Merge branch 'cputime' of git://git390.osdl.marist.edu/pub/scm/linux-2.6

* 'cputime' of git://git390.osdl.marist.edu/pub/scm/linux-2.6:
  [PATCH] fast vdso implementation for CLOCK_THREAD_CPUTIME_ID
  [PATCH] improve idle cputime accounting
  [PATCH] improve precision of idle time detection.
  [PATCH] improve precision of process accounting.
  [PATCH] idle cputime accounting
  [PATCH] fix scaled & unscaled cputime accounting

19 files changed, 563 insertions, 368 deletions

diff --git a/arch/s390/include/asm/cpu.h b/arch/s390/include/asm/cpu.h
index e5a6a9ba3adf..d60a2eefb17b 100644
--- a/arch/s390/include/asm/cpu.h
+++ b/arch/s390/include/asm/cpu.h
@@ -14,7 +14,6 @@
 
 struct s390_idle_data {
 	spinlock_t lock;
-	unsigned int in_idle;
 	unsigned long long idle_count;
 	unsigned long long idle_enter;
 	unsigned long long idle_time;
@@ -22,12 +21,12 @@ struct s390_idle_data {
 
 DECLARE_PER_CPU(struct s390_idle_data, s390_idle);
 
-void s390_idle_leave(void);
+void vtime_start_cpu(void);
 
 static inline void s390_idle_check(void)
 {
-	if ((&__get_cpu_var(s390_idle))->in_idle)
-		s390_idle_leave();
+	if ((&__get_cpu_var(s390_idle))->idle_enter != 0ULL)
+		vtime_start_cpu();
 }
 
 #endif /* _ASM_S390_CPU_H_ */
diff --git a/arch/s390/include/asm/cputime.h b/arch/s390/include/asm/cputime.h
index 133ce054fc89..521726430afa 100644
--- a/arch/s390/include/asm/cputime.h
+++ b/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 */
diff --git a/arch/s390/include/asm/lowcore.h b/arch/s390/include/asm/lowcore.h
index 0bc51d52a899..ffdef5fe8587 100644
--- a/arch/s390/include/asm/lowcore.h
+++ b/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
@@ -106,8 +106,10 @@
 #define __LC_IPLDEV                     0xDB8
 #define __LC_CURRENT			0xDD8
 #define __LC_INT_CLOCK			0xDE8
+#define __LC_VDSO_PER_CPU		0xE38
 #endif /* __s390x__ */
 
+#define __LC_PASTE			0xE40
 
 #define __LC_PANIC_MAGIC		0xE00
 #ifndef __s390x__
@@ -252,11 +254,11 @@ 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 +345,11 @@ 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 */
@@ -381,7 +383,12 @@ struct _lowcore
         /* whether the kernel died with panic() or not */
         __u32        panic_magic;              /* 0xe00 */
 
-	__u8         pad13[0x11b8-0xe04];      /* 0xe04 */
+	/* Per cpu primary space access list */
+	__u8	     pad_0xe04[0xe3c-0xe04];   /* 0xe04 */
+	__u32	     vdso_per_cpu_data;	       /* 0xe3c */
+	__u32	     paste[16];		       /* 0xe40 */
+
+	__u8	     pad13[0x11b8-0xe80];      /* 0xe80 */
 
 	/* 64 bit extparam used for pfault, diag 250 etc  */
 	__u64        ext_params2;               /* 0x11B8 */
diff --git a/arch/s390/include/asm/system.h b/arch/s390/include/asm/system.h
index 024ef42ed6d7..3a8b26eb1f2e 100644
--- a/arch/s390/include/asm/system.h
+++ b/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")
diff --git a/arch/s390/include/asm/thread_info.h b/arch/s390/include/asm/thread_info.h
index c1eaf9604da7..c544aa524535 100644
--- a/arch/s390/include/asm/thread_info.h
+++ b/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;
 };
 
