[PATCH] improve precision of process accounting.

The unit of the cputime accouting values that are stored per process is
currently a microsecond. The CPU timer has a maximum granularity of
2**-12 microseconds. There is no benefit in storing the per process values
in the lesser precision and there is the disadvantage that the backend
has to do the rounding to microseconds. The better solution is to use
the maximum granularity of the CPU timer as cputime unit.
Signed-off-by: NMartin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/include/asm/cputime.h

@@ -11,7 +11,7 @@
 
 #include <asm/div64.h>
 
-/* We want to use micro-second resolution. */
+/* We want to use full resolution of the CPU timer: 2**-12 micro-seconds. */
 
 typedef unsigned long long cputime_t;
 typedef unsigned long long cputime64_t;
@@ -53,9 +53,9 @@ __div(unsigned long long n, unsigned int base)
 #define cputime_ge(__a, __b)		((__a) >= (__b))
 #define cputime_lt(__a, __b)		((__a) <  (__b))
 #define cputime_le(__a, __b)		((__a) <= (__b))
-#define cputime_to_jiffies(__ct)	(__div((__ct), 1000000 / HZ))
+#define cputime_to_jiffies(__ct)	(__div((__ct), 4096000000ULL / HZ))
 #define cputime_to_scaled(__ct)		(__ct)
-#define jiffies_to_cputime(__hz)	((cputime_t)(__hz) * (1000000 / HZ))
+#define jiffies_to_cputime(__hz)	((cputime_t)(__hz) * (4096000000ULL / HZ))
 
 #define cputime64_zero			(0ULL)
 #define cputime64_add(__a, __b)		((__a) + (__b))
@@ -64,7 +64,7 @@ __div(unsigned long long n, unsigned int base)
 static inline u64
 cputime64_to_jiffies64(cputime64_t cputime)
 {
-	do_div(cputime, 1000000 / HZ);
+	do_div(cputime, 4096000000ULL / HZ);
 	return cputime;
 }
 
@@ -74,13 +74,13 @@ cputime64_to_jiffies64(cputime64_t cputime)
 static inline unsigned int
 cputime_to_msecs(const cputime_t cputime)
 {
-	return __div(cputime, 1000);
+	return __div(cputime, 4096000);
 }
 
 static inline cputime_t
 msecs_to_cputime(const unsigned int m)
 {
-	return (cputime_t) m * 1000;
+	return (cputime_t) m * 4096000;
 }
 
 /*
@@ -89,13 +89,13 @@ msecs_to_cputime(const unsigned int m)
 static inline unsigned int
 cputime_to_secs(const cputime_t cputime)
 {
-	return __div(cputime, 1000000);
+	return __div(cputime, 2048000000) >> 1;
 }
 
 static inline cputime_t
 secs_to_cputime(const unsigned int s)
 {
-	return (cputime_t) s * 1000000;
+	return (cputime_t) s * 4096000000ULL;
 }
 
 /*
@@ -104,7 +104,7 @@ secs_to_cputime(const unsigned int s)
 static inline cputime_t
 timespec_to_cputime(const struct timespec *value)
 {
-        return value->tv_nsec / 1000 + (u64) value->tv_sec * 1000000;
+	return value->tv_nsec * 4096 / 1000 + (u64) value->tv_sec * 4096000000ULL;
 }
 
 static inline void
@@ -114,12 +114,12 @@ cputime_to_timespec(const cputime_t cputime, struct timespec *value)
 	register_pair rp;
 
 	rp.pair = cputime >> 1;
-	asm ("dr %0,%1" : "+d" (rp) : "d" (1000000 >> 1));
-	value->tv_nsec = rp.subreg.even * 1000;
+	asm ("dr %0,%1" : "+d" (rp) : "d" (2048000000UL));
+	value->tv_nsec = rp.subreg.even * 1000 / 4096;
 	value->tv_sec = rp.subreg.odd;
 #else
-	value->tv_nsec = (cputime % 1000000) * 1000;
-	value->tv_sec = cputime / 1000000;
+	value->tv_nsec = (cputime % 4096000000ULL) * 1000 / 4096;
+	value->tv_sec = cputime / 4096000000ULL;
 #endif
 }
 
