@vger.kernel.org, Linux-Kernel Mailinglist , Andrew Morton , jkacur@redhat.com, Thomas Gleixner , Arnd Bergmann , Christoph Hellwig , =?ISO-8859-14?B?RnLpZOlyaWM=?= Weisbecker , Alexander Viro Subject: Re: [PATCH 1/2] BKL: Remove BKL from default_llseek() Message-ID: <20091119101330.5a09d6b0@lxorguk.ukuu.org.uk> In-Reply-To: <20091119024008.GB15128@shareable.org> References: <1258560457-15129-1-git-send-email-jblunck@suse.de> <20091118171524.4d2f8cec@lxorguk.ukuu.org.uk> <20091118172730.GD28723@shareable.org> <200911181835.55007.oliver@neukum.org> <20091118175003.GF28723@shareable.org> <20091118181606.1c8caab1@lxorguk.ukuu.org.uk> <20091119024008.GB15128@shareable.org> X-Mailer: Claws Mail 3.7.3 (GTK+ 2.14.7; x86_64-redhat-linux-gnu) Mime-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org > That was implicit in my question: Are there any seekable devices where > bytes read != offset delta, and if yes, is that correct behaviour, a > bug, or a silly interface that should go away? It's neither a bug nor silly, although some of our users of it are a bit strange. Consider a block based device. It makes complete sense then that a short read (eg user space passing a small buffer) causes the seek position to move on one block. Ditto a block based system with EOF markers might do that (eg some tape systems). In the weirdness (but API) category we have things like the MSR driver which returns an entire MSR for each byte offset. Alan