Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1751670Ab2FKTl2 (ORCPT ); Mon, 11 Jun 2012 15:41:28 -0400 Received: from mail-qc0-f174.google.com ([209.85.216.174]:61743 "EHLO mail-qc0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751570Ab2FKTlY (ORCPT ); Mon, 11 Jun 2012 15:41:24 -0400 Date: Mon, 11 Jun 2012 22:41:17 +0300 From: Ido Yariv To: Ingo Molnar Cc: hpa@zytor.com, linux-kernel@vger.kernel.org, shai@scalemp.com, kiran.thirumalai@gmail.com, tglx@linutronix.de, linux-tip-commits@vger.kernel.org, Fengguang Wu Subject: Re: [tip:x86/platform] x86/vsmp: Ignore IOAPIC IRQ affinity if possible Message-ID: <20120611194117.GA19302@NoteStation.sbx07338.cambrma.wayport.net> References: <1338675095-27260-2-git-send-email-ido@wizery.com> <20120608085229.GA10638@gmail.com> <20120608154301.GA11242@NoteStation> <20120611085843.GL31556@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20120611085843.GL31556@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 Content-Length: 3777 Lines: 119 Hi Ingo, On Mon, Jun 11, 2012 at 10:58:43AM +0200, Ingo Molnar wrote: > > * Ido Yariv wrote: > > > This seems to break when CONFIG_PARAVIRT is set but CONFIG_SMP > > isn't. Since there's little point in optimizing IOAPIC routing > > for UP kernels, how about the following fix? > > Looks good to me, I'll give it some testing. I'm afraid this patch is missing one more config dependency. Since no_irq_affinity is declared in proc.c, we need to verify that CONFIG_PROC_FS is set before referencing this variable. The below patch fixes this. Thanks to Fengguang Wu for reporting this. Thanks, Ido. >From 2a865725447a84a7e3921b01b3683b665e896971 Mon Sep 17 00:00:00 2001 From: Ravikiran Thirumalai Date: Sun, 3 Jun 2012 01:11:35 +0300 Subject: [PATCH] x86/vsmp: Ignore IOAPIC IRQ affinity if possible vSMP can route interrupts more optimally based on internal knowledge the OS does not have. In order to support this optimization, all CPUs must be able to handle all possible IOAPIC interrupts. Fix this by setting the vector allocation domain for all CPUs and by enabling this feature in vSMP. Signed-off-by: Ravikiran Thirumalai Signed-off-by: Shai Fultheim [ido@wizery.com: rebased, simplified, and reworded the commit message] Signed-off-by: Ido Yariv --- arch/x86/kernel/vsmp_64.c | 30 ++++++++++++++++++++++++++---- 1 file changed, 26 insertions(+), 4 deletions(-) diff --git a/arch/x86/kernel/vsmp_64.c b/arch/x86/kernel/vsmp_64.c index 59eea85..3f0285a 100644 --- a/arch/x86/kernel/vsmp_64.c +++ b/arch/x86/kernel/vsmp_64.c @@ -16,6 +16,7 @@ #include #include #include +#include #include #include @@ -95,6 +96,18 @@ static void __init set_vsmp_pv_ops(void) ctl = readl(address + 4); printk(KERN_INFO "vSMP CTL: capabilities:0x%08x control:0x%08x\n", cap, ctl); + + /* If possible, let the vSMP foundation route the interrupt optimally */ +#ifdef CONFIG_SMP + if (cap & ctl & BIT(8)) { + ctl &= ~BIT(8); +#ifdef CONFIG_PROC_FS + /* Don't let users change irq affinity via procfs */ + no_irq_affinity = 1; +#endif + } +#endif + if (cap & ctl & (1 << 4)) { /* Setup irq ops and turn on vSMP IRQ fastpath handling */ pv_irq_ops.irq_disable = PV_CALLEE_SAVE(vsmp_irq_disable); @@ -102,12 +115,11 @@ static void __init set_vsmp_pv_ops(void) pv_irq_ops.save_fl = PV_CALLEE_SAVE(vsmp_save_fl); pv_irq_ops.restore_fl = PV_CALLEE_SAVE(vsmp_restore_fl); pv_init_ops.patch = vsmp_patch; - ctl &= ~(1 << 4); - writel(ctl, address + 4); - ctl = readl(address + 4); - printk(KERN_INFO "vSMP CTL: control set to:0x%08x\n", ctl); } + writel(ctl, address + 4); + ctl = readl(address + 4); + pr_info("vSMP CTL: control set to:0x%08x\n", ctl); early_iounmap(address, 8); } @@ -192,10 +204,20 @@ static int apicid_phys_pkg_id(int initial_apic_id, int index_msb) return hard_smp_processor_id() >> index_msb; } +/* + * In vSMP, all cpus should be capable of handling interrupts, regardless of + * the APIC used. + */ +static void fill_vector_allocation_domain(int cpu, struct cpumask *retmask) +{ + cpumask_setall(retmask); +} + static void vsmp_apic_post_init(void) { /* need to update phys_pkg_id */ apic->phys_pkg_id = apicid_phys_pkg_id; + apic->vector_allocation_domain = fill_vector_allocation_domain; } void __init vsmp_init(void) -- 1.7.10.2 -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/