From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753067Ab0LALYq (ORCPT ); Wed, 1 Dec 2010 06:24:46 -0500 Received: from h5.dl5rb.org.uk ([81.2.74.5]:37162 "EHLO h5.dl5rb.org.uk" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752193Ab0LALYp (ORCPT ); Wed, 1 Dec 2010 06:24:45 -0500 Date: Wed, 1 Dec 2010 11:24:10 +0000 From: Ralf Baechle To: Anoop P Cc: linux-mips@linux-mips.org, dvomlehn@cisco.com, David Howells , Thomas Gleixner , David Daney , linux-kernel@vger.kernel.org Subject: Re: [PATCH 3/3] Allow setup_irq call for VPE1 timer. Message-ID: <20101201112410.GL2916@linux-mips.org> References: <1290697632-6139-1-git-send-email-anoop.pa@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <1290697632-6139-1-git-send-email-anoop.pa@gmail.com> User-Agent: Mutt/1.5.21 (2010-09-15) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Nov 25, 2010 at 08:37:12PM +0530, Anoop P wrote: > From: Anoop P A > > VSMP configuration can have seperate timer interrupts for each VPE.Need to setup IRQ for VPE1 timer. > +#ifndef CONFIG_MIPS_MT_SMP > if (cp0_timer_irq_installed) > return 0; > - > +#endif > cp0_timer_irq_installed = 1; > > setup_irq(irq, &c0_compare_irqaction); On the stylistic side adding an #ifdef gives me wrinkles. With CONFIG_MIPS_MT_SMP this patch results in sharing c0_compare_irqaction between multiple interrupts which is broken. Struct irqaction contains the interrupt number, all registered irqaction structs are part of a chained list via its ->next member and also there is a per interrupt proc directory. To fix this properly you'll have to introduce do a bit of bookkeeping - you want to register each interrupt only once - and allocate a struct irqaction per registered timer interrupt. The allocation is made a little trickier by kmalloc not being available yet by the time this code is getting invoked via time_init() so you'll have to move it to run via the late_time_init hook like x86: static __init void x86_late_time_init(void) { ... do the real work ... } /* ... */ void __init time_init(void) { late_time_init = x86_late_time_init; } Which makes me wonder if there is a reason why we need to have both time_init() and late_time_init() - can't we just move the time_init()? Ralf