1 files changed, 274 insertions, 17 deletions

diff --git a/arch/powerpc/kernel/time.c b/arch/powerpc/kernel/time.c
index 2a7ddc57937..24e3ad756de 100644
--- a/arch/powerpc/kernel/time.c
+++ b/arch/powerpc/kernel/time.c
@@ -51,6 +51,7 @@
 #include <linux/percpu.h>
 #include <linux/rtc.h>
 #include <linux/jiffies.h>
+#include <linux/posix-timers.h>
 
 #include <asm/io.h>
 #include <asm/processor.h>
@@ -98,6 +99,7 @@ unsigned long tb_ticks_per_jiffy;
 unsigned long tb_ticks_per_usec = 100; /* sane default */
 EXPORT_SYMBOL(tb_ticks_per_usec);
 unsigned long tb_ticks_per_sec;
+EXPORT_SYMBOL(tb_ticks_per_sec);	/* for cputime_t conversions */
 u64 tb_to_xs;
 unsigned tb_to_us;
 
@@ -135,6 +137,224 @@ unsigned long tb_last_stamp;
  */
 DEFINE_PER_CPU(unsigned long, last_jiffy);
 
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+/*
+ * Factors for converting from cputime_t (timebase ticks) to
+ * jiffies, milliseconds, seconds, and clock_t (1/USER_HZ seconds).
+ * These are all stored as 0.64 fixed-point binary fractions.
+ */
+u64 __cputime_jiffies_factor;
+EXPORT_SYMBOL(__cputime_jiffies_factor);
+u64 __cputime_msec_factor;
+EXPORT_SYMBOL(__cputime_msec_factor);
+u64 __cputime_sec_factor;
+EXPORT_SYMBOL(__cputime_sec_factor);
+u64 __cputime_clockt_factor;
+EXPORT_SYMBOL(__cputime_clockt_factor);
+
+static void calc_cputime_factors(void)
+{
+	struct div_result res;
+
+	div128_by_32(HZ, 0, tb_ticks_per_sec, &res);
+	__cputime_jiffies_factor = res.result_low;
+	div128_by_32(1000, 0, tb_ticks_per_sec, &res);
+	__cputime_msec_factor = res.result_low;
+	div128_by_32(1, 0, tb_ticks_per_sec, &res);
+	__cputime_sec_factor = res.result_low;
+	div128_by_32(USER_HZ, 0, tb_ticks_per_sec, &res);
+	__cputime_clockt_factor = res.result_low;
+}
+
+/*
+ * Read the PURR on systems that have it, otherwise the timebase.
+ */
+static u64 read_purr(void)
+{
+	if (cpu_has_feature(CPU_FTR_PURR))
+		return mfspr(SPRN_PURR);
+	return mftb();
+}
+
+/*
+ * Account time for a transition between system, hard irq
+ * or soft irq state.
+ */
+void account_system_vtime(struct task_struct *tsk)
+{
+	u64 now, delta;
+	unsigned long flags;
+
+	local_irq_save(flags);
+	now = read_purr();
+	delta = now - get_paca()->startpurr;
+	get_paca()->startpurr = now;
+	if (!in_interrupt()) {
+		delta += get_paca()->system_time;
+		get_paca()->system_time = 0;
+	}
+	account_system_time(tsk, 0, delta);
+	local_irq_restore(flags);
+}
+
+/*
+ * Transfer the user and system times accumulated in the paca
+ * by the exception entry and exit code to the generic process
+ * user and system time records.
+ * Must be called with interrupts disabled.
+ */
+void account_process_vtime(struct task_struct *tsk)
+{
+	cputime_t utime;
+
+	utime = get_paca()->user_time;
+	get_paca()->user_time = 0;
+	account_user_time(tsk, utime);
+}
+
+static void account_process_time(struct pt_regs *regs)
+{
+	int cpu = smp_processor_id();
+
+	account_process_vtime(current);
+	run_local_timers();
+	if (rcu_pending(cpu))
+		rcu_check_callbacks(cpu, user_mode(regs));
+	scheduler_tick();
+ 	run_posix_cpu_timers(current);
+}
+
+#ifdef CONFIG_PPC_SPLPAR
+/*
+ * Stuff for accounting stolen time.
+ */
+struct cpu_purr_data {
+	int	initialized;			/* thread is running */
+	u64	tb0;			/* timebase at origin time */
+	u64	purr0;			/* PURR at origin time */
+	u64	tb;			/* last TB value read */
+	u64	purr;			/* last PURR value read */
+	u64	stolen;			/* stolen time so far */
+	spinlock_t lock;
+};
+
+static DEFINE_PER_CPU(struct cpu_purr_data, cpu_purr_data);
