powerpc: Implement accurate task and CPU time accounting

This implements accurate task and cpu time accounting for 64-bit
powerpc kernels.  Instead of accounting a whole jiffy of time to a
task on a timer interrupt because that task happened to be running at
the time, we now account time in units of timebase ticks according to
the actual time spent by the task in user mode and kernel mode.  We
also count the time spent processing hardware and software interrupts
accurately.  This is conditional on CONFIG_VIRT_CPU_ACCOUNTING.  If
that is not set, we do tick-based approximate accounting as before.

To get this accurate information, we read either the PURR (processor
utilization of resources register) on POWER5 machines, or the timebase
on other machines on

* each entry to the kernel from usermode
* each exit to usermode
* transitions between process context, hard irq context and soft irq
  context in kernel mode
* context switches.

On POWER5 systems with shared-processor logical partitioning we also
read both the PURR and the timebase at each timer interrupt and
context switch in order to determine how much time has been taken by
the hypervisor to run other partitions ("steal" time).  Unfortunately,
since we need values of the PURR on both threads at the same time to
accurately calculate the steal time, and since we can only calculate
steal time on a per-core basis, the apportioning of the steal time
between idle time (time which we ceded to the hypervisor in the idle
loop) and actual stolen time is somewhat approximate at the moment.

This is all based quite heavily on what s390 does, and it uses the
generic interfaces that were added by the s390 developers,
i.e. account_system_time(), account_user_time(), etc.

This patch doesn't add any new interfaces between the kernel and
userspace, and doesn't change the units in which time is reported to
userspace by things such as /proc/stat, /proc/<pid>/stat, getrusage(),
times(), etc.  Internally the various task and cpu times are stored in
timebase units, but they are converted to USER_HZ units (1/100th of a
second) when reported to userspace.  Some precision is therefore lost
but there should not be any accumulating error, since the internal
accumulation is at full precision.

Signed-off-by: Paul Mackerras <paulus@samba.org>

8 files changed, 280 insertions, 3 deletions

diff --git a/include/asm-powerpc/cputable.h b/include/asm-powerpc/cputable.h
index 90d005bb4d1c..99d12ff6346c 100644
--- a/include/asm-powerpc/cputable.h
+++ b/include/asm-powerpc/cputable.h
@@ -117,6 +117,7 @@ extern void do_cpu_ftr_fixups(unsigned long offset);
 #define CPU_FTR_MMCRA_SIHV		ASM_CONST(0x0000080000000000)
 #define CPU_FTR_CI_LARGE_PAGE		ASM_CONST(0x0000100000000000)
 #define CPU_FTR_PAUSE_ZERO		ASM_CONST(0x0000200000000000)
+#define CPU_FTR_PURR			ASM_CONST(0x0000400000000000)
 #else
 /* ensure on 32b processors the flags are available for compiling but
  * don't do anything */
@@ -132,6 +133,7 @@ extern void do_cpu_ftr_fixups(unsigned long offset);
 #define CPU_FTR_LOCKLESS_TLBIE		ASM_CONST(0x0)
 #define CPU_FTR_MMCRA_SIHV		ASM_CONST(0x0)
 #define CPU_FTR_CI_LARGE_PAGE		ASM_CONST(0x0)
+#define CPU_FTR_PURR			ASM_CONST(0x0)
 #endif
 
