Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1758226Ab3EGMPH (ORCPT ); Tue, 7 May 2013 08:15:07 -0400 Received: from smtp.citrix.com ([66.165.176.89]:32181 "EHLO SMTP.CITRIX.COM" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1756681Ab3EGMPF (ORCPT ); Tue, 7 May 2013 08:15:05 -0400 X-IronPort-AV: E=Sophos;i="4.87,628,1363132800"; d="scan'208";a="23709250" Date: Tue, 7 May 2013 13:15:02 +0100 From: Stefano Stabellini X-X-Sender: sstabellini@kaball.uk.xensource.com To: Ian Campbell CC: Stefano Stabellini , "xen-devel@lists.xensource.com" , "linux@arm.linux.org.uk" , "arnd@arndb.de" , "marc.zyngier@arm.com" , "konrad.wilk@oracle.com" , "will.deacon@arm.com" , "linux-kernel@vger.kernel.org" , "nico@linaro.org" , "olof@lixom.net" , "linux-arm-kernel@lists.infradead.org" , "cov@codeaurora.org" Subject: Re: [Xen-devel] [PATCH RFC v2 2/4] arm: introduce CONFIG_PARAVIRT and pv_time_ops In-Reply-To: <1367918609.26321.43.camel@zakaz.uk.xensource.com> Message-ID: References: <1367851878-21629-2-git-send-email-stefano.stabellini@eu.citrix.com> <1367918609.26321.43.camel@zakaz.uk.xensource.com> User-Agent: Alpine 2.02 (DEB 1266 2009-07-14) MIME-Version: 1.0 Content-Type: text/plain; charset="US-ASCII" Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 6109 Lines: 181 On Tue, 7 May 2013, Ian Campbell wrote: > On Mon, 2013-05-06 at 15:51 +0100, Stefano Stabellini wrote: > > Introduce CONFIG_PARAVIRT on ARM. > > What about PARAVIRT_TIME_ACCOUNTING? I'm not sure what it is but it > looks like a more lightweight version of pv stolen time? PARAVIRT_TIME_ACCOUNTING selects PARAVIRT on x86 :-) > > The only paravirt interface supported is pv_time_ops.steal_clock. > > No runtime pvops patching yet. > > Or indeed ever, I think. The use cases for patching on x86 are not > things which carry over to ARM with virt extensions. Agreed > > This allows us to make us of steal_account_process_tick for stolen ticks > > accounting. > > > > > > Signed-off-by: Stefano Stabellini > > CC: linux@arm.linux.org.uk > > CC: will.deacon@arm.com > > CC: nico@linaro.org > > CC: marc.zyngier@arm.com > > CC: cov@codeaurora.org > > CC: arnd@arndb.de > > CC: olof@lixom.net > > --- > > arch/arm/Kconfig | 9 +++++++++ > > arch/arm/include/asm/paravirt.h | 19 +++++++++++++++++++ > > arch/arm/kernel/Makefile | 1 + > > arch/arm/kernel/paravirt.c | 32 ++++++++++++++++++++++++++++++++ > > 4 files changed, 61 insertions(+), 0 deletions(-) > > create mode 100644 arch/arm/include/asm/paravirt.h > > create mode 100644 arch/arm/kernel/paravirt.c > > > > diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig > > index 344e299..35cb10a 100644 > > --- a/arch/arm/Kconfig > > +++ b/arch/arm/Kconfig > > @@ -1887,12 +1887,21 @@ config XEN_DOM0 > > def_bool y > > depends on XEN > > > > +config PARAVIRT > > + bool "Enable paravirtualization code" > > + ---help--- > > + This changes the kernel so it can modify itself when it is run > > + under a hypervisor, potentially improving performance significantly > > + over full virtualization. However, when run without a hypervisor > > + the kernel is theoretically slower and slightly larger. > > I'm not sure this description (carried over from x86) are really true > for ARM. e.g. the downsides there when not virtualised are in the PV MMU > (pte operations) and interrupt masking stuff, which should never make > its way onto ARM. Right > I think it would be a worthwhile change to refactor the stolen time > handling out from under