 /*
diff --git a/arch/s390/include/asm/timer.h b/arch/s390/include/asm/timer.h
index 61705d60f995..e4bcab739c19 100644
--- a/arch/s390/include/asm/timer.h
+++ b/arch/s390/include/asm/timer.h
@@ -23,20 +23,18 @@ struct vtimer_list {
 	__u64 expires;
 	__u64 interval;
 
-	spinlock_t lock;
-	unsigned long magic;
-
 	void (*function)(unsigned long);
 	unsigned long data;
 };
 
-/* the offset value will wrap after ca. 71 years */
+/* the vtimer value will wrap after ca. 71 years */
 struct vtimer_queue {
 	struct list_head list;
 	spinlock_t lock;
-	__u64 to_expire;	  /* current event expire time */
-	__u64 offset;		  /* list offset to zero */
-	__u64 idle;		  /* temp var for idle */
+	__u64 timer;		/* last programmed timer */
+	__u64 elapsed;		/* elapsed time of timer expire values */
+	__u64 idle;		/* temp var for idle */
+	int do_spt;		/* =1: reprogram cpu timer in idle */
 };
 
 extern void init_virt_timer(struct vtimer_list *timer);
@@ -48,8 +46,8 @@ extern int del_virt_timer(struct vtimer_list *timer);
 extern void init_cpu_vtimer(void);
 extern void vtime_init(void);
 
-extern void vtime_start_cpu_timer(void);
-extern void vtime_stop_cpu_timer(void);
+extern void vtime_stop_cpu(void);
+extern void vtime_start_leave(void);
 
 #endif /* __KERNEL__ */
 
diff --git a/arch/s390/include/asm/vdso.h b/arch/s390/include/asm/vdso.h
index a44f4fe16a35..7bdd7c8ebc91 100644
--- a/arch/s390/include/asm/vdso.h
+++ b/arch/s390/include/asm/vdso.h
@@ -12,9 +12,9 @@
 #ifndef __ASSEMBLY__
 
 /*
- * Note about this structure:
+ * Note about the vdso_data and vdso_per_cpu_data structures:
  *
- * NEVER USE THIS IN USERSPACE CODE DIRECTLY. The layout of this
+ * NEVER USE THEM IN USERSPACE CODE DIRECTLY. The layout of the
  * structure is supposed to be known only to the function in the vdso
  * itself and may change without notice.
  */
@@ -28,10 +28,21 @@ struct vdso_data {
 	__u64 wtom_clock_nsec;		/*				0x28 */
 	__u32 tz_minuteswest;		/* Minutes west of Greenwich	0x30 */
 	__u32 tz_dsttime;		/* Type of dst correction	0x34 */
+	__u32 ectg_available;
+};
+
+struct vdso_per_cpu_data {
+	__u64 ectg_timer_base;
+	__u64 ectg_user_time;
 };
 
 extern struct vdso_data *vdso_data;
 
+#ifdef CONFIG_64BIT
+int vdso_alloc_per_cpu(int cpu, struct _lowcore *lowcore);
+void vdso_free_per_cpu(int cpu, struct _lowcore *lowcore);
+#endif
+
 #endif /* __ASSEMBLY__ */
 