@@ -131,7 +131,7 @@ cputime_to_timespec(const cputime_t cputime, struct timespec *value)
 static inline cputime_t
 timeval_to_cputime(const struct timeval *value)
 {
-        return value->tv_usec + (u64) value->tv_sec * 1000000;
+	return value->tv_usec * 4096 + (u64) value->tv_sec * 4096000000ULL;
 }
 
 static inline void
@@ -141,12 +141,12 @@ cputime_to_timeval(const cputime_t cputime, struct timeval *value)
 	register_pair rp;
 
 	rp.pair = cputime >> 1;
-	asm ("dr %0,%1" : "+d" (rp) : "d" (1000000 >> 1));
-	value->tv_usec = rp.subreg.even;
+	asm ("dr %0,%1" : "+d" (rp) : "d" (2048000000UL));
+	value->tv_usec = rp.subreg.even / 4096;
 	value->tv_sec = rp.subreg.odd;
 #else
-	value->tv_usec = cputime % 1000000;
-	value->tv_sec = cputime / 1000000;
+	value->tv_usec = cputime % 4096000000ULL;
+	value->tv_sec = cputime / 4096000000ULL;
 #endif
 }
 
@@ -156,13 +156,13 @@ cputime_to_timeval(const cputime_t cputime, struct timeval *value)
 static inline clock_t
 cputime_to_clock_t(cputime_t cputime)
 {
-	return __div(cputime, 1000000 / USER_HZ);
+	return __div(cputime, 4096000000ULL / USER_HZ);
 }
 
 static inline cputime_t
 clock_t_to_cputime(unsigned long x)
 {
-	return (cputime_t) x * (1000000 / USER_HZ);
+	return (cputime_t) x * (4096000000ULL / USER_HZ);
 }
 
 /*
@@ -171,7 +171,7 @@ clock_t_to_cputime(unsigned long x)
 static inline clock_t
 cputime64_to_clock_t(cputime64_t cputime)
 {
-       return __div(cputime, 1000000 / USER_HZ);
+       return __div(cputime, 4096000000ULL / USER_HZ);
 }
 
 #endif /* _S390_CPUTIME_H */

arch/s390/include/asm/lowcore.h

@@ -67,11 +67,11 @@
 #define __LC_SYNC_ENTER_TIMER		0x248
 #define __LC_ASYNC_ENTER_TIMER		0x250
 #define __LC_EXIT_TIMER			0x258
-#define __LC_LAST_UPDATE_TIMER		0x260
-#define __LC_USER_TIMER			0x268
-#define __LC_SYSTEM_TIMER		0x270
-#define __LC_LAST_UPDATE_CLOCK		0x278
-#define __LC_STEAL_CLOCK		0x280
+#define __LC_USER_TIMER			0x260
+#define __LC_SYSTEM_TIMER		0x268
+#define __LC_STEAL_TIMER		0x270
+#define __LC_LAST_UPDATE_TIMER		0x278
+#define __LC_LAST_UPDATE_CLOCK		0x280
 #define __LC_RETURN_MCCK_PSW            0x288
 #define __LC_KERNEL_STACK               0xC40
 #define __LC_THREAD_INFO		0xC44
@@ -89,11 +89,11 @@
 #define __LC_SYNC_ENTER_TIMER		0x250
 #define __LC_ASYNC_ENTER_TIMER		0x258
 #define __LC_EXIT_TIMER			0x260
-#define __LC_LAST_UPDATE_TIMER		0x268
-#define __LC_USER_TIMER			0x270
-#define __LC_SYSTEM_TIMER		0x278
-#define __LC_LAST_UPDATE_CLOCK		0x280
-#define __LC_STEAL_CLOCK		0x288
+#define __LC_USER_TIMER			0x268
+#define __LC_SYSTEM_TIMER		0x270
+#define __LC_STEAL_TIMER		0x278
+#define __LC_LAST_UPDATE_TIMER		0x280
+#define __LC_LAST_UPDATE_CLOCK		0x288
 #define __LC_RETURN_MCCK_PSW            0x290
 #define __LC_KERNEL_STACK               0xD40
 #define __LC_THREAD_INFO		0xD48
@@ -252,11 +252,11 @@ struct _lowcore
 	__u64        sync_enter_timer;         /* 0x248 */
 	__u64        async_enter_timer;        /* 0x250 */
 	__u64        exit_timer;               /* 0x258 */
-	__u64        last_update_timer;        /* 0x260 */
-	__u64        user_timer;               /* 0x268 */
-	__u64        system_timer;             /* 0x270 */
-	__u64        last_update_clock;        /* 0x278 */
-	__u64        steal_clock;              /* 0x280 */
+	__u64	     user_timer;	       /* 0x260 */
+	__u64	     system_timer;	       /* 0x268 */
+	__u64	     steal_timer;	       /* 0x270 */
+	__u64	     last_update_timer;        /* 0x278 */
+	__u64	     last_update_clock;        /* 0x280 */
         psw_t        return_mcck_psw;          /* 0x288 */
 	__u8         pad8[0xc00-0x290];        /* 0x290 */
 
