[PATCH] cputime: microsecond based cputime for s390

This patch adds the architecture magic to replace the jiffies based cputime
with microsecond based cputime and it adds code to calculate involuntary
wait time.  With this patch the numbers reported by top and ps when running
on LPAR or z/VM are finally not junk anymore.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>
Signed-off-by: Andrew Morton <akpm@osdl.org>
Signed-off-by: Linus Torvalds <torvalds@osdl.org>

7 files changed, 378 insertions, 58 deletions

diff --git a/arch/s390/Kconfig b/arch/s390/Kconfig
index 85ecab32be8915..8394c4e5107332 100644
--- a/arch/s390/Kconfig
+++ b/arch/s390/Kconfig
@@ -367,6 +367,13 @@ config VIRT_TIMER
 	  This provides a kernel interface for virtual CPU timers.
 	  Default is disabled.
 
+config VIRT_CPU_ACCOUNTING
+	bool "Base user process accounting on virtual cpu timer"
+	depends on VIRT_TIMER
+	help
+	  Select this option to use CPU timer deltas to do user
+	  process accounting.
+
 config APPLDATA_BASE
 	bool "Linux - VM Monitor Stream, base infrastructure"
 	depends on PROC_FS && VIRT_TIMER=y
diff --git a/arch/s390/kernel/binfmt_elf32.c b/arch/s390/kernel/binfmt_elf32.c
index ed234bf06d1c00..03ba5893f17b7f 100644
--- a/arch/s390/kernel/binfmt_elf32.c
+++ b/arch/s390/kernel/binfmt_elf32.c
@@ -202,9 +202,8 @@ MODULE_AUTHOR("Gerhard Tonn <ton@de.ibm.com>");
 static __inline__ void
 cputime_to_compat_timeval(const cputime_t cputime, struct compat_timeval *value)
 {
-	unsigned long jiffies = cputime_to_jiffies(cputime);
-	value->tv_usec = (jiffies % HZ) * (1000000L / HZ);
-	value->tv_sec = jiffies / HZ;
+	value->tv_usec = cputime % 1000000;
+	value->tv_sec = cputime / 1000000;
 }
 
 #include "../../../fs/binfmt_elf.c"
diff --git a/arch/s390/kernel/entry.S b/arch/s390/kernel/entry.S
index f72f9ec9be55a5..c0e09b33febe6f 100644
--- a/arch/s390/kernel/entry.S
+++ b/arch/s390/kernel/entry.S
@@ -66,6 +66,27 @@ STACK_SIZE  = 1 << STACK_SHIFT
  *    R15 - kernel stack pointer
  */
 
+	.macro  STORE_TIMER lc_offset
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	stpt	\lc_offset
+#endif
+	.endm
+
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	.macro  UPDATE_VTIME lc_from,lc_to,lc_sum
+	lm	%r10,%r11,\lc_from
+	sl	%r10,\lc_to
+	sl	%r11,\lc_to+4
+	bc	3,BASED(0f)
+	sl	%r10,BASED(.Lc_1)
+0:	al	%r10,\lc_sum
+	al	%r11,\lc_sum+4
+	bc	12,BASED(1f)
+	al	%r10,BASED(.Lc_1)
+1:	stm	%r10,%r11,\lc_sum
+	.endm
+#endif
+
 	.macro	SAVE_ALL_BASE savearea
 	stm	%r12,%r15,\savearea
 	l	%r13,__LC_SVC_NEW_PSW+4	# load &system_call to %r13
@@ -118,6 +139,7 @@ STACK_SIZE  = 1 << STACK_SHIFT
 	ni	__LC_RETURN_PSW+1,0xfd	# clear wait state bit
 	.endif
 	lm	%r0,%r15,SP_R0(%r15)	# load gprs 0-15 of user
+	STORE_TIMER __LC_EXIT_TIMER
 	lpsw	__LC_RETURN_PSW		# back to caller
 	.endm
 
@@ -159,9 +181,21 @@ __critical_start:
 