+
+static void snapshot_tb_and_purr(void *data)
+{
+	struct cpu_purr_data *p = &__get_cpu_var(cpu_purr_data);
+
+	p->tb0 = mftb();
+	p->purr0 = mfspr(SPRN_PURR);
+	p->tb = p->tb0;
+	p->purr = 0;
+	wmb();
+	p->initialized = 1;
+}
+
+/*
+ * Called during boot when all cpus have come up.
+ */
+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);
+}
+
+void calculate_steal_time(void)
+{
+	u64 tb, purr, t0;
+	s64 stolen;
+	struct cpu_purr_data *p0, *pme, *phim;
+	int cpu;
+
+	if (!cpu_has_feature(CPU_FTR_PURR))
+		return;
+	cpu = smp_processor_id();
+	pme = &per_cpu(cpu_purr_data, cpu);
+	if (!pme->initialized)
+		return;		/* this can happen in early boot */
+	p0 = &per_cpu(cpu_purr_data, cpu & ~1);
+	phim = &per_cpu(cpu_purr_data, cpu ^ 1);
+	spin_lock(&p0->lock);
+	tb = mftb();
+	purr = mfspr(SPRN_PURR) - pme->purr0;
+	if (!phim->initialized || !cpu_online(cpu ^ 1)) {
+		stolen = (tb - pme->tb) - (purr - pme->purr);
+	} else {
+		t0 = pme->tb0;
+		if (phim->tb0 < t0)
+			t0 = phim->tb0;
+		stolen = phim->tb - t0 - phim->purr - purr - p0->stolen;
+	}
+	if (stolen > 0) {
+		account_steal_time(current, stolen);
+		p0->stolen += stolen;
+	}
+	pme->tb = tb;
+	pme->purr = purr;
+	spin_unlock(&p0->lock);
+}
+
+/*
+ * Must be called before the cpu is added to the online map when
+ * a cpu is being brought up at runtime.
+ */
+static void snapshot_purr(void)
+{
+	int cpu;
+	u64 purr;
+	struct cpu_purr_data *p0, *pme, *phim;
+	unsigned long flags;
+
+	if (!cpu_has_feature(CPU_FTR_PURR))
+		return;
+	cpu = smp_processor_id();
+	pme = &per_cpu(cpu_purr_data, cpu);
+	p0 = &per_cpu(cpu_purr_data, cpu & ~1);
+	phim = &per_cpu(cpu_purr_data, cpu ^ 1);
+	spin_lock_irqsave(&p0->lock, flags);
+	pme->tb = pme->tb0 = mftb();
+	purr = mfspr(SPRN_PURR);
+	if (!phim->initialized) {
+		pme->purr = 0;
+		pme->purr0 = purr;
+	} else {
+		/* set p->purr and p->purr0 for no change in p0->stolen */
+		pme->purr = phim->tb - phim->tb0 - phim->purr - p0->stolen;
+		pme->purr0 = purr - pme->purr;
+	}
+	pme->initialized = 1;
+	spin_unlock_irqrestore(&p0->lock, flags);
+}
+
+#endif /* CONFIG_PPC_SPLPAR */
+
+#else /* ! CONFIG_VIRT_CPU_ACCOUNTING */
+#define calc_cputime_factors()
+#define account_process_time(regs)	update_process_times(user_mode(regs))
+#define calculate_steal_time()		do { } while (0)
+#endif
+
+#if !(defined(CONFIG_VIRT_CPU_ACCOUNTING) && defined(CONFIG_PPC_SPLPAR))
+#define snapshot_purr()			do { } while (0)
+#endif
+
+/*
+ * Called when a cpu comes up after the system has finished booting,
+ * i.e. as a result of a hotplug cpu action.
+ */
+void snapshot_timebase(void)
+{
+	__get_cpu_var(last_jiffy) = get_tb();
+	snapshot_purr();
+}
+
 void __delay(unsigned long loops)
 {
 	unsigned long start;
@@ -283,9 +503,9 @@ static inline void update_gtod(u64 new_tb_stamp, u64 new_stamp_xsec,
 	 * the two values of tb_update_count match and are even then the
 	 * tb_to_xs and stamp_xsec values are consistent.  If not, then it
 	 * loops back and reads them again until this criteria is met.
+	 * We expect the caller to have done the first increment of
+	 * vdso_data->tb_update_count already.
 	 */
-	++(vdso_data->tb_update_count);
-	smp_wmb();
 	vdso_data->tb_orig_stamp = new_tb_stamp;
 	vdso_data->stamp_xsec = new_stamp_xsec;
 	vdso_data->tb_to_xs = new_tb_to_xs;
@@ -310,20 +530,15 @@ static __inline__ void timer_recalc_offset(u64 cur_tb)
 	unsigned long offset;
 	u64 new_stamp_xsec;
 	u64 tlen, t2x;
+	u64 tb, xsec_old, xsec_new;
+	struct gettimeofday_vars *varp;
 