@@ -343,11 +343,11 @@ struct _lowcore
 	__u64        sync_enter_timer;         /* 0x250 */
 	__u64        async_enter_timer;        /* 0x258 */
 	__u64        exit_timer;               /* 0x260 */
-	__u64        last_update_timer;        /* 0x268 */
-	__u64        user_timer;               /* 0x270 */
-	__u64        system_timer;             /* 0x278 */
-	__u64        last_update_clock;        /* 0x280 */
-	__u64        steal_clock;              /* 0x288 */
+	__u64	     user_timer;	       /* 0x268 */
+	__u64	     system_timer;	       /* 0x270 */
+	__u64	     steal_timer;	       /* 0x278 */
+	__u64	     last_update_timer;        /* 0x280 */
+	__u64	     last_update_clock;        /* 0x288 */
         psw_t        return_mcck_psw;          /* 0x290 */
         __u8         pad8[0xc00-0x2a0];        /* 0x2a0 */
         /* System info area */

arch/s390/include/asm/system.h

@@ -99,7 +99,7 @@ static inline void restore_access_regs(unsigned int *acrs)
 	prev = __switch_to(prev,next);					     \
 } while (0)
 
-extern void account_vtime(struct task_struct *);
+extern void account_vtime(struct task_struct *, struct task_struct *);
 extern void account_tick_vtime(struct task_struct *);
 extern void account_system_vtime(struct task_struct *);
 
@@ -121,7 +121,7 @@ static inline void cmma_init(void) { }
 
 #define finish_arch_switch(prev) do {					     \
 	set_fs(current->thread.mm_segment);				     \
-	account_vtime(prev);						     \
+	account_vtime(prev, current);					     \
 } while (0)
 
 #define nop() asm volatile("nop")

arch/s390/include/asm/thread_info.h

@@ -47,6 +47,8 @@ struct thread_info {
 	unsigned int		cpu;		/* current CPU */
 	int			preempt_count;	/* 0 => preemptable, <0 => BUG */
 	struct restart_block	restart_block;
+	__u64			user_timer;
+	__u64			system_timer;
 };
 
 /*

arch/s390/kernel/vtime.c

@@ -31,11 +31,10 @@ static DEFINE_PER_CPU(struct vtimer_queue, virt_cpu_timer);
  * Update process times based on virtual cpu times stored by entry.S
  * to the lowcore fields user_timer, system_timer & steal_clock.
  */
-void account_process_tick(struct task_struct *tsk, int user_tick)
+static void do_account_vtime(struct task_struct *tsk, int hardirq_offset)
 {
-	cputime_t cputime;
-	__u64 timer, clock;
-	int rcu_user_flag;
+	struct thread_info *ti = task_thread_info(tsk);
+	__u64 timer, clock, user, system, steal;
 
 	timer = S390_lowcore.last_update_timer;
 	clock = S390_lowcore.last_update_clock;
@@ -44,59 +43,47 @@ void account_process_tick(struct task_struct *tsk, int user_tick)
 		      : "=m" (S390_lowcore.last_update_timer),
 		        "=m" (S390_lowcore.last_update_clock) );
 	S390_lowcore.system_timer += timer - S390_lowcore.last_update_timer;
-	S390_lowcore.steal_clock += S390_lowcore.last_update_clock - clock;
+	S390_lowcore.steal_timer += S390_lowcore.last_update_clock - clock;
 
-	cputime = S390_lowcore.user_timer >> 12;
-	rcu_user_flag = cputime != 0;
-	S390_lowcore.user_timer -= cputime << 12;
-	S390_lowcore.steal_clock -= cputime << 12;
-	account_user_time(tsk, cputime, cputime);
+	user = S390_lowcore.user_timer - ti->user_timer;
+	S390_lowcore.steal_timer -= user;
+	ti->user_timer = S390_lowcore.user_timer;
+	account_user_time(tsk, user, user);
 
-	cputime =  S390_lowcore.system_timer >> 12;
-	S390_lowcore.system_timer -= cputime << 12;
-	S390_lowcore.steal_clock -= cputime << 12;
+	system = S390_lowcore.system_timer - ti->system_timer;
+	S390_lowcore.steal_timer -= system;
+	ti->system_timer = S390_lowcore.system_timer;
 	if (idle_task(smp_processor_id()) != current)
-		account_system_time(tsk, HARDIRQ_OFFSET, cputime, cputime);
+		account_system_time(tsk, hardirq_offset, system, system);
 	else
-		account_idle_time(cputime);
+		account_idle_time(system);
 
