From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1755214Ab0FXLXv (ORCPT ); Thu, 24 Jun 2010 07:23:51 -0400 Received: from mx2.mail.elte.hu ([157.181.151.9]:47244 "EHLO mx2.mail.elte.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754968Ab0FXLXu (ORCPT ); Thu, 24 Jun 2010 07:23:50 -0400 Date: Thu, 24 Jun 2010 13:23:40 +0200 From: Ingo Molnar To: Peter Zijlstra Cc: Andi Kleen , Huang Ying , "H.PeterA" <"nvin hpa"@zytor.com>, linux-kernel@vger.kernel.org Subject: Re: [RFC][PATCH] irq_work Message-ID: <20100624112340.GA13502@elte.hu> References: <1277363069.3947.30.camel@yhuang-dev.sh.intel.com> <1277363993.1875.847.camel@laptop> <1277364453.3947.64.camel@yhuang-dev.sh.intel.com> <1277364727.1875.857.camel@laptop> <20100624102725.GA578@basil.fritz.box> <1277375409.1875.943.camel@laptop> <20100624105254.GB578@basil.fritz.box> <1277377121.1875.948.camel@laptop> <20100624110830.GC578@basil.fritz.box> <1277377852.1875.950.camel@laptop> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1277377852.1875.950.camel@laptop> User-Agent: Mutt/1.5.20 (2009-08-17) X-ELTE-SpamScore: 1.0 X-ELTE-SpamLevel: s X-ELTE-SpamCheck: no X-ELTE-SpamVersion: ELTE 2.0 X-ELTE-SpamCheck-Details: score=1.0 required=5.9 tests=BAYES_50 autolearn=no SpamAssassin version=3.2.5 1.0 BAYES_50 BODY: Bayesian spam probability is 40 to 60% [score: 0.4442] Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org * Peter Zijlstra wrote: > On Thu, 2010-06-24 at 13:08 +0200, Andi Kleen wrote: > > > And I really want hardirq context for perf callbacks, some code actually > > > relies on it (I used to have the fallback in the timer softirq and that > > > > Surely that could be fixed? *requiring* hard irq context sounds weird. > > possibly, but there is no reason what so ever to use softirq here. > > > > broke thing at some point). > > > > I have one case that needs to sleep (but only when interrupting user code) > > They key thing in it really is to switch stacks back to process. > > softirq can't sleep either, you need a trampoline anyway. What might make sense is to offer two types of callbacks: one that is immediate whenever an event triggers - and another that is sleepable and is executed from process context. Having an intermediate softirq level might be over-design indeed. Thanks, Ingo