Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S932611AbXBWTvY (ORCPT ); Fri, 23 Feb 2007 14:51:24 -0500 Received: (majordomo@vger.kernel.org) by vger.kernel.org id S932616AbXBWTvX (ORCPT ); Fri, 23 Feb 2007 14:51:23 -0500 Received: from e36.co.us.ibm.com ([32.97.110.154]:51426 "EHLO e36.co.us.ibm.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S932611AbXBWTvW (ORCPT ); Fri, 23 Feb 2007 14:51:22 -0500 Subject: Re: sparc generic time / clockevents From: john stultz To: David Miller Cc: tglx@linutronix.de, lkml , Peter Keilty In-Reply-To: <20070223.015520.125893182.davem@davemloft.net> References: <1172165959.25076.192.camel@localhost.localdomain> <20070223.012519.48528691.davem@davemloft.net> <1172224579.25076.219.camel@localhost.localdomain> <20070223.015520.125893182.davem@davemloft.net> Content-Type: text/plain Date: Fri, 23 Feb 2007 11:51:18 -0800 Message-Id: <1172260279.6261.20.camel@localhost> Mime-Version: 1.0 X-Mailer: Evolution 2.8.1 Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 5053 Lines: 174 On Fri, 2007-02-23 at 01:55 -0800, David Miller wrote: > From: Thomas Gleixner > Date: Fri, 23 Feb 2007 10:56:18 +0100 > > > On Fri, 2007-02-23 at 01:25 -0800, David Miller wrote: > > > > Yes, all you need is to omit the CLOCK_EVT_FEAT_PERIODIC flag when you > > > > register your device. > > > > > > Thanks a lot Thomas. > > > > > > I noticed while doing this work that the generic clock code is > > > incompatible with the time interpolator, since both provide a > > > do_{get,set}timeofday() implementation. Just FYI... > > > > John, can you have a look at this ? > > Note we may not care :-) All the interpolator does is provide > a generic tick based gettimeofday implementation, and the generic > clock code does essentially the same thing. > > And the generic clock code must in fact provide the implementations > in order to deal with dyntick issues. > > The only other platform using the time interpolator is IA64, > and after my dyntick implementation on sparc64 the one and > only user will be IA64 :-) Yea. I actually have some in-progress patches from Peter Keilty that convert ia64 and sparc64 time_interpolators to clocksources, then removes the time_interpolator code. The ia64 conversion is more complicated due to the fsyscall asm, but I think the sparc64 conversion (below) is pretty straight forward. I've only built tested this, so I have no clue if it actually works. Any thoughts? thanks -john From: Peter Keilty Initial sparc64 conversion to generic clocksource/timekeeping code. Signed-off-by: Peter Keilty Signed-off-by: John Stultz Kconfig | 2 +- defconfig | 2 +- kernel/time.c | 34 +++++++++++++++++++++++----------- 3 files changed, 25 insertions(+), 13 deletions(-) linux-2.6.21-rc1_timeofday-arch-sparc64_C7.patch ============================================ diff --git a/arch/sparc64/Kconfig b/arch/sparc64/Kconfig index f75a686..13a4547 100644 --- a/arch/sparc64/Kconfig +++ b/arch/sparc64/Kconfig @@ -34,7 +34,7 @@ config LOCKDEP_SUPPORT bool default y -config TIME_INTERPOLATION +config GENERIC_TIME bool default y diff --git a/arch/sparc64/defconfig b/arch/sparc64/defconfig index 0f44a6a..33e061a 100644 --- a/arch/sparc64/defconfig +++ b/arch/sparc64/defconfig @@ -9,7 +9,7 @@ CONFIG_64BIT=y CONFIG_MMU=y CONFIG_STACKTRACE_SUPPORT=y CONFIG_LOCKDEP_SUPPORT=y -CONFIG_TIME_INTERPOLATION=y +CONFIG_GENERIC_TIME=y CONFIG_ARCH_MAY_HAVE_PC_FDC=y # CONFIG_ARCH_HAS_ILOG2_U32 is not set # CONFIG_ARCH_HAS_ILOG2_U64 is not set diff --git a/arch/sparc64/kernel/time.c b/arch/sparc64/kernel/time.c index f84da4f..497b29b 100644 --- a/arch/sparc64/kernel/time.c +++ b/arch/sparc64/kernel/time.c @@ -31,6 +31,7 @@ #include #include #include #include +#include #include #include @@ -620,7 +621,7 @@ #endif if (!mregs && !dregs) { prom_printf("Something wrong, clock regs not mapped yet.\n"); prom_halt(); - } + } if (mregs) { spin_lock_irq(&mostek_lock); @@ -820,7 +821,7 @@ #endif } set_system_time(); - + local_irq_restore(flags); return 0; @@ -975,22 +976,33 @@ static struct notifier_block sparc64_cpu #endif /* CONFIG_CPU_FREQ */ -static struct time_interpolator sparc64_cpu_interpolator = { - .source = TIME_SOURCE_CPU, - .shift = 16, - .mask = 0xffffffffffffffffLL +static cycle_t read_sparc64_cpuclock(void) +{ + return (cycle_t)get_cycles(); +} + +static struct clocksource clocksource_sparc64_cpuclock = { + .name = "sparc64_cpuclock", + .rating = 300, + .read = read_sparc64_cpuclock, + .mask = 0xffffffffffffffffLL, + .mult = 0, /*to be caluclated*/ + .shift = 16, + .flags = CLOCK_SOURCE_IS_CONTINUOUS, }; + /* The quotient formula is taken from the IA64 port. */ #define SPARC64_NSEC_PER_CYC_SHIFT 10UL void __init time_init(void) { unsigned long clock = sparc64_init_timers(); - sparc64_cpu_interpolator.frequency = clock; - register_time_interpolator(&sparc64_cpu_interpolator); + clocksource_sparc64_cpuclock.mult = clocksource_hz2mult(clock, + clocksource_sparc64_cpuclock.shift); + clocksource_register(&clocksource_sparc64_cpuclock); - /* Now that the interpolator is registered, it is + /* Now that the clocksource is registered, it is * safe to start the timer ticking. */ sparc64_start_timers(); @@ -1025,11 +1037,11 @@ #endif unsigned long flags; u8 tmp; - /* + /* * Not having a register set can lead to trouble. * Also starfire doesn't have a tod clock. */ - if (!mregs && !dregs) + if (!mregs && !dregs) return -1; if (mregs) { - 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/