 	if (__USE_RTC())
 		return;
 	tlen = current_tick_length();
 	offset = cur_tb - do_gtod.varp->tb_orig_stamp;
-	if (tlen == last_tick_len && offset < 0x80000000u) {
-		/* check that we're still in sync; if not, resync */
-		struct timeval tv;
-		__do_gettimeofday(&tv, cur_tb);
-		if (tv.tv_sec <= xtime.tv_sec &&
-		    (tv.tv_sec < xtime.tv_sec ||
-		     tv.tv_usec * 1000 <= xtime.tv_nsec))
-			return;
-	}
+	if (tlen == last_tick_len && offset < 0x80000000u)
+		return;
 	if (tlen != last_tick_len) {
 		t2x = mulhdu(tlen << TICKLEN_SHIFT, ticklen_to_xs);
 		last_tick_len = tlen;
@@ -332,6 +547,21 @@ static __inline__ void timer_recalc_offset(u64 cur_tb)
 	new_stamp_xsec = (u64) xtime.tv_nsec * XSEC_PER_SEC;
 	do_div(new_stamp_xsec, 1000000000);
 	new_stamp_xsec += (u64) xtime.tv_sec * XSEC_PER_SEC;
+
+	++vdso_data->tb_update_count;
+	smp_mb();
+
+	/*
+	 * Make sure time doesn't go backwards for userspace gettimeofday.
+	 */
+	tb = get_tb();
+	varp = do_gtod.varp;
+	xsec_old = mulhdu(tb - varp->tb_orig_stamp, varp->tb_to_xs)
+		+ varp->stamp_xsec;
+	xsec_new = mulhdu(tb - cur_tb, t2x) + new_stamp_xsec;
+	if (xsec_new < xsec_old)
+		new_stamp_xsec += xsec_old - xsec_new;
+
 	update_gtod(cur_tb, new_stamp_xsec, t2x);
 }
 
@@ -382,6 +612,7 @@ static void iSeries_tb_recal(void)
 						new_tb_ticks_per_jiffy, sign, tick_diff );
 				tb_ticks_per_jiffy = new_tb_ticks_per_jiffy;
 				tb_ticks_per_sec   = new_tb_ticks_per_sec;
+				calc_cputime_factors();
 				div128_by_32( XSEC_PER_SEC, 0, tb_ticks_per_sec, &divres );
 				do_gtod.tb_ticks_per_sec = tb_ticks_per_sec;
 				tb_to_xs = divres.result_low;
@@ -430,6 +661,7 @@ void timer_interrupt(struct pt_regs * regs)
 	irq_enter();
 
 	profile_tick(CPU_PROFILING, regs);
+	calculate_steal_time();
 