 	.globl  system_call
 system_call:
+	STORE_TIMER __LC_SYNC_ENTER_TIMER
+sysc_saveall:
 	SAVE_ALL_BASE __LC_SAVE_AREA
         SAVE_ALL __LC_SVC_OLD_PSW,__LC_SAVE_AREA,1
 	lh	%r7,0x8a	  # get svc number from lowcore
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+sysc_vtime:
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	bz	BASED(sysc_do_svc)
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_SYNC_ENTER_TIMER,__LC_USER_TIMER
+sysc_stime:
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+sysc_update:
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
+#endif
 sysc_do_svc:
 	l	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	sla	%r7,2             # *4 and test for svc 0
@@ -391,10 +425,19 @@ pgm_check_handler:
  * we just ignore the PER event (FIXME: is there anything we have to do
  * for LPSW?).
  */
+	STORE_TIMER __LC_SYNC_ENTER_TIMER
 	SAVE_ALL_BASE __LC_SAVE_AREA
         tm      __LC_PGM_INT_CODE+1,0x80 # check whether we got a per exception
         bnz     BASED(pgm_per)           # got per exception -> special case
 	SAVE_ALL __LC_PGM_OLD_PSW,__LC_SAVE_AREA,1
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	bz	BASED(pgm_no_vtime)
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_SYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
+pgm_no_vtime:
+#endif
 	l	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
         l       %r3,__LC_PGM_ILC         # load program interruption code
 	la	%r8,0x7f
@@ -425,6 +468,14 @@ pgm_per:
 #
 pgm_per_std:
 	SAVE_ALL __LC_PGM_OLD_PSW,__LC_SAVE_AREA,1
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	bz	BASED(pgm_no_vtime2)
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_SYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
+pgm_no_vtime2:
+#endif
 	l	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	l	%r1,__TI_task(%r9)
 	mvc	__THREAD_per+__PER_atmid(2,%r1),__LC_PER_ATMID
@@ -442,6 +493,14 @@ pgm_per_std:
 #
 pgm_svcper:
 	SAVE_ALL __LC_SVC_OLD_PSW,__LC_SAVE_AREA,1
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	bz	BASED(pgm_no_vtime3)
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_SYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
+pgm_no_vtime3:
+#endif
 	lh	%r7,0x8a		# get svc number from lowcore
 	l	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	l	%r1,__TI_task(%r9)
@@ -458,9 +517,18 @@ pgm_svcper:
 
         .globl io_int_handler
 io_int_handler:
+	STORE_TIMER __LC_ASYNC_ENTER_TIMER
 	stck	__LC_INT_CLOCK
 	SAVE_ALL_BASE __LC_SAVE_AREA+16
         SAVE_ALL __LC_IO_OLD_PSW,__LC_SAVE_AREA+16,0
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	bz	BASED(io_no_vtime)
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_ASYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_ASYNC_ENTER_TIMER
+io_no_vtime:
+#endif
 	l	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
         l       %r1,BASED(.Ldo_IRQ)        # load address of do_IRQ
         la      %r2,SP_PTREGS(%r15) # address of register-save area
@@ -549,9 +617,18 @@ io_sigpending:
 
         .globl  ext_int_handler
 ext_int_handler:
+	STORE_TIMER __LC_ASYNC_ENTER_TIMER
 	stck	__LC_INT_CLOCK
 	SAVE_ALL_BASE __LC_SAVE_AREA+16
         SAVE_ALL __LC_EXT_OLD_PSW,__LC_SAVE_AREA+16,0
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	bz	BASED(ext_no_vtime)
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_ASYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_ASYNC_ENTER_TIMER
+ext_no_vtime:
+#endif
 	l	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	la	%r2,SP_PTREGS(%r15)    # address of register-save area
 	lh	%r3,__LC_EXT_INT_CODE  # get interruption code
@@ -565,8 +642,17 @@ ext_int_handler:
 
         .globl mcck_int_handler
 mcck_int_handler:
+	STORE_TIMER __LC_ASYNC_ENTER_TIMER
 	SAVE_ALL_BASE __LC_SAVE_AREA+32
         SAVE_ALL __LC_MCK_OLD_PSW,__LC_SAVE_AREA+32,0
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	bz	BASED(mcck_no_vtime)
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_ASYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_ASYNC_ENTER_TIMER
+mcck_no_vtime:
+#endif
 	l       %r1,BASED(.Ls390_mcck)
 	basr    %r14,%r1	  # call machine check handler
 mcck_return:
@@ -661,17 +747,47 @@ cleanup_critical:
 	br	%r14
 