 #endif /* __KERNEL__ */
diff --git a/arch/s390/kernel/asm-offsets.c b/arch/s390/kernel/asm-offsets.c
index e641f60bac99..67a60016babb 100644
--- a/arch/s390/kernel/asm-offsets.c
+++ b/arch/s390/kernel/asm-offsets.c
@@ -48,6 +48,11 @@ int main(void)
 	DEFINE(__VDSO_WTOM_SEC, offsetof(struct vdso_data, wtom_clock_sec));
 	DEFINE(__VDSO_WTOM_NSEC, offsetof(struct vdso_data, wtom_clock_nsec));
 	DEFINE(__VDSO_TIMEZONE, offsetof(struct vdso_data, tz_minuteswest));
+	DEFINE(__VDSO_ECTG_OK, offsetof(struct vdso_data, ectg_available));
+	DEFINE(__VDSO_ECTG_BASE,
+	       offsetof(struct vdso_per_cpu_data, ectg_timer_base));
+	DEFINE(__VDSO_ECTG_USER,
+	       offsetof(struct vdso_per_cpu_data, ectg_user_time));
 	/* constants used by the vdso */
 	DEFINE(CLOCK_REALTIME, CLOCK_REALTIME);
 	DEFINE(CLOCK_MONOTONIC, CLOCK_MONOTONIC);
diff --git a/arch/s390/kernel/entry.S b/arch/s390/kernel/entry.S
index 55de521aef77..1268aa2991bf 100644
--- a/arch/s390/kernel/entry.S
+++ b/arch/s390/kernel/entry.S
@@ -583,8 +583,8 @@ kernel_per:
 
 	.globl io_int_handler
 io_int_handler:
-	stpt	__LC_ASYNC_ENTER_TIMER
 	stck	__LC_INT_CLOCK
+	stpt	__LC_ASYNC_ENTER_TIMER
 	SAVE_ALL_BASE __LC_SAVE_AREA+16
 	SAVE_ALL_ASYNC __LC_IO_OLD_PSW,__LC_SAVE_AREA+16
 	CREATE_STACK_FRAME __LC_IO_OLD_PSW,__LC_SAVE_AREA+16
@@ -723,8 +723,8 @@ io_notify_resume:
 
 	.globl	ext_int_handler
 ext_int_handler:
-	stpt	__LC_ASYNC_ENTER_TIMER
 	stck	__LC_INT_CLOCK
+	stpt	__LC_ASYNC_ENTER_TIMER
 	SAVE_ALL_BASE __LC_SAVE_AREA+16
 	SAVE_ALL_ASYNC __LC_EXT_OLD_PSW,__LC_SAVE_AREA+16
 	CREATE_STACK_FRAME __LC_EXT_OLD_PSW,__LC_SAVE_AREA+16
@@ -750,6 +750,7 @@ __critical_end:
 
 	.globl mcck_int_handler
 mcck_int_handler:
+	stck	__LC_INT_CLOCK
 	spt	__LC_CPU_TIMER_SAVE_AREA	# revalidate cpu timer
 	lm	%r0,%r15,__LC_GPREGS_SAVE_AREA	# revalidate gprs
 	SAVE_ALL_BASE __LC_SAVE_AREA+32
diff --git a/arch/s390/kernel/entry64.S b/arch/s390/kernel/entry64.S
index 16bb4fd1a403..c6fbde13971a 100644
--- a/arch/s390/kernel/entry64.S
+++ b/arch/s390/kernel/entry64.S
@@ -177,8 +177,11 @@ _TIF_WORK_INT = (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_NEED_RESCHED | \
 	.if !\sync
 	ni	\psworg+1,0xfd		# clear wait state bit
 	.endif
-	lmg	%r0,%r15,SP_R0(%r15)	# load gprs 0-15 of user
+	lg	%r14,__LC_VDSO_PER_CPU
+	lmg	%r0,%r13,SP_R0(%r15)	# load gprs 0-13 of user
 	stpt	__LC_EXIT_TIMER
+	mvc	__VDSO_ECTG_BASE(16,%r14),__LC_EXIT_TIMER
+	lmg	%r14,%r15,SP_R14(%r15)	# load grps 14-15 of user
 	lpswe	\psworg			# back to caller
 	.endm
 