 #ifdef CONFIG_PPC_ISERIES
 	get_lppaca()->int_dword.fields.decr_int = 0;
@@ -451,7 +683,7 @@ void timer_interrupt(struct pt_regs * regs)
 		 * is the case.
 		 */
 		if (!cpu_is_offline(cpu))
-			update_process_times(user_mode(regs));
+			account_process_time(regs);
 
 		/*
 		 * No need to check whether cpu is offline here; boot_cpuid
@@ -508,13 +740,27 @@ void wakeup_decrementer(void)
 void __init smp_space_timers(unsigned int max_cpus)
 {
 	int i;
+	unsigned long half = tb_ticks_per_jiffy / 2;
 	unsigned long offset = tb_ticks_per_jiffy / max_cpus;
 	unsigned long previous_tb = per_cpu(last_jiffy, boot_cpuid);
 
 	/* make sure tb > per_cpu(last_jiffy, cpu) for all cpus always */
 	previous_tb -= tb_ticks_per_jiffy;
-	for_each_cpu(i) {
-		if (i != boot_cpuid) {
+	/*
+	 * The stolen time calculation for POWER5 shared-processor LPAR
+	 * systems works better if the two threads' timebase interrupts
+	 * are staggered by half a jiffy with respect to each other.
+	 */
+	for_each_possible_cpu(i) {
+		if (i == boot_cpuid)
+			continue;
+		if (i == (boot_cpuid ^ 1))
+			per_cpu(last_jiffy, i) =
+				per_cpu(last_jiffy, boot_cpuid) - half;
+		else if (i & 1)
+			per_cpu(last_jiffy, i) =
+				per_cpu(last_jiffy, i ^ 1) + half;
+		else {
 			previous_tb += offset;
 			per_cpu(last_jiffy, i) = previous_tb;
 		}
@@ -564,6 +810,10 @@ int do_settimeofday(struct timespec *tv)
 	}
 #endif
 
+	/* Make userspace gettimeofday spin until we're done. */
+	++vdso_data->tb_update_count;
+	smp_mb();
+
 	/*
 	 * Subtract off the number of nanoseconds since the
 	 * beginning of the last tick.
@@ -706,6 +956,7 @@ void __init time_init(void)
 	tb_ticks_per_sec = ppc_tb_freq;
 	tb_ticks_per_usec = ppc_tb_freq / 1000000;
 	tb_to_us = mulhwu_scale_factor(ppc_tb_freq, 1000000);
+	calc_cputime_factors();
 
 	/*
 	 * Calculate the length of each tick in ns.  It will not be
@@ -724,10 +975,16 @@ void __init time_init(void)
 	 * It is computed as:
 	 * ticklen_to_xs = 2^N / (tb_ticks_per_jiffy * 1e9)
 	 * where N = 64 + 20 - TICKLEN_SCALE - TICKLEN_SHIFT
-	 * so as to give the result as a 0.64 fixed-point fraction.
+	 * which turns out to be N = 51 - SHIFT_HZ.
+	 * This gives the result as a 0.64 fixed-point fraction.
+	 * That value is reduced by an offset amounting to 1 xsec per
+	 * 2^31 timebase ticks to avoid problems with time going backwards
+	 * by 1 xsec when we do timer_recalc_offset due to losing the
+	 * fractional xsec.  That offset is equal to ppc_tb_freq/2^51
+	 * since there are 2^20 xsec in a second.
 	 */
-	div128_by_32(1ULL << (64 + 20 - TICKLEN_SCALE - TICKLEN_SHIFT), 0,
-		     tb_ticks_per_jiffy, &res);
+	div128_by_32((1ULL << 51) - ppc_tb_freq, 0,
+		     tb_ticks_per_jiffy << SHIFT_HZ, &res);
 	div128_by_32(res.result_high, res.result_low, NSEC_PER_SEC, &res);
 	ticklen_to_xs = res.result_low;