 cleanup_system_call:
-	mvc	__LC_RETURN_PSW(4),0(%r12)
-	clc	4(4,%r12),BASED(cleanup_table_system_call)
-	bne	BASED(0f)
+	mvc	__LC_RETURN_PSW(8),0(%r12)
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	clc	__LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn+4)
+	bh	BASED(0f)
+	mvc	__LC_SYNC_ENTER_TIMER(8),__LC_ASYNC_ENTER_TIMER
+0:	clc	__LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn+8)
+	bhe	BASED(cleanup_vtime)
+#endif
+	clc	__LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn)
+	bh	BASED(0f)
 	mvc	__LC_SAVE_AREA(16),__LC_SAVE_AREA+16
 0:	st	%r13,__LC_SAVE_AREA+20
 	SAVE_ALL __LC_SVC_OLD_PSW,__LC_SAVE_AREA,1
 	st	%r15,__LC_SAVE_AREA+28
 	lh	%r7,0x8a
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+cleanup_vtime:
+	clc	__LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn+12)
+	bhe	BASED(cleanup_stime)
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	bz	BASED(cleanup_novtime)
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_SYNC_ENTER_TIMER,__LC_USER_TIMER
+cleanup_stime:
+	clc	__LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn+16)
+	bh	BASED(cleanup_update)
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+cleanup_update:
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
+cleanup_novtime:
+#endif
 	mvc	__LC_RETURN_PSW+4(4),BASED(cleanup_table_system_call+4)
 	la	%r12,__LC_RETURN_PSW
 	br	%r14
+cleanup_system_call_insn:
+	.long	sysc_saveall + 0x80000000
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	.long   system_call + 0x80000000
+	.long   sysc_vtime + 0x80000000
+	.long   sysc_stime + 0x80000000
+	.long   sysc_update + 0x80000000
+#endif
 
 cleanup_sysc_return:
 	mvc	__LC_RETURN_PSW(4),0(%r12)
@@ -680,15 +796,23 @@ cleanup_sysc_return:
 	br	%r14
 
 cleanup_sysc_leave:
-	clc	4(4,%r12),BASED(cleanup_sysc_leave_lpsw)
+	clc	4(4,%r12),BASED(cleanup_sysc_leave_insn)
+	be	BASED(0f)
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	mvc	__LC_EXIT_TIMER(8),__LC_ASYNC_ENTER_TIMER
+	clc	4(4,%r12),BASED(cleanup_sysc_leave_insn+4)
 	be	BASED(0f)
+#endif
 	mvc	__LC_RETURN_PSW(8),SP_PSW(%r15)
 	mvc	__LC_SAVE_AREA+16(16),SP_R12(%r15)
 	lm	%r0,%r11,SP_R0(%r15)
 	l	%r15,SP_R15(%r15)
 0:	la	%r12,__LC_RETURN_PSW
 	br	%r14
-cleanup_sysc_leave_lpsw:
+cleanup_sysc_leave_insn:
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	.long	sysc_leave + 14 + 0x80000000
+#endif
 	.long	sysc_leave + 10 + 0x80000000
 
 /*
@@ -704,6 +828,7 @@ cleanup_sysc_leave_lpsw:
 .L0x028:       .short 0x028
 .L0x030:       .short 0x030
 .L0x038:       .short 0x038
+.Lc_1:         .long  1
 
 /*
  * Symbol constants
diff --git a/arch/s390/kernel/entry64.S b/arch/s390/kernel/entry64.S
index adbe2a5f5f72b0..51527ab8c8f974 100644
--- a/arch/s390/kernel/entry64.S
+++ b/arch/s390/kernel/entry64.S
@@ -58,6 +58,21 @@ _TIF_WORK_INT = (_TIF_SIGPENDING | _TIF_NEED_RESCHED)
 
 #define BASED(name) name-system_call(%r13)
 
+	.macro  STORE_TIMER lc_offset
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	stpt	\lc_offset
+#endif
+	.endm
+
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	.macro  UPDATE_VTIME lc_from,lc_to,lc_sum
+	lg	%r10,\lc_from
+	slg	%r10,\lc_to
+	alg	%r10,\lc_sum
+	stg	%r10,\lc_sum
+	.endm
+#endif
+
 /*
  * Register usage in interrupt handlers:
  *    R9  - pointer to current task structure
@@ -117,6 +132,7 @@ _TIF_WORK_INT = (_TIF_SIGPENDING | _TIF_NEED_RESCHED)
 	ni	__LC_RETURN_PSW+1,0xfd	# clear wait state bit
 	.endif
 	lmg	%r0,%r15,SP_R0(%r15)	# load gprs 0-15 of user
+	STORE_TIMER __LC_EXIT_TIMER
 	lpswe	__LC_RETURN_PSW		# back to caller
 	.endm
 