 #ifndef __ASSEMBLY__
@@ -316,7 +318,7 @@ enum {
 	    CPU_FTR_HPTE_TABLE | CPU_FTR_PPCAS_ARCH_V2 |
 	    CPU_FTR_MMCRA | CPU_FTR_SMT |
 	    CPU_FTR_COHERENT_ICACHE | CPU_FTR_LOCKLESS_TLBIE |
-	    CPU_FTR_MMCRA_SIHV,
+	    CPU_FTR_MMCRA_SIHV | CPU_FTR_PURR,
 	CPU_FTRS_CELL = CPU_FTR_SPLIT_ID_CACHE | CPU_FTR_USE_TB |
 	    CPU_FTR_HPTE_TABLE | CPU_FTR_PPCAS_ARCH_V2 |
 	    CPU_FTR_ALTIVEC_COMP | CPU_FTR_MMCRA | CPU_FTR_SMT |
diff --git a/include/asm-powerpc/cputime.h b/include/asm-powerpc/cputime.h
index 6d68ad7e0ea3..a21185d47883 100644
--- a/include/asm-powerpc/cputime.h
+++ b/include/asm-powerpc/cputime.h
@@ -1 +1,203 @@
+/*
+ * Definitions for measuring cputime on powerpc machines.
+ *
+ * Copyright (C) 2006 Paul Mackerras, IBM Corp.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version
+ * 2 of the License, or (at your option) any later version.
+ *
+ * If we have CONFIG_VIRT_CPU_ACCOUNTING, we measure cpu time in
+ * the same units as the timebase.  Otherwise we measure cpu time
+ * in jiffies using the generic definitions.
+ */
+
+#ifndef __POWERPC_CPUTIME_H
+#define __POWERPC_CPUTIME_H
+
+#ifndef CONFIG_VIRT_CPU_ACCOUNTING
 #include <asm-generic/cputime.h>
+#else
+
+#include <linux/types.h>
+#include <linux/time.h>
+#include <asm/div64.h>
+#include <asm/time.h>
+#include <asm/param.h>
+
+typedef u64 cputime_t;
+typedef u64 cputime64_t;
+
+#define cputime_zero			((cputime_t)0)
+#define cputime_max			((~((cputime_t)0) >> 1) - 1)
+#define cputime_add(__a, __b)		((__a) +  (__b))
+#define cputime_sub(__a, __b)		((__a) -  (__b))
+#define cputime_div(__a, __n)		((__a) /  (__n))
+#define cputime_halve(__a)		((__a) >> 1)
+#define cputime_eq(__a, __b)		((__a) == (__b))
+#define cputime_gt(__a, __b)		((__a) >  (__b))
+#define cputime_ge(__a, __b)		((__a) >= (__b))
+#define cputime_lt(__a, __b)		((__a) <  (__b))
+#define cputime_le(__a, __b)		((__a) <= (__b))
+
+#define cputime64_zero			((cputime64_t)0)
+#define cputime64_add(__a, __b)		((__a) + (__b))
+#define cputime_to_cputime64(__ct)	(__ct)
+
+#ifdef __KERNEL__
+
+/*
+ * Convert cputime <-> jiffies
+ */
+extern u64 __cputime_jiffies_factor;
+
+static inline unsigned long cputime_to_jiffies(const cputime_t ct)
+{
+	return mulhdu(ct, __cputime_jiffies_factor);
+}
+
+static inline cputime_t jiffies_to_cputime(const unsigned long jif)
+{
+	cputime_t ct;
+	unsigned long sec;
+
+	/* have to be a little careful about overflow */
+	ct = jif % HZ;
+	sec = jif / HZ;
+	if (ct) {
+		ct *= tb_ticks_per_sec;
+		do_div(ct, HZ);
+	}
+	if (sec)
+		ct += (cputime_t) sec * tb_ticks_per_sec;
+	return ct;
+}
+
+static inline u64 cputime64_to_jiffies64(const cputime_t ct)
+{
+	return mulhdu(ct, __cputime_jiffies_factor);
+}
+
+/*
+ * Convert cputime <-> milliseconds
+ */
+extern u64 __cputime_msec_factor;
+
+static inline unsigned long cputime_to_msecs(const cputime_t ct)
+{
+	return mulhdu(ct, __cputime_msec_factor);
+}
+
+static inline cputime_t msecs_to_cputime(const unsigned long ms)
+{
+	cputime_t ct;
+	unsigned long sec;
+
+	/* have to be a little careful about overflow */
+	ct = ms % 1000;
+	sec = ms / 1000;
+	if (ct) {
+		ct *= tb_ticks_per_sec;
+		do_div(ct, 1000);
+	}
+	if (sec)
+		ct += (cputime_t) sec * tb_ticks_per_sec;
+	return ct;
+}
+
+/*
+ * Convert cputime <-> seconds
+ */
+extern u64 __cputime_sec_factor;
+
+static inline unsigned long cputime_to_secs(const cputime_t ct)
+{
+	return mulhdu(ct, __cputime_sec_factor);
+}
+
+static inline cputime_t secs_to_cputime(const unsigned long sec)
+{
+	return (cputime_t) sec * tb_ticks_per_sec;
+}
+
+/*
+ * Convert cputime <-> timespec
+ */
+static inline void cputime_to_timespec(const cputime_t ct, struct timespec *p)
+{
+	u64 x = ct;
+	unsigned int frac;
+
+	frac = do_div(x, tb_ticks_per_sec);
+	p->tv_sec = x;
+	x = (u64) frac * 1000000000;
+	do_div(x, tb_ticks_per_sec);
+	p->tv_nsec = x;
+}
+
+static inline cputime_t timespec_to_cputime(const struct timespec *p)
+{
+	cputime_t ct;
+
+	ct = (u64) p->tv_nsec * tb_ticks_per_sec;
+	do_div(ct, 1000000000);
+	return ct + (u64) p->tv_sec * tb_ticks_per_sec;
+}
+
+/*
+ * Convert cputime <-> timeval
+ */
+static inline void cputime_to_timeval(const cputime_t ct, struct timeval *p)
+{
+	u64 x = ct;
+	unsigned int frac;
+
+	frac = do_div(x, tb_ticks_per_sec);
+	p->tv_sec = x;
+	x = (u64) frac * 1000000;
+	do_div(x, tb_ticks_per_sec);
+	p->tv_usec = x;
+}
+
+static inline cputime_t timeval_to_cputime(const struct timeval *p)
+{
+	cputime_t ct;
+
+	ct = (u64) p->tv_usec * tb_ticks_per_sec;
+	do_div(ct, 1000000);
+	return ct + (u64) p->tv_sec * tb_ticks_per_sec;
+}
+
+/*
+ * Convert cputime <-> clock_t (units of 1/USER_HZ seconds)
+ */
+extern u64 __cputime_clockt_factor;
+
+static inline unsigned long cputime_to_clock_t(const cputime_t ct)
+{
+	return mulhdu(ct, __cputime_clockt_factor);
+}
+
+static inline cputime_t clock_t_to_cputime(const unsigned long clk)
+{
+	cputime_t ct;
+	unsigned long sec;
+
+	/* have to be a little careful about overflow */
+	ct = clk % USER_HZ;
+	sec = clk / USER_HZ;
+	if (ct) {
+		ct *= tb_ticks_per_sec;
+		do_div(ct, USER_HZ);
+	}
+	if (sec)
+		ct += (cputime_t) sec * tb_ticks_per_sec;
+	return ct;
+}
+
+#define cputime64_to_clock_t(ct)	cputime_to_clock_t((cputime_t)(ct))
+
+#endif /* __KERNEL__ */
+#endif /* CONFIG_VIRT_CPU_ACCOUNTING */
+#endif /* __POWERPC_CPUTIME_H */
diff --git a/include/asm-powerpc/irq.h b/include/asm-powerpc/irq.h
index 8eb7e857ec4c..51f87d9993b6 100644
--- a/include/asm-powerpc/irq.h
+++ b/include/asm-powerpc/irq.h
@@ -479,6 +479,10 @@ extern int distribute_irqs;
 struct irqaction;
 struct pt_regs;
 