@@ -559,8 +562,8 @@ kernel_per:
  */
 	.globl io_int_handler
 io_int_handler:
-	stpt	__LC_ASYNC_ENTER_TIMER
 	stck	__LC_INT_CLOCK
+	stpt	__LC_ASYNC_ENTER_TIMER
 	SAVE_ALL_BASE __LC_SAVE_AREA+32
 	SAVE_ALL_ASYNC __LC_IO_OLD_PSW,__LC_SAVE_AREA+32
 	CREATE_STACK_FRAME __LC_IO_OLD_PSW,__LC_SAVE_AREA+32
@@ -721,8 +724,8 @@ io_notify_resume:
  */
 	.globl	ext_int_handler
 ext_int_handler:
-	stpt	__LC_ASYNC_ENTER_TIMER
 	stck	__LC_INT_CLOCK
+	stpt	__LC_ASYNC_ENTER_TIMER
 	SAVE_ALL_BASE __LC_SAVE_AREA+32
 	SAVE_ALL_ASYNC __LC_EXT_OLD_PSW,__LC_SAVE_AREA+32
 	CREATE_STACK_FRAME __LC_EXT_OLD_PSW,__LC_SAVE_AREA+32
@@ -746,6 +749,7 @@ __critical_end:
  */
 	.globl mcck_int_handler
 mcck_int_handler:
+	stck	__LC_INT_CLOCK
 	la	%r1,4095		# revalidate r1
 	spt	__LC_CPU_TIMER_SAVE_AREA-4095(%r1)	# revalidate cpu timer
 	lmg	%r0,%r15,__LC_GPREGS_SAVE_AREA-4095(%r1)# revalidate gprs
@@ -979,23 +983,23 @@ cleanup_sysc_return:
 
 cleanup_sysc_leave:
 	clc	8(8,%r12),BASED(cleanup_sysc_leave_insn)
-	je	2f
-	mvc	__LC_EXIT_TIMER(8),__LC_ASYNC_ENTER_TIMER
+	je	3f
 	clc	8(8,%r12),BASED(cleanup_sysc_leave_insn+8)
-	je	2f
-	mvc	__LC_RETURN_PSW(16),SP_PSW(%r15)
+	jhe	0f
+	mvc	__LC_EXIT_TIMER(8),__LC_ASYNC_ENTER_TIMER
+0:	mvc	__LC_RETURN_PSW(16),SP_PSW(%r15)
 	cghi	%r12,__LC_MCK_OLD_PSW
-	jne	0f
+	jne	1f
 	mvc	__LC_SAVE_AREA+64(32),SP_R12(%r15)
-	j	1f
-0:	mvc	__LC_SAVE_AREA+32(32),SP_R12(%r15)
-1:	lmg	%r0,%r11,SP_R0(%r15)
+	j	2f
+1:	mvc	__LC_SAVE_AREA+32(32),SP_R12(%r15)
+2:	lmg	%r0,%r11,SP_R0(%r15)
 	lg	%r15,SP_R15(%r15)
-2:	la	%r12,__LC_RETURN_PSW
+3:	la	%r12,__LC_RETURN_PSW
 	br	%r14
 cleanup_sysc_leave_insn:
 	.quad	sysc_done - 4
-	.quad	sysc_done - 8
+	.quad	sysc_done - 16
 
 cleanup_io_return:
 	mvc	__LC_RETURN_PSW(8),0(%r12)
@@ -1005,23 +1009,23 @@ cleanup_io_return:
 
 cleanup_io_leave:
 	clc	8(8,%r12),BASED(cleanup_io_leave_insn)
-	je	2f
-	mvc	__LC_EXIT_TIMER(8),__LC_ASYNC_ENTER_TIMER
+	je	3f
 	clc	8(8,%r12),BASED(cleanup_io_leave_insn+8)
-	je	2f
-	mvc	__LC_RETURN_PSW(16),SP_PSW(%r15)
+	jhe	0f
+	mvc	__LC_EXIT_TIMER(8),__LC_ASYNC_ENTER_TIMER
+0:	mvc	__LC_RETURN_PSW(16),SP_PSW(%r15)
 	cghi	%r12,__LC_MCK_OLD_PSW
-	jne	0f
+	jne	1f
 	mvc	__LC_SAVE_AREA+64(32),SP_R12(%r15)
-	j	1f
-0:	mvc	__LC_SAVE_AREA+32(32),SP_R12(%r15)
-1:	lmg	%r0,%r11,SP_R0(%r15)
+	j	2f
+1:	mvc	__LC_SAVE_AREA+32(32),SP_R12(%r15)
+2:	lmg	%r0,%r11,SP_R0(%r15)
 	lg	%r15,SP_R15(%r15)
-2:	la	%r12,__LC_RETURN_PSW
+3:	la	%r12,__LC_RETURN_PSW
 	br	%r14
 cleanup_io_leave_insn:
 	.quad	io_done - 4
-	.quad	io_done - 8
+	.quad	io_done - 16
 