@@ -156,9 +172,21 @@ __critical_start:
 
 	.globl  system_call
 system_call:
+	STORE_TIMER __LC_SYNC_ENTER_TIMER
+sysc_saveall:
 	SAVE_ALL_BASE __LC_SAVE_AREA
         SAVE_ALL __LC_SVC_OLD_PSW,__LC_SAVE_AREA,1
 	llgh    %r7,__LC_SVC_INT_CODE # get svc number from lowcore
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+sysc_vtime:
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	jz	sysc_do_svc
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_SYNC_ENTER_TIMER,__LC_USER_TIMER
+sysc_stime:
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+sysc_update:
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
+#endif
 sysc_do_svc:
 	lg	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
         slag    %r7,%r7,2         # *4 and test for svc 0
@@ -441,10 +469,19 @@ pgm_check_handler:
  * we just ignore the PER event (FIXME: is there anything we have to do
  * for LPSW?).
  */
+	STORE_TIMER __LC_SYNC_ENTER_TIMER
 	SAVE_ALL_BASE __LC_SAVE_AREA
         tm      __LC_PGM_INT_CODE+1,0x80 # check whether we got a per exception
         jnz     pgm_per                  # got per exception -> special case
 	SAVE_ALL __LC_PGM_OLD_PSW,__LC_SAVE_AREA,1
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	jz	pgm_no_vtime
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_SYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
+pgm_no_vtime:
+#endif
 	lg	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	lgf     %r3,__LC_PGM_ILC	 # load program interruption code
 	lghi	%r8,0x7f
@@ -475,6 +512,14 @@ pgm_per:
 #
 pgm_per_std:
 	SAVE_ALL __LC_PGM_OLD_PSW,__LC_SAVE_AREA,1
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	jz	pgm_no_vtime2
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_SYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
+pgm_no_vtime2:
+#endif
 	lg	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	lg	%r1,__TI_task(%r9)
 	mvc	__THREAD_per+__PER_atmid(2,%r1),__LC_PER_ATMID
@@ -492,6 +537,14 @@ pgm_per_std:
 #
 pgm_svcper:
 	SAVE_ALL __LC_SVC_OLD_PSW,__LC_SAVE_AREA,1
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	jz	pgm_no_vtime3
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_SYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
+pgm_no_vtime3:
+#endif
 	llgh    %r7,__LC_SVC_INT_CODE	# get svc number from lowcore
 	lg	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	lg	%r1,__TI_task(%r9)
@@ -507,9 +560,18 @@ pgm_svcper:
  */
         .globl io_int_handler
 io_int_handler:
+	STORE_TIMER __LC_ASYNC_ENTER_TIMER
 	stck	__LC_INT_CLOCK
 	SAVE_ALL_BASE __LC_SAVE_AREA+32
         SAVE_ALL __LC_IO_OLD_PSW,__LC_SAVE_AREA+32,0
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	jz	io_no_vtime
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_ASYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_ASYNC_ENTER_TIMER
+io_no_vtime:
+#endif
 	lg	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
         la      %r2,SP_PTREGS(%r15)    # address of register-save area
 	brasl   %r14,do_IRQ            # call standard irq handler
@@ -595,9 +657,18 @@ io_sigpending:
  */
         .globl  ext_int_handler
 ext_int_handler:
+	STORE_TIMER __LC_ASYNC_ENTER_TIMER
 	stck	__LC_INT_CLOCK
 	SAVE_ALL_BASE __LC_SAVE_AREA+32
         SAVE_ALL __LC_EXT_OLD_PSW,__LC_SAVE_AREA+32,0
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	jz	ext_no_vtime
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_ASYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_ASYNC_ENTER_TIMER
+ext_no_vtime:
+#endif
 	lg	%r9,__LC_THREAD_INFO	# load pointer to thread_info struct
 	la	%r2,SP_PTREGS(%r15)    # address of register-save area
 	llgh	%r3,__LC_EXT_INT_CODE  # get interruption code
@@ -609,8 +680,17 @@ ext_int_handler:
  */
         .globl mcck_int_handler
 mcck_int_handler:
+	STORE_TIMER __LC_ASYNC_ENTER_TIMER
 	SAVE_ALL_BASE __LC_SAVE_AREA+64
         SAVE_ALL __LC_MCK_OLD_PSW,__LC_SAVE_AREA+64,0
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	jz	mcck_no_vtime
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_ASYNC_ENTER_TIMER,__LC_USER_TIMER
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_ASYNC_ENTER_TIMER
+mcck_no_vtime:
+#endif
 	brasl   %r14,s390_do_machine_check
 mcck_return:
         RESTORE_ALL 0
@@ -700,17 +780,47 @@ cleanup_critical:
 	br	%r14
 
