From: Ingo Molnar Subject: Re: [patch] rt: fix ipi kfree(), introduce IPI_SOFTIRQ Date: Thu, 12 Feb 2009 09:22:07 +0100 Message-ID: <20090212082207.GB26838@elte.hu> References: <20090212005032.GA4788@nowhere> <20090212021257.GB4697@nowhere> <20090212081801.GA22979@elte.hu> <20090212081923.GA26838@elte.hu> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Thomas Gleixner , LKML , rt-users , Steven Rostedt , Peter Zijlstra , Carsten Emde , Clark Williams To: Frederic Weisbecker , Peter Zijlstra Return-path: Received: from mx2.mail.elte.hu ([157.181.151.9]:55680 "EHLO mx2.mail.elte.hu" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750879AbZBLIWZ (ORCPT ); Thu, 12 Feb 2009 03:22:25 -0500 Content-Disposition: inline In-Reply-To: <20090212081923.GA26838@elte.hu> Sender: linux-rt-users-owner@vger.kernel.org List-ID: * Ingo Molnar wrote: > ------------------> > Subject: rt: fix ipi kfree(), introduce IPI_SOFTIRQ > From: Ingo Molnar > Date: Thu Feb 12 09:06:11 CET 2009 > > in 2.6.28 generic_smp_call_function_interrupt() grew a kfree(), > which is a rather complex, sleepable method under -rt. But the > IPI code runs as a hardirq - which cannot run such code. > > So defer this work to a softirq context instead. It still stays > on the same CPU so the percpu IPI assumptions are upheld. On a second thought ... I think we could eliminate the kfree() instead, and keep the atomicity of IPI cross-calls. Linus expressed doubts about the IPI kmalloc()/kfree() pair we do in the generic SMP IPI code, suggesting that it probably does not help performance all that much - so such a change might be upstream-able as well and would keep -rt closer to mainline. Ingo