 /*
  * Integer constants
diff --git a/arch/s390/kernel/head64.S b/arch/s390/kernel/head64.S
index 3ccd36b24b8f..f9f70aa15244 100644
--- a/arch/s390/kernel/head64.S
+++ b/arch/s390/kernel/head64.S
@@ -87,6 +87,8 @@ startup_continue:
 	lg	%r12,.Lparmaddr-.LPG1(%r13)	# pointer to parameter area
 					# move IPL device to lowcore
 	mvc	__LC_IPLDEV(4),IPL_DEVICE+4-PARMAREA(%r12)
+	lghi	%r0,__LC_PASTE
+	stg	%r0,__LC_VDSO_PER_CPU
 #
 # Setup stack
 #
diff --git a/arch/s390/kernel/process.c b/arch/s390/kernel/process.c
index 04f8c67a6101..b6110bdf8dc2 100644
--- a/arch/s390/kernel/process.c
+++ b/arch/s390/kernel/process.c
@@ -38,6 +38,7 @@
 #include <linux/utsname.h>
 #include <linux/tick.h>
 #include <linux/elfcore.h>
+#include <linux/kernel_stat.h>
 #include <asm/uaccess.h>
 #include <asm/pgtable.h>
 #include <asm/system.h>
@@ -45,7 +46,6 @@
 #include <asm/processor.h>
 #include <asm/irq.h>
 #include <asm/timer.h>
-#include <asm/cpu.h>
 #include "entry.h"
 
 asmlinkage void ret_from_fork(void) asm ("ret_from_fork");
@@ -75,36 +75,6 @@ unsigned long thread_saved_pc(struct task_struct *tsk)
 	return sf->gprs[8];
 }
 
-DEFINE_PER_CPU(struct s390_idle_data, s390_idle) = {
-	.lock = __SPIN_LOCK_UNLOCKED(s390_idle.lock)
-};
-
-static int s390_idle_enter(void)
-{
-	struct s390_idle_data *idle;
-
-	idle = &__get_cpu_var(s390_idle);
-	spin_lock(&idle->lock);
-	idle->idle_count++;
-	idle->in_idle = 1;
-	idle->idle_enter = get_clock();
-	spin_unlock(&idle->lock);
-	vtime_stop_cpu_timer();
-	return NOTIFY_OK;
-}
-
-void s390_idle_leave(void)
-{
-	struct s390_idle_data *idle;
-
-	vtime_start_cpu_timer();
-	idle = &__get_cpu_var(s390_idle);
-	spin_lock(&idle->lock);
-	idle->idle_time += get_clock() - idle->idle_enter;
-	idle->in_idle = 0;
-	spin_unlock(&idle->lock);
-}
-
 extern void s390_handle_mcck(void);
 /*
  * The idle loop on a S390...
@@ -117,10 +87,6 @@ static void default_idle(void)
 		local_irq_enable();
 		return;
 	}
-	if (s390_idle_enter() == NOTIFY_BAD) {
-		local_irq_enable();
-		return;
-	}
 #ifdef CONFIG_HOTPLUG_CPU
 	if (cpu_is_offline(smp_processor_id())) {
 		preempt_enable_no_resched();
@@ -130,7 +96,6 @@ static void default_idle(void)
 	local_mcck_disable();
 	if (test_thread_flag(TIF_MCCK_PENDING)) {
 		local_mcck_enable();
-		s390_idle_leave();
 		local_irq_enable();
 		s390_handle_mcck();
 		return;
@@ -138,9 +103,9 @@ static void default_idle(void)
 	trace_hardirqs_on();
 	/* Don't trace preempt off for idle. */
 	stop_critical_timings();
-	/* Wait for external, I/O or machine check interrupt. */
-	__load_psw_mask(psw_kernel_bits | PSW_MASK_WAIT |
-			PSW_MASK_IO | PSW_MASK_EXT);
+	/* Stop virtual timer and halt the cpu. */
+	vtime_stop_cpu();
+	/* Reenable preemption tracer. */
 	start_critical_timings();
 }
 