 cleanup_system_call:
-	mvc	__LC_RETURN_PSW(8),0(%r12)
-	clc	8(8,%r12),BASED(cleanup_table_system_call)
-	jne	0f
+	mvc	__LC_RETURN_PSW(16),0(%r12)
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	clc	__LC_RETURN_PSW+8(8),BASED(cleanup_system_call_insn+8)
+	jh	0f
+	mvc	__LC_SYNC_ENTER_TIMER(8),__LC_ASYNC_ENTER_TIMER
+0:	clc	__LC_RETURN_PSW+8(8),BASED(cleanup_system_call_insn+16)
+	jhe	cleanup_vtime
+#endif
+	clc	__LC_RETURN_PSW+8(8),BASED(cleanup_system_call_insn)
+	jh	0f
 	mvc	__LC_SAVE_AREA(32),__LC_SAVE_AREA+32
 0:	stg	%r13,__LC_SAVE_AREA+40
 	SAVE_ALL __LC_SVC_OLD_PSW,__LC_SAVE_AREA,1
 	stg	%r15,__LC_SAVE_AREA+56
 	llgh	%r7,__LC_SVC_INT_CODE
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+cleanup_vtime:
+	clc	__LC_RETURN_PSW+8(8),BASED(cleanup_system_call_insn+24)
+	jhe	cleanup_stime
+	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
+	jz	cleanup_novtime
+	UPDATE_VTIME __LC_EXIT_TIMER,__LC_SYNC_ENTER_TIMER,__LC_USER_TIMER
+cleanup_stime:
+	clc	__LC_RETURN_PSW+8(8),BASED(cleanup_system_call_insn+32)
+	jh	cleanup_update
+	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
+cleanup_update:
+	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
+cleanup_novtime:
+#endif
 	mvc	__LC_RETURN_PSW+8(8),BASED(cleanup_table_system_call+8)
 	la	%r12,__LC_RETURN_PSW
 	br	%r14
+cleanup_system_call_insn:
+	.quad	sysc_saveall
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	.quad   system_call
+	.quad   sysc_vtime
+	.quad   sysc_stime
+	.quad   sysc_update
+#endif
 
 cleanup_sysc_return:
 	mvc	__LC_RETURN_PSW(8),0(%r12)
@@ -719,15 +829,23 @@ cleanup_sysc_return:
 	br	%r14
 
 cleanup_sysc_leave:
-	clc	8(8,%r12),BASED(cleanup_sysc_leave_lpsw)
+	clc	8(8,%r12),BASED(cleanup_sysc_leave_insn)
+	je	0f
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	mvc	__LC_EXIT_TIMER(8),__LC_ASYNC_ENTER_TIMER
+	clc	8(8,%r12),BASED(cleanup_sysc_leave_insn+8)
 	je	0f
+#endif
 	mvc	__LC_RETURN_PSW(16),SP_PSW(%r15)
 	mvc	__LC_SAVE_AREA+32(32),SP_R12(%r15)
 	lmg	%r0,%r11,SP_R0(%r15)
 	lg	%r15,SP_R15(%r15)
 0:	la	%r12,__LC_RETURN_PSW
 	br	%r14
-cleanup_sysc_leave_lpsw:
+cleanup_sysc_leave_insn:
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	.quad	sysc_leave + 16
+#endif
 	.quad	sysc_leave + 12
 
 /*
diff --git a/arch/s390/kernel/irq.c b/arch/s390/kernel/irq.c
index 75275fe905b134..59bfceabaebebe 100644
--- a/arch/s390/kernel/irq.c
+++ b/arch/s390/kernel/irq.c
@@ -71,6 +71,10 @@ asmlinkage void do_softirq(void)
 
 	local_irq_save(flags);
 
+	account_system_vtime(current);
+
+	local_bh_disable();
+
 	if (local_softirq_pending()) {
 		/* Get current stack pointer. */
 		asm volatile("la %0,0(15)" : "=a" (old));
@@ -93,6 +97,10 @@ asmlinkage void do_softirq(void)
 			__do_softirq();
 	}
 