-	cputime = S390_lowcore.steal_clock;
-	if ((__s64) cputime > 0) {
-		cputime >>= 12;
-		S390_lowcore.steal_clock -= cputime << 12;
+	steal = S390_lowcore.steal_timer;
+	if ((s64) steal > 0) {
+		S390_lowcore.steal_timer = 0;
 		if (idle_task(smp_processor_id()) != current)
-			account_steal_time(cputime);
+			account_steal_time(steal);
 		else
-			account_idle_time(cputime);
+			account_idle_time(steal);
 	}
 }
 
-/*
- * Update process times based on virtual cpu times stored by entry.S
- * to the lowcore fields user_timer, system_timer & steal_clock.
- */
-void account_vtime(struct task_struct *tsk)
+void account_vtime(struct task_struct *prev, struct task_struct *next)
 {
-	cputime_t cputime;
-	__u64 timer;
-
-	timer = S390_lowcore.last_update_timer;
-	asm volatile ("  STPT %0"    /* Store current cpu timer value */
-		      : "=m" (S390_lowcore.last_update_timer) );
-	S390_lowcore.system_timer += timer - S390_lowcore.last_update_timer;
-
-	cputime = S390_lowcore.user_timer >> 12;
-	S390_lowcore.user_timer -= cputime << 12;
-	S390_lowcore.steal_clock -= cputime << 12;
-	account_user_time(tsk, cputime, cputime);
+	struct thread_info *ti;
+
+	do_account_vtime(prev, 0);
+	ti = task_thread_info(prev);
+	ti->user_timer = S390_lowcore.user_timer;
+	ti->system_timer = S390_lowcore.system_timer;
+	ti = task_thread_info(next);
+	S390_lowcore.user_timer = ti->user_timer;
+	S390_lowcore.system_timer = ti->system_timer;
+}
 
-	cputime =  S390_lowcore.system_timer >> 12;
-	S390_lowcore.system_timer -= cputime << 12;
-	S390_lowcore.steal_clock -= cputime << 12;
-	if (idle_task(smp_processor_id()) != current)
-		account_system_time(tsk, 0, cputime, cputime);
-	else
-		account_idle_time(cputime);
+void account_process_tick(struct task_struct *tsk, int user_tick)
+{
+	do_account_vtime(tsk, HARDIRQ_OFFSET);
 }
 
 /*
@@ -105,21 +92,21 @@ void account_vtime(struct task_struct *tsk)
  */
 void account_system_vtime(struct task_struct *tsk)
 {
-	cputime_t cputime;
-	__u64 timer;
+	struct thread_info *ti = task_thread_info(tsk);
+	__u64 timer, system;
 
 	timer = S390_lowcore.last_update_timer;
 	asm volatile ("  STPT %0"    /* Store current cpu timer value */
 		      : "=m" (S390_lowcore.last_update_timer) );
 	S390_lowcore.system_timer += timer - S390_lowcore.last_update_timer;
 
-	cputime =  S390_lowcore.system_timer >> 12;
-	S390_lowcore.system_timer -= cputime << 12;
-	S390_lowcore.steal_clock -= cputime << 12;
+	system = S390_lowcore.system_timer - ti->system_timer;
+	S390_lowcore.steal_timer -= system;
+	ti->system_timer = S390_lowcore.system_timer;
 	if (in_irq() || idle_task(smp_processor_id()) != current)
-		account_system_time(tsk, 0, cputime, cputime);
+		account_system_time(tsk, 0, system, system);
 	else
-		account_idle_time(cputime);
+		account_idle_time(system);
 }
 EXPORT_SYMBOL_GPL(account_system_vtime);
 
@@ -490,8 +477,8 @@ void init_cpu_vtimer(void)
 	/* kick the virtual timer */
 	S390_lowcore.exit_timer = VTIMER_MAX_SLICE;
 	S390_lowcore.last_update_timer = VTIMER_MAX_SLICE;
-	asm volatile ("SPT %0" : : "m" (S390_lowcore.last_update_timer));
 	asm volatile ("STCK %0" : "=m" (S390_lowcore.last_update_clock));
+	asm volatile ("SPT %0" : : "m" (S390_lowcore.last_update_timer));
 
 	/* enable cpu timer interrupts */
 	__ctl_set_bit(0,10);