diff --git a/arch/s390/kernel/s390_ext.c b/arch/s390/kernel/s390_ext.c
index e019b419efc6..a0d2d55d7fb3 100644
--- a/arch/s390/kernel/s390_ext.c
+++ b/arch/s390/kernel/s390_ext.c
@@ -119,8 +119,8 @@ void do_extint(struct pt_regs *regs, unsigned short code)
 	struct pt_regs *old_regs;
 
 	old_regs = set_irq_regs(regs);
-	irq_enter();
 	s390_idle_check();
+	irq_enter();
 	if (S390_lowcore.int_clock >= S390_lowcore.clock_comparator)
 		/* Serve timer interrupts first. */
 		clock_comparator_work();
diff --git a/arch/s390/kernel/setup.c b/arch/s390/kernel/setup.c
index b7a1efd5522c..d825f4950e4e 100644
--- a/arch/s390/kernel/setup.c
+++ b/arch/s390/kernel/setup.c
@@ -427,6 +427,8 @@ setup_lowcore(void)
 		/* enable extended save area */
 		__ctl_set_bit(14, 29);
 	}
+#else
+	lc->vdso_per_cpu_data = (unsigned long) &lc->paste[0];
 #endif
 	set_prefix((u32)(unsigned long) lc);
 }
diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c
index 3ed5c7a83c6c..9c0ccb532a45 100644
--- a/arch/s390/kernel/smp.c
+++ b/arch/s390/kernel/smp.c
@@ -47,6 +47,7 @@
 #include <asm/lowcore.h>
 #include <asm/sclp.h>
 #include <asm/cpu.h>
+#include <asm/vdso.h>
 #include "entry.h"
 
 /*
@@ -500,6 +501,9 @@ static int __cpuinit smp_alloc_lowcore(int cpu)
 			goto out;
 		lowcore->extended_save_area_addr = (u32) save_area;
 	}
+#else
+	if (vdso_alloc_per_cpu(cpu, lowcore))
+		goto out;
 #endif
 	lowcore_ptr[cpu] = lowcore;
 	return 0;
@@ -522,6 +526,8 @@ static void smp_free_lowcore(int cpu)
 #ifndef CONFIG_64BIT
 	if (MACHINE_HAS_IEEE)
 		free_page((unsigned long) lowcore->extended_save_area_addr);
+#else
+	vdso_free_per_cpu(cpu, lowcore);
 #endif
 	free_page(lowcore->panic_stack - PAGE_SIZE);
 	free_pages(lowcore->async_stack - ASYNC_SIZE, ASYNC_ORDER);
@@ -664,6 +670,7 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
 	lowcore = (void *) __get_free_pages(GFP_KERNEL | GFP_DMA, lc_order);
 	panic_stack = __get_free_page(GFP_KERNEL);
 	async_stack = __get_free_pages(GFP_KERNEL, ASYNC_ORDER);
+	BUG_ON(!lowcore || !panic_stack || !async_stack);
 #ifndef CONFIG_64BIT
 	if (MACHINE_HAS_IEEE)
 		save_area = get_zeroed_page(GFP_KERNEL);
@@ -677,6 +684,8 @@ void __init smp_prepare_cpus(unsigned int max_cpus)
 #ifndef CONFIG_64BIT
 	if (MACHINE_HAS_IEEE)
 		lowcore->extended_save_area_addr = (u32) save_area;
+#else
+	BUG_ON(vdso_alloc_per_cpu(smp_processor_id(), lowcore));
 #endif
 	set_prefix((u32)(unsigned long) lowcore);
 	local_mcck_enable();
@@ -845,9 +854,11 @@ static ssize_t show_idle_count(struct sys_device *dev,
 	unsigned long long idle_count;
 