+#define __ARCH_HAS_DO_SOFTIRQ
+
+extern void __do_softirq(void);
+
 #ifdef CONFIG_IRQSTACKS
 /*
  * Per-cpu stacks for handling hard and soft interrupts.
@@ -491,8 +495,6 @@ extern void call_do_softirq(struct thread_info *tp);
 extern int call___do_IRQ(int irq, struct pt_regs *regs,
 		struct thread_info *tp);
 
-#define __ARCH_HAS_DO_SOFTIRQ
-
 #else
 #define irq_ctx_init()
 
diff --git a/include/asm-powerpc/paca.h b/include/asm-powerpc/paca.h
index ec94b51074fc..4465b95ebef0 100644
--- a/include/asm-powerpc/paca.h
+++ b/include/asm-powerpc/paca.h
@@ -96,6 +96,11 @@ struct paca_struct {
 	u64 saved_r1;			/* r1 save for RTAS calls */
 	u64 saved_msr;			/* MSR saved here by enter_rtas */
 	u8 proc_enabled;		/* irq soft-enable flag */
+
+	/* Stuff for accurate time accounting */
+	u64 user_time;			/* accumulated usermode TB ticks */
+	u64 system_time;		/* accumulated system TB ticks */
+	u64 startpurr;			/* PURR/TB value snapshot */
 };
 
 extern struct paca_struct paca[];
diff --git a/include/asm-powerpc/ppc_asm.h b/include/asm-powerpc/ppc_asm.h
index ab8688d39024..dd1c0a913d5f 100644
--- a/include/asm-powerpc/ppc_asm.h
+++ b/include/asm-powerpc/ppc_asm.h
@@ -15,6 +15,48 @@
 #define SZL			(BITS_PER_LONG/8)
 
