From df211c8a478e16232ebbc16e650b19163183fcb6 Mon Sep 17 00:00:00 2001 From: Nathan Lynch Date: Wed, 23 May 2007 10:51:25 +1000 Subject: [POWERPC] Remove spinlock from struct cpu_purr_data cpu_purr_data is a per-cpu array used to account for stolen time on partitioned systems. It used to be the case that cpus accessed each others' cpu_purr_data, so each entry was protected by a spinlock. However, the code was reworked ("Simplify stolen time calculation") with the result that each cpu accesses its own cpu_purr_data and not those of other cpus. This means we can get rid of the spinlock as long as we're careful to disable interrupts when accessing cpu_purr_data in process context. Signed-off-by: Nathan Lynch Signed-off-by: Paul Mackerras --- arch/powerpc/kernel/time.c | 24 +++++++++++++++--------- 1 file changed, 15 insertions(+), 9 deletions(-) (limited to 'arch/powerpc/kernel/time.c') diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c index 2c8564d54e4..0a8a820672f 100644 --- a/arch/powerpc/kernel/time.c +++ b/arch/powerpc/kernel/time.c @@ -222,19 +222,28 @@ struct cpu_purr_data { int initialized; /* thread is running */ u64 tb; /* last TB value read */ u64 purr; /* last PURR value read */ - spinlock_t lock; }; +/* + * Each entry in the cpu_purr_data array is manipulated only by its + * "owner" cpu -- usually in the timer interrupt but also occasionally + * in process context for cpu online. As long as cpus do not touch + * each others' cpu_purr_data, disabling local interrupts is + * sufficient to serialize accesses. + */ static DEFINE_PER_CPU(struct cpu_purr_data, cpu_purr_data); static void snapshot_tb_and_purr(void *data) { + unsigned long flags; struct cpu_purr_data *p = &__get_cpu_var(cpu_purr_data); + local_irq_save(flags); p->tb = mftb(); p->purr = mfspr(SPRN_PURR); wmb(); p->initialized = 1; + local_irq_restore(flags); } /* @@ -242,15 +251,14 @@ static void snapshot_tb_and_purr(void *data) */ void snapshot_timebases(void) { - int cpu; - if (!cpu_has_feature(CPU_FTR_PURR)) return; - for_each_possible_cpu(cpu) - spin_lock_init(&per_cpu(cpu_purr_data, cpu).lock); on_each_cpu(snapshot_tb_and_purr, NULL, 0, 1); } +/* + * Must be called with interrupts disabled. + */ void calculate_steal_time(void) { u64 tb, purr; @@ -262,7 +270,6 @@ void calculate_steal_time(void) pme = &per_cpu(cpu_purr_data, smp_processor_id()); if (!pme->initialized) return; /* this can happen in early boot */ - spin_lock(&pme->lock); tb = mftb(); purr = mfspr(SPRN_PURR); stolen = (tb - pme->tb) - (purr - pme->purr); @@ -270,7 +277,6 @@ void calculate_steal_time(void) account_steal_time(current, stolen); pme->tb = tb; pme->purr = purr; - spin_unlock(&pme->lock); } /* @@ -284,12 +290,12 @@ static void snapshot_purr(void) if (!cpu_has_feature(CPU_FTR_PURR)) return; + local_irq_save(flags); pme = &per_cpu(cpu_purr_data, smp_processor_id()); - spin_lock_irqsave(&pme->lock, flags); pme->tb = mftb(); pme->purr = mfspr(SPRN_PURR); pme->initialized = 1; - spin_unlock_irqrestore(&pme->lock, flags); + local_irq_restore(flags); } #endif /* CONFIG_PPC_SPLPAR */ -- cgit v1.2.3-70-g09d2 From 4cefebb1b497a84d61f2fa29d497df75a84b69d4 Mon Sep 17 00:00:00 2001 From: Michael Neuling Date: Fri, 8 Jun 2007 13:18:50 +1000 Subject: [POWERPC] Fix stolen time for SMT without LPAR For POWERPC, stolen time accounts for cycles lost to the hypervisor or PURR cycles attributed to the other SMT thread. Hence, when a PURR is available, we should still calculate stolen time, irrespective of being virtualised. Signed-off-by: Michael Neuling Signed-off-by: Paul Mackerras --- arch/powerpc/kernel/time.c | 2 +- include/asm-powerpc/time.h | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-) (limited to 'arch/powerpc/kernel/time.c') diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c index 0a8a820672f..43c687a1d76 100644 --- a/arch/powerpc/kernel/time.c +++ b/arch/powerpc/kernel/time.c @@ -214,7 +214,6 @@ static void account_process_time(struct pt_regs *regs) run_posix_cpu_timers(current); } -#ifdef CONFIG_PPC_SPLPAR /* * Stuff for accounting stolen time. */ @@ -279,6 +278,7 @@ void calculate_steal_time(void) pme->purr = purr; } +#ifdef CONFIG_PPC_SPLPAR /* * Must be called before the cpu is added to the online map when * a cpu is being brought up at runtime. diff --git a/include/asm-powerpc/time.h b/include/asm-powerpc/time.h index 3fd57c048f5..2d00e13c981 100644 --- a/include/asm-powerpc/time.h +++ b/include/asm-powerpc/time.h @@ -232,7 +232,7 @@ extern void account_process_vtime(struct task_struct *tsk); #define account_process_vtime(tsk) do { } while (0) #endif -#if defined(CONFIG_VIRT_CPU_ACCOUNTING) && defined(CONFIG_PPC_SPLPAR) +#if defined(CONFIG_VIRT_CPU_ACCOUNTING) extern void calculate_steal_time(void); extern void snapshot_timebases(void); #else -- cgit v1.2.3-70-g09d2 From 71712b455374a73af042fcfb5002fef5fd25ba44 Mon Sep 17 00:00:00 2001 From: Tony Breeds Date: Fri, 22 Jun 2007 16:54:30 +1000 Subject: [POWERPC] Move iSeries_tb_recal into its own late_initcall. Currently iSeries will recalibrate the cputime_factors in the first settimeofday() call. It seems the reason for doing this is to ensure a resaonable time delta after time_init(). On current kernels (with udev), this call is made 40-60 seconds into the boot process, by moving it to a late initcall it is called approximately 5 seconds after time_init() is called. This is sufficient to recalibrate the timebase. Signed-off-by: Tony Breeds CC: Stephen Rothwell Signed-off-by: Paul Mackerras --- arch/powerpc/kernel/time.c | 30 +++++++++++++++++++----------- arch/powerpc/platforms/iseries/setup.c | 6 ++---- include/asm-powerpc/time.h | 2 ++ 3 files changed, 23 insertions(+), 15 deletions(-) (limited to 'arch/powerpc/kernel/time.c') diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c index 43c687a1d76..66d2db7495a 100644 --- a/arch/powerpc/kernel/time.c +++ b/arch/powerpc/kernel/time.c @@ -77,9 +77,8 @@ /* keep track of when we need to update the rtc */ time_t last_rtc_update; #ifdef CONFIG_PPC_ISERIES -unsigned long iSeries_recal_titan = 0; -unsigned long iSeries_recal_tb = 0; -static unsigned long first_settimeofday = 1; +static unsigned long __initdata iSeries_recal_titan; +static signed long __initdata iSeries_recal_tb; #endif /* The decrementer counts down by 128 every 128ns on a 601. */ @@ -556,10 +555,15 @@ EXPORT_SYMBOL(profile_pc); * returned by the service processor for the timebase frequency. */ -static void iSeries_tb_recal(void) +static int __init iSeries_tb_recal(void) { struct div_result divres; unsigned long titan, tb; + + /* Make sure we only run on iSeries */ + if (!firmware_has_feature(FW_FEATURE_ISERIES)) + return -ENODEV; + tb = get_tb(); titan = HvCallXm_loadTod(); if ( iSeries_recal_titan ) { @@ -600,8 +604,18 @@ static void iSeries_tb_recal(void) } iSeries_recal_titan = titan; iSeries_recal_tb = tb; + + return 0; } -#endif +late_initcall(iSeries_tb_recal); + +/* Called from platform early init */ +void __init iSeries_time_init_early(void) +{ + iSeries_recal_tb = get_tb(); + iSeries_recal_titan = HvCallXm_loadTod(); +} +#endif /* CONFIG_PPC_ISERIES */ /* * For iSeries shared processors, we have to let the hypervisor @@ -765,12 +779,6 @@ int do_settimeofday(struct timespec *tv) * to the RTC again, or write to the RTC but then they don't call * settimeofday