+	account_system_vtime(current);
+
+	__local_bh_enable();
+
 	local_irq_restore(flags);
 }
 
diff --git a/arch/s390/kernel/time.c b/arch/s390/kernel/time.c
index 2fea300f4b479c..995e2cd38e773f 100644
--- a/arch/s390/kernel/time.c
+++ b/arch/s390/kernel/time.c
@@ -150,28 +150,6 @@ int do_settimeofday(struct timespec *tv)
 
 EXPORT_SYMBOL(do_settimeofday);
 
-#ifndef CONFIG_ARCH_S390X
-
-static inline __u32
-__calculate_ticks(__u64 elapsed)
-{
-	register_pair rp;
-
-	rp.pair = elapsed >> 1;
-	asm ("dr %0,%1" : "+d" (rp) : "d" (CLK_TICKS_PER_JIFFY >> 1));
-	return rp.subreg.odd;
-}
-
-#else /* CONFIG_ARCH_S390X */
-
-static inline __u32
-__calculate_ticks(__u64 elapsed)
-{
-	return elapsed / CLK_TICKS_PER_JIFFY;
-}
-
-#endif /* CONFIG_ARCH_S390X */
-
 
 #ifdef CONFIG_PROFILING
 #define s390_do_profile(regs)	profile_tick(CPU_PROFILING, regs)
@@ -187,14 +165,14 @@ __calculate_ticks(__u64 elapsed)
 void account_ticks(struct pt_regs *regs)
 {
 	__u64 tmp;
-	__u32 ticks;
+	__u32 ticks, xticks;
 
 	/* Calculate how many ticks have passed. */
 	if (S390_lowcore.int_clock < S390_lowcore.jiffy_timer)
 		return;
 	tmp = S390_lowcore.int_clock - S390_lowcore.jiffy_timer;
 	if (tmp >= 2*CLK_TICKS_PER_JIFFY) {  /* more than two ticks ? */
-		ticks = __calculate_ticks(tmp) + 1;
+		ticks = __div(tmp, CLK_TICKS_PER_JIFFY) + 1;
 		S390_lowcore.jiffy_timer +=
 			CLK_TICKS_PER_JIFFY * (__u64) ticks;
 	} else if (tmp >= CLK_TICKS_PER_JIFFY) {
@@ -216,11 +194,9 @@ void account_ticks(struct pt_regs *regs)
 	 */
 	write_seqlock(&xtime_lock);
 	if (S390_lowcore.jiffy_timer > xtime_cc) {
-		__u32 xticks;
-
 		tmp = S390_lowcore.jiffy_timer - xtime_cc;
 		if (tmp >= 2*CLK_TICKS_PER_JIFFY) {
-			xticks = __calculate_ticks(tmp);
+			xticks = __div(tmp, CLK_TICKS_PER_JIFFY);
 			xtime_cc += (__u64) xticks * CLK_TICKS_PER_JIFFY;
 		} else {
 			xticks = 1;
@@ -230,14 +206,18 @@ void account_ticks(struct pt_regs *regs)
 			do_timer(regs);
 	}
 	write_sequnlock(&xtime_lock);
-	while (ticks--)
-		update_process_times(user_mode(regs));
 #else
-	while (ticks--) {
+	for (xticks = ticks; xticks > 0; xticks--)
 		do_timer(regs);
+#endif
+
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+	account_user_vtime(current);
+#else
+	while (ticks--)
 		update_process_times(user_mode(regs));
-	}
 #endif
+
 	s390_do_profile(regs);
 }
 
diff --git a/arch/s390/kernel/vtime.c b/arch/s390/kernel/vtime.c
index 02d2179e40828a..63cdfec3ba99b7 100644
--- a/arch/s390/kernel/vtime.c
+++ b/arch/s390/kernel/vtime.c
@@ -17,6 +17,8 @@
 #include <linux/types.h>
 #include <linux/timex.h>
 #include <linux/notifier.h>
+#include <linux/kernel_stat.h>
+#include <linux/rcupdate.h>
 
 #include <asm/s390_ext.h>
 #include <asm/timer.h>
@@ -25,7 +27,95 @@
 static ext_int_info_t ext_int_info_timer;
 DEFINE_PER_CPU(struct vtimer_queue, virt_cpu_timer);
 