 	idle = &per_cpu(s390_idle, dev->id);
-	spin_lock_irq(&idle->lock);
+	spin_lock(&idle->lock);
 	idle_count = idle->idle_count;
-	spin_unlock_irq(&idle->lock);
+	if (idle->idle_enter)
+		idle_count++;
+	spin_unlock(&idle->lock);
 	return sprintf(buf, "%llu\n", idle_count);
 }
 static SYSDEV_ATTR(idle_count, 0444, show_idle_count, NULL);
@@ -856,18 +867,17 @@ static ssize_t show_idle_time(struct sys_device *dev,
 				struct sysdev_attribute *attr, char *buf)
 {
 	struct s390_idle_data *idle;
-	unsigned long long new_time;
+	unsigned long long now, idle_time, idle_enter;
 
 	idle = &per_cpu(s390_idle, dev->id);
-	spin_lock_irq(&idle->lock);
-	if (idle->in_idle) {
-		new_time = get_clock();
-		idle->idle_time += new_time - idle->idle_enter;
-		idle->idle_enter = new_time;
-	}
-	new_time = idle->idle_time;
-	spin_unlock_irq(&idle->lock);
-	return sprintf(buf, "%llu\n", new_time >> 12);
+	spin_lock(&idle->lock);
+	now = get_clock();
+	idle_time = idle->idle_time;
+	idle_enter = idle->idle_enter;
+	if (idle_enter != 0ULL && idle_enter < now)
+		idle_time += now - idle_enter;
+	spin_unlock(&idle->lock);
+	return sprintf(buf, "%llu\n", idle_time >> 12);
 }
 static SYSDEV_ATTR(idle_time_us, 0444, show_idle_time, NULL);
 
diff --git a/arch/s390/kernel/vdso.c b/arch/s390/kernel/vdso.c
index 10a6ccef4412..25a6a82f1c02 100644
--- a/arch/s390/kernel/vdso.c
+++ b/arch/s390/kernel/vdso.c
@@ -31,9 +31,6 @@
 #include <asm/sections.h>
 #include <asm/vdso.h>
 
-/* Max supported size for symbol names */
-#define MAX_SYMNAME	64
-
 #if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
 extern char vdso32_start, vdso32_end;
 static void *vdso32_kbase = &vdso32_start;
@@ -71,6 +68,119 @@ static union {
 struct vdso_data *vdso_data = &vdso_data_store.data;
 