to perform this operation. */ -#ifdef CONFIG_PPC_ISERIES - if (firmware_has_feature(FW_FEATURE_ISERIES) && first_settimeofday) { - iSeries_tb_recal(); - first_settimeofday = 0; - } -#endif /* Make userspace gettimeofday spin until we're done. */ ++vdso_data->tb_update_count; diff --git a/arch/powerpc/platforms/iseries/setup.c b/arch/powerpc/platforms/iseries/setup.c index 7f5dcee814d..13a8b1908de 100644 --- a/arch/powerpc/platforms/iseries/setup.c +++ b/arch/powerpc/platforms/iseries/setup.c @@ -79,8 +79,6 @@ extern void iSeries_pci_final_fixup(void); static void iSeries_pci_final_fixup(void) { } #endif -extern unsigned long iSeries_recal_tb; -extern unsigned long iSeries_recal_titan; struct MemoryBlock { unsigned long absStart; @@ -292,8 +290,8 @@ static void __init iSeries_init_early(void) { DBG(" -> iSeries_init_early()\n"); - iSeries_recal_tb = get_tb(); - iSeries_recal_titan = HvCallXm_loadTod(); + /* Snapshot the timebase, for use in later recalibration */ + iSeries_time_init_early(); /* * Initialize the DMA/TCE management diff --git a/include/asm-powerpc/time.h b/include/asm-powerpc/time.h index 2d00e13c981..d7f5ddfbaac 100644 --- a/include/asm-powerpc/time.h +++ b/include/asm-powerpc/time.h @@ -240,5 +240,7 @@ extern void snapshot_timebases(void); #define snapshot_timebases() do { } while (0) #endif +extern void iSeries_time_init_early(void); + #endif /* __KERNEL__ */ #endif /* __POWERPC_TIME_H */ -- cgit v1.2.3-70-g09d2 From fc9069fe56d9594b7a6da0b9f9ca0eafd756260f Mon Sep 17 00:00:00 2001 From: Tony Breeds Date: Wed, 4 Jul 2007 14:04:31 +1000 Subject: [POWERPC] Modify sched_clock() to make CONFIG_PRINTK_TIME more sane When booting a current kernel with CONFIG_PRINTK_TIME enabled you'll see messages like: [ 0.000000] time_init: decrementer frequency = 188.044000 MHz [ 0.000000] time_init: processor frequency = 1504.352000 MHz [3712914.436297] Console: colour dummy device 80x25 This cause by the initialisation of tb_to_ns_scale in time_init(), suddenly the multiplication in sched_clock() now does something :). This patch modifies sched_clock() to report the offset since the machine booted so the same printk's now look like: [ 0.000000] time_init: decrementer frequency = 188.044000 MHz [ 0.000000] time_init: processor frequency = 1504.352000 MHz [ 0.000135] Console: colour dummy device 80x25 Effectivly including the uptime in printk()s. This patch makes tb_to_ns_scale and tb_to_ns_shift static and read_mostly for good measure. Signed-off-by: Tony Breeds Signed-off-by: Paul Mackerras --- arch/powerpc/kernel/time.c | 9 ++++++--- 1 file changed, 6 insertions(+), 3 deletions(-) (limited to 'arch/powerpc/kernel/time.c') diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c index 66d2db7495a..e5df167f782 100644 --- a/arch/powerpc/kernel/time.c +++ b/arch/powerpc/kernel/time.c @@ -112,8 +112,9 @@ u64 ticklen_to_xs; /* 0.64 fraction */ DEFINE_SPINLOCK(rtc_lock); EXPORT_SYMBOL_GPL(rtc_lock); -u64 tb_to_ns_scale; -unsigned tb_to_ns_shift; +static u64 tb_to_ns_scale __read_mostly; +static unsigned tb_to_ns_shift __read_mostly; +static unsigned long boot_tb __read_mostly; struct gettimeofday_struct do_gtod; @@ -755,7 +756,7 @@ unsigned long long sched_clock(void) { if (__USE_RTC()) return get_rtc(); - return mulhdu(get_tb(), tb_to_ns_scale) << tb_to_ns_shift; + return mulhdu(get_tb() - boot_tb, tb_to_ns_scale) << tb_to_ns_shift; } int do_settimeofday(struct timespec *tv) @@ -974,6 +975,8 @@ void __init time_init(void) } tb_to_ns_scale = scale; tb_to_ns_shift = shift; + /* Save the current timebase to pretty up CONFIG_PRINTK_TIME */ + boot_tb = get_tb(); tm = get_boot_time(); -- cgit v1.2.3-70-g09d2