From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id ; Wed, 7 Aug 2002 23:23:03 -0400 Received: (majordomo@vger.kernel.org) by vger.kernel.org id ; Wed, 7 Aug 2002 23:22:55 -0400 Received: from samba.sourceforge.net ([198.186.203.85]:16363 "EHLO lists.samba.org") by vger.kernel.org with ESMTP id ; Wed, 7 Aug 2002 23:22:53 -0400 From: Rusty Russell To: "David S. Miller" Cc: rusty@rustcorp.com.au, george@mvista.com, kuznet@ms2.inr.ac.ru, linux-kernel@vger.kernel.org Subject: Re: softirq parameters In-reply-to: Your message of "Wed, 07 Aug 2002 11:24:24 MST." <20020807.112424.60322440.davem@redhat.com> Date: Thu, 08 Aug 2002 12:29:30 +1000 Message-Id: <20020808032826.AAE5F4A86@lists.samba.org> Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org In message <20020807.112424.60322440.davem@redhat.com> you write: > From: Matthew Wilcox > Date: Wed, 7 Aug 2002 19:35:04 +0100 > > Hrm, I didn't realise what the implementation of percpu was... does it > work for modules? > > Good question, I bet it doesn't based upon how I understand > it to work. Yes, it doesn't work for modules. The IBM guys have a kmalloc_percpu thing which will, but I'm not trying to change the world. This is one reason for changing to DECLARE_PER_CPU and DEFINE_PER_CPU macros (the other is because RTH wants to use __thread for per-cpu variables). Rusty. -- Anyone who quotes me in their sig is an idiot. -- Rusty Russell.