 /*
+ * Stuff for accurate CPU time accounting.
+ * These macros handle transitions between user and system state
+ * in exception entry and exit and accumulate time to the
+ * user_time and system_time fields in the paca.
+ */
+
+#ifndef CONFIG_VIRT_CPU_ACCOUNTING
+#define ACCOUNT_CPU_USER_ENTRY(ra, rb)
+#define ACCOUNT_CPU_USER_EXIT(ra, rb)
+#else
+#define ACCOUNT_CPU_USER_ENTRY(ra, rb)					\
+	beq	2f;			/* if from kernel mode */	\
+BEGIN_FTR_SECTION;							\
+	mfspr	ra,SPRN_PURR;		/* get processor util. reg */	\
+END_FTR_SECTION_IFSET(CPU_FTR_PURR);					\
+BEGIN_FTR_SECTION;							\
+	mftb	ra;			/* or get TB if no PURR */	\
+END_FTR_SECTION_IFCLR(CPU_FTR_PURR);					\
+	ld	rb,PACA_STARTPURR(r13);				\
+	std	ra,PACA_STARTPURR(r13);					\
+	subf	rb,rb,ra;		/* subtract start value */	\
+	ld	ra,PACA_USER_TIME(r13);					\
+	add	ra,ra,rb;		/* add on to user time */	\
+	std	ra,PACA_USER_TIME(r13);					\
+2:
+
+#define ACCOUNT_CPU_USER_EXIT(ra, rb)					\
+BEGIN_FTR_SECTION;							\
+	mfspr	ra,SPRN_PURR;		/* get processor util. reg */	\
+END_FTR_SECTION_IFSET(CPU_FTR_PURR);					\
+BEGIN_FTR_SECTION;							\
+	mftb	ra;			/* or get TB if no PURR */	\
+END_FTR_SECTION_IFCLR(CPU_FTR_PURR);					\
+	ld	rb,PACA_STARTPURR(r13);				\
+	std	ra,PACA_STARTPURR(r13);					\
+	subf	rb,rb,ra;		/* subtract start value */	\
+	ld	ra,PACA_SYSTEM_TIME(r13);				\
+	add	ra,ra,rb;		/* add on to user time */	\
+	std	ra,PACA_SYSTEM_TIME(r13);
+#endif
+
+/*
  * Macros for storing registers into and loading registers from
  * exception frames.
  */
diff --git a/include/asm-powerpc/system.h b/include/asm-powerpc/system.h
index d9bf53653b10..41b7a5b3d701 100644
--- a/include/asm-powerpc/system.h
+++ b/include/asm-powerpc/system.h
@@ -424,5 +424,9 @@ static inline void create_function_call(unsigned long addr, void * func)
 	create_branch(addr, func_addr, BRANCH_SET_LINK);
 }
 
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+extern void account_system_vtime(struct task_struct *);
+#endif
+
 #endif /* __KERNEL__ */
 #endif /* _ASM_POWERPC_SYSTEM_H */
diff --git a/include/asm-powerpc/time.h b/include/asm-powerpc/time.h
index baddc9ab57ad..912118db13ae 100644
--- a/include/asm-powerpc/time.h
+++ b/include/asm-powerpc/time.h
@@ -41,6 +41,7 @@ extern time_t last_rtc_update;
 
 extern void generic_calibrate_decr(void);
 extern void wakeup_decrementer(void);
+extern void snapshot_timebase(void);
 
 /* Some sane defaults: 125 MHz timebase, 1GHz processor */
 extern unsigned long ppc_proc_freq;
@@ -221,5 +222,19 @@ struct cpu_usage {
 
 DECLARE_PER_CPU(struct cpu_usage, cpu_usage_array);
 
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+extern void account_process_vtime(struct task_struct *tsk);
+#else
+#define account_process_vtime(tsk)		do { } while (0)
+#endif
+
+#if defined(CONFIG_VIRT_CPU_ACCOUNTING) && defined(CONFIG_PPC_SPLPAR)
+extern void calculate_steal_time(void);
+extern void snapshot_timebases(void);
+#else
+#define calculate_steal_time()			do { } while (0)
+#define snapshot_timebases()			do { } while (0)
+#endif
+
 #endif /* __KERNEL__ */
 #endif /* __PPC64_TIME_H */
diff --git a/include/asm-ppc/time.h b/include/asm-ppc/time.h
index 321fb75b5f22..c86112323c9f 100644
--- a/include/asm-ppc/time.h
+++ b/include/asm-ppc/time.h
@@ -153,5 +153,10 @@ extern __inline__ unsigned binary_tbl(void) {
 ({unsigned z; asm ("mulhwu %0,%1,%2" : "=r" (z) : "r" (x), "r" (y)); z;})
 
 unsigned mulhwu_scale_factor(unsigned, unsigned);
+
+#define account_process_vtime(tsk)		do { } while (0)
+#define calculate_steal_time()			do { } while (0)
+#define snapshot_timebases()			do { } while (0)
+
 #endif /* __ASM_TIME_H__ */
 #endif /* __KERNEL__ */