 /*
+ * Setup vdso data page.
+ */
+static void vdso_init_data(struct vdso_data *vd)
+{
+	unsigned int facility_list;
+
+	facility_list = stfl();
+	vd->ectg_available = switch_amode && (facility_list & 1);
+}
+
+#ifdef CONFIG_64BIT
+/*
+ * Setup per cpu vdso data page.
+ */
+static void vdso_init_per_cpu_data(int cpu, struct vdso_per_cpu_data *vpcd)
+{
+}
+
+/*
+ * Allocate/free per cpu vdso data.
+ */
+#ifdef CONFIG_64BIT
+#define SEGMENT_ORDER	2
+#else
+#define SEGMENT_ORDER	1
+#endif
+
+int vdso_alloc_per_cpu(int cpu, struct _lowcore *lowcore)
+{
+	unsigned long segment_table, page_table, page_frame;
+	u32 *psal, *aste;
+	int i;
+
+	lowcore->vdso_per_cpu_data = __LC_PASTE;
+
+	if (!switch_amode || !vdso_enabled)
+		return 0;
+
+	segment_table = __get_free_pages(GFP_KERNEL, SEGMENT_ORDER);
+	page_table = get_zeroed_page(GFP_KERNEL | GFP_DMA);
+	page_frame = get_zeroed_page(GFP_KERNEL);
+	if (!segment_table || !page_table || !page_frame)
+		goto out;
+
+	clear_table((unsigned long *) segment_table, _SEGMENT_ENTRY_EMPTY,
+		    PAGE_SIZE << SEGMENT_ORDER);
+	clear_table((unsigned long *) page_table, _PAGE_TYPE_EMPTY,
+		    256*sizeof(unsigned long));
+
+	*(unsigned long *) segment_table = _SEGMENT_ENTRY + page_table;
+	*(unsigned long *) page_table = _PAGE_RO + page_frame;
+
+	psal = (u32 *) (page_table + 256*sizeof(unsigned long));
+	aste = psal + 32;
+
+	for (i = 4; i < 32; i += 4)
+		psal[i] = 0x80000000;
+
+	lowcore->paste[4] = (u32)(addr_t) psal;
+	psal[0] = 0x20000000;
+	psal[2] = (u32)(addr_t) aste;
+	*(unsigned long *) (aste + 2) = segment_table +
+		_ASCE_TABLE_LENGTH + _ASCE_USER_BITS + _ASCE_TYPE_SEGMENT;
+	aste[4] = (u32)(addr_t) psal;
+	lowcore->vdso_per_cpu_data = page_frame;
+
+	vdso_init_per_cpu_data(cpu, (struct vdso_per_cpu_data *) page_frame);
+	return 0;
+
+out:
+	free_page(page_frame);
+	free_page(page_table);
+	free_pages(segment_table, SEGMENT_ORDER);
+	return -ENOMEM;
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+void vdso_free_per_cpu(int cpu, struct _lowcore *lowcore)
+{
+	unsigned long segment_table, page_table, page_frame;
+	u32 *psal, *aste;
+
+	if (!switch_amode || !vdso_enabled)
+		return;
+
+	psal = (u32 *)(addr_t) lowcore->paste[4];
+	aste = (u32 *)(addr_t) psal[2];
+	segment_table = *(unsigned long *)(aste + 2) & PAGE_MASK;
+	page_table = *(unsigned long *) segment_table;
+	page_frame = *(unsigned long *) page_table;
+
+	free_page(page_frame);
+	free_page(page_table);
+	free_pages(segment_table, SEGMENT_ORDER);
+}
+#endif /* CONFIG_HOTPLUG_CPU */
+
+static void __vdso_init_cr5(void *dummy)
+{
+	unsigned long cr5;
+
+	cr5 = offsetof(struct _lowcore, paste);
+	__ctl_load(cr5, 5, 5);
+}
+
+static void vdso_init_cr5(void)
+{
+	if (switch_amode && vdso_enabled)
+		on_each_cpu(__vdso_init_cr5, NULL, 1);
+}
+#endif /* CONFIG_64BIT */
+
+/*
  * This is called from binfmt_elf, we create the special vma for the
  * vDSO and insert it into the mm struct tree
  */
@@ -172,6 +282,9 @@ static int __init vdso_init(void)
 {
 	int i;
 
+	if (!vdso_enabled)
+		return 0;
+	vdso_init_data(vdso_data);
 #if defined(CONFIG_32BIT) || defined(CONFIG_COMPAT)
 	/* Calculate the size of the 32 bit vDSO */
 	vdso32_pages = ((&vdso32_end - &vdso32_start
@@ -208,6 +321,10 @@ static int __init vdso_init(void)
 	}
 	vdso64_pagelist[vdso64_pages - 1] = virt_to_page(vdso_data);