-void start_cpu_timer(void)
+#ifdef CONFIG_VIRT_CPU_ACCOUNTING
+/*
+ * Update process times based on virtual cpu times stored by entry.S
+ * to the lowcore fields user_timer, system_timer & steal_clock.
+ */
+void account_user_vtime(struct task_struct *tsk)
+{
+	cputime_t cputime;
+	__u64 timer, clock;
+	int rcu_user_flag;
+
+	timer = S390_lowcore.last_update_timer;
+	clock = S390_lowcore.last_update_clock;
+	asm volatile ("  STPT %0\n"    /* Store current cpu timer value */
+		      "  STCK %1"      /* Store current tod clock value */
+		      : "=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;
+
+	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 =  S390_lowcore.system_timer >> 12;
+	S390_lowcore.system_timer -= cputime << 12;
+	S390_lowcore.steal_clock -= cputime << 12;
+	account_system_time(tsk, HARDIRQ_OFFSET, cputime);
+
+	cputime = S390_lowcore.steal_clock;
+	if ((__s64) cputime > 0) {
+		cputime >>= 12;
+		S390_lowcore.steal_clock -= cputime << 12;
+		account_steal_time(tsk, cputime);
+	}
+
+	run_local_timers();
+	if (rcu_pending(smp_processor_id()))
+		rcu_check_callbacks(smp_processor_id(), rcu_user_flag);
+	scheduler_tick();
+}
+
+/*
+ * Update process times based on virtual cpu times stored by entry.S
+ * to the lowcore fields user_timer, system_timer & steal_clock.
+ */
+void account_system_vtime(struct task_struct *tsk)
+{
+	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.system_timer >> 12;
+	S390_lowcore.system_timer -= cputime << 12;
+	S390_lowcore.steal_clock -= cputime << 12;
+	account_system_time(tsk, 0, cputime);
+}
+
+static inline void set_vtimer(__u64 expires)
+{
+	__u64 timer;
+
+	asm volatile ("  STPT %0\n"  /* Store current cpu timer value */
+		      "  SPT %1"     /* Set new value immediatly afterwards */
+		      : "=m" (timer) : "m" (expires) );
+	S390_lowcore.system_timer += S390_lowcore.last_update_timer - timer;
+	S390_lowcore.last_update_timer = expires;
+
+	/* store expire time for this CPU timer */
+	per_cpu(virt_cpu_timer, smp_processor_id()).to_expire = expires;
+}
+#else
+static inline void set_vtimer(__u64 expires)
+{
+	S390_lowcore.last_update_timer = expires;
+	asm volatile ("SPT %0" : : "m" (S390_lowcore.last_update_timer));
+
+	/* store expire time for this CPU timer */
+	per_cpu(virt_cpu_timer, smp_processor_id()).to_expire = expires;
+}
+#endif
+
+static void start_cpu_timer(void)
 {
 	struct vtimer_queue *vt_list;
 
@@ -33,7 +123,7 @@ void start_cpu_timer(void)
 	set_vtimer(vt_list->idle);
 }
 
-void stop_cpu_timer(void)
+static void stop_cpu_timer(void)
 {
 	__u64 done;
 	struct vtimer_queue *vt_list;
@@ -71,19 +161,11 @@ void stop_cpu_timer(void)
 	set_vtimer(VTIMER_MAX_SLICE);
 }
 
-void set_vtimer(__u64 expires)
-{
-	asm volatile ("SPT %0" : : "m" (expires));
-
-	/* store expire time for this CPU timer */
-	per_cpu(virt_cpu_timer, smp_processor_id()).to_expire = expires;
-}
-
 /*
  * Sorted add to a list. List is linear searched until first bigger
  * element is found.
  */
-void list_add_sorted(struct vtimer_list *timer, struct list_head *head)
+static void list_add_sorted(struct vtimer_list *timer, struct list_head *head)
 {
 	struct vtimer_list *event;
 
@@ -429,11 +511,12 @@ void init_cpu_vtimer(void)
 {
 	struct vtimer_queue *vt_list;
 	unsigned long cr0;
-	__u64 timer;
 
 	/* kick the virtual timer */
-	timer = VTIMER_MAX_SLICE;
-	asm volatile ("SPT %0" : : "m" (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));
 	__ctl_store(cr0, 0, 0);
 	cr0 |= 0x400;
 	__ctl_load(cr0, 0, 0);