the rather wide reaching umbrella of the x86 > PARAVIRT option. (assuming PARAVIRT_TIME_ACCOUNTING isn't already that) Actually PARAVIRT doesn't mean much in common code, the only thing it covers is stolen time. What I mean to say is that just because we are introducing something called "PARAVIRT" on ARM, it doesn't mean that it has to come with all sort of baggage. > > + > > config XEN > > bool "Xen guest support on ARM (EXPERIMENTAL)" > > depends on ARM && AEABI && OF > > depends on CPU_V7 && !CPU_V6 > > depends on !GENERIC_ATOMIC64 > > select ARM_PSCI > > + select PARAVIRT > > help > > Say Y if you want to run Linux in a Virtual Machine on Xen on ARM. > > > > diff --git a/arch/arm/include/asm/paravirt.h b/arch/arm/include/asm/paravirt.h > > new file mode 100644 > > index 0000000..3b95bc6 > > --- /dev/null > > +++ b/arch/arm/include/asm/paravirt.h > > @@ -0,0 +1,19 @@ > > +#ifndef _ASM_ARM_PARAVIRT_H > > +#define _ASM_ARM_PARAVIRT_H > > + > > +struct static_key; > > +extern struct static_key paravirt_steal_enabled; > > +extern struct static_key paravirt_steal_rq_enabled; > > + > > +struct pv_time_ops { > > + unsigned long long (*steal_clock)(int cpu); > > +}; > > +extern struct pv_time_ops pv_time_ops; > > + > > +static inline u64 paravirt_steal_clock(int cpu) > > +{ > > + return pv_time_ops.steal_clock(cpu); > > +} > > + > > + > > +#endif > > diff --git a/arch/arm/kernel/Makefile b/arch/arm/kernel/Makefile > > index dd9d90a..6764f60 100644 > > --- a/arch/arm/kernel/Makefile > > +++ b/arch/arm/kernel/Makefile > > @@ -86,5 +86,6 @@ ifeq ($(CONFIG_ARM_PSCI),y) > > obj-y += psci.o > > obj-$(CONFIG_SMP) += psci_smp.o > > endif > > +obj-$(CONFIG_PARAVIRT) += paravirt.o > > > > extra-y := $(head-y) vmlinux.lds > > diff --git a/arch/arm/kernel/paravirt.c b/arch/arm/kernel/paravirt.c > > new file mode 100644 > > index 0000000..3e73fc8 > > --- /dev/null > > +++ b/arch/arm/kernel/paravirt.c > > @@ -0,0 +1,32 @@ > > +/* > > + * This program is free software; you can redistribute it and/or modify > > + * it under the terms of the GNU General Public License version 2 as > > + * published by the Free Software Foundation. > > + * > > + * This program is distributed in the hope that it will be useful, > > + * but WITHOUT ANY WARRANTY; without even the implied warranty of > > + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the > > + * GNU General Public License for more details. > > + * > > + * Copyright (C) 2013 Citrix Systems > > + * > > + * Author: Stefano Stabellini > > + */ > > + > > +#include > > +#include > > +#include > > +#include > > + > > +struct static_key paravirt_steal_enabled; > > +struct static_key paravirt_steal_rq_enabled; > > + > > +static u64 native_steal_clock(int cpu) > > +{ > > + return 0; > > +} > > + > > +struct pv_time_ops pv_time_ops = { > > + .steal_clock = native_steal_clock, > > +}; > > +EXPORT_SYMBOL_GPL(pv_time_ops); > > This foo_ops.bar and native_bar thing is a bit of a hangover from the > paravirt patching infrastructure on x86 and it doesn't really apply > here. > > Given that the call to paravirt_steal_time call is already protected by > this static_key stuff I think it would be safe to leave the hook as NULL > in the case where it is unused. Good point > Given all the different clock sources on ARM is there not an existing > ops struct where this could live? I am not sure, but I would be happy to move it something arch-specific. -- 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/