[S390] entry[64].S improvements

Another round of cleanup for entry[64].S, in particular the program check
handler looks more reasonable now. The code size for the 31 bit kernel
has been reduced by 616 byte and by 528 byte for the 64 bit version.
Even better the code is a bit faster as well.

Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

8 files changed, 1002 insertions, 1237 deletions

diff --git a/arch/s390/include/asm/lowcore.h b/arch/s390/include/asm/lowcore.h
index 3b97964e0e96..707f2306725b 100644
--- a/arch/s390/include/asm/lowcore.h
+++ b/arch/s390/include/asm/lowcore.h
@@ -97,47 +97,52 @@ struct _lowcore {
 	__u32	gpregs_save_area[16];		/* 0x0180 */
 	__u32	cregs_save_area[16];		/* 0x01c0 */
 
+	/* Save areas. */
+	__u32	save_area_sync[8];		/* 0x0200 */
+	__u32	save_area_async[8];		/* 0x0220 */
+	__u32	save_area_restart[1];		/* 0x0240 */
+	__u8	pad_0x0244[0x0248-0x0244];	/* 0x0244 */
+
 	/* Return psws. */
-	__u32	save_area[16];			/* 0x0200 */
-	psw_t	return_psw;			/* 0x0240 */
-	psw_t	return_mcck_psw;		/* 0x0248 */
+	psw_t	return_psw;			/* 0x0248 */
+	psw_t	return_mcck_psw;		/* 0x0250 */
 
 	/* CPU time accounting values */
-	__u64	sync_enter_timer;		/* 0x0250 */
-	__u64	async_enter_timer;		/* 0x0258 */
-	__u64	mcck_enter_timer;		/* 0x0260 */
-	__u64	exit_timer;			/* 0x0268 */
-	__u64	user_timer;			/* 0x0270 */
-	__u64	system_timer;			/* 0x0278 */
-	__u64	steal_timer;			/* 0x0280 */
-	__u64	last_update_timer;		/* 0x0288 */
-	__u64	last_update_clock;		/* 0x0290 */
+	__u64	sync_enter_timer;		/* 0x0258 */
+	__u64	async_enter_timer;		/* 0x0260 */
+	__u64	mcck_enter_timer;		/* 0x0268 */
+	__u64	exit_timer;			/* 0x0270 */
+	__u64	user_timer;			/* 0x0278 */
+	__u64	system_timer;			/* 0x0280 */
+	__u64	steal_timer;			/* 0x0288 */
+	__u64	last_update_timer;		/* 0x0290 */
+	__u64	last_update_clock;		/* 0x0298 */
 
 	/* Current process. */
-	__u32	current_task;			/* 0x0298 */
-	__u32	thread_info;			/* 0x029c */
-	__u32	kernel_stack;			/* 0x02a0 */
+	__u32	current_task;			/* 0x02a0 */
+	__u32	thread_info;			/* 0x02a4 */
+	__u32	kernel_stack;			/* 0x02a8 */
 
 	/* Interrupt and panic stack. */
-	__u32	async_stack;			/* 0x02a4 */
-	__u32	panic_stack;			/* 0x02a8 */
+	__u32	async_stack;			/* 0x02ac */
+	__u32	panic_stack;			/* 0x02b0 */
 
 	/* Address space pointer. */
-	__u32	kernel_asce;			/* 0x02ac */
-	__u32	user_asce;			/* 0x02b0 */
-	__u32	current_pid;			/* 0x02b4 */
+	__u32	kernel_asce;			/* 0x02b4 */
+	__u32	user_asce;			/* 0x02b8 */
+	__u32	current_pid;			/* 0x02bc */
 
 	/* SMP info area */
-	__u32	cpu_nr;				/* 0x02b8 */
-	__u32	softirq_pending;		/* 0x02bc */
-	__u32	percpu_offset;			/* 0x02c0 */
-	__u32	ext_call_fast;			/* 0x02c4 */
-	__u64	int_clock;			/* 0x02c8 */
-	__u64	mcck_clock;			/* 0x02d0 */
-	__u64	clock_comparator;		/* 0x02d8 */
-	__u32	machine_flags;			/* 0x02e0 */
-	__u32	ftrace_func;			/* 0x02e4 */
-	__u8	pad_0x02e8[0x0300-0x02e8];	/* 0x02e8 */
+	__u32	cpu_nr;				/* 0x02c0 */
+	__u32	softirq_pending;		/* 0x02c4 */
+	__u32	percpu_offset;			/* 0x02c8 */
+	__u32	ext_call_fast;			/* 0x02cc */
+	__u64	int_clock;			/* 0x02d0 */
+	__u64	mcck_clock;			/* 0x02d8 */
+	__u64	clock_comparator;		/* 0x02e0 */
+	__u32	machine_flags;			/* 0x02e8 */
+	__u32	ftrace_func;			/* 0x02ec */
+	__u8	pad_0x02f8[0x0300-0x02f0];	/* 0x02f0 */
 
 	/* Interrupt response block */
 	__u8	irb[64];			/* 0x0300 */
@@ -229,57 +234,62 @@ struct _lowcore {
 	psw_t	mcck_new_psw;			/* 0x01e0 */
 	psw_t	io_new_psw;			/* 0x01f0 */
 
-	/* Entry/exit save area & return psws. */
-	__u64	save_area[16];			/* 0x0200 */
-	psw_t	return_psw;			/* 0x0280 */
-	psw_t	return_mcck_psw;		/* 0x0290 */
+	/* Save areas. */
+	__u64	save_area_sync[8];		/* 0x0200 */
+	__u64	save_area_async[8];		/* 0x0240 */
+	__u64	save_area_restart[1];		/* 0x0280 */
+	__u8	pad_0x0288[0x0290-0x0288];	/* 0x0288 */
+
+	/* Return psws. */
+	psw_t	return_psw;			/* 0x0290 */
+	psw_t	return_mcck_psw;		/* 0x02a0 */
 
 	/* CPU accounting and timing values. */
-	__u64	sync_enter_timer;		/* 0x02a0 */
-	__u64	async_enter_timer;		/* 0x02a8 */
-	__u64	mcck_enter_timer;		/* 0x02b0 */
-	__u64	exit_timer;			/* 0x02b8 */
-	__u64	user_timer;			/* 0x02c0 */
-	__u64	system_timer;			/* 0x02c8 */
-	__u64	steal_timer;			/* 0x02d0 */
-	__u64	last_update_timer;		/* 0x02d8 */
-	__u64	last_update_clock;		/* 0x02e0 */
+	__u64	sync_enter_timer;		/* 0x02b0 */
+	__u64	async_enter_timer;		/* 0x02b8 */
+	__u64	mcck_enter_timer;		/* 0x02c0 */
+	__u64	exit_timer;			/* 0x02c8 */
+	__u64	user_timer;			/* 0x02d0 */
+	__u64	system_timer;			/* 0x02d8 */
+	__u64	steal_timer;			/* 0x02e0 */
+	__u64	last_update_timer;		/* 0x02e8 */
+	__u64	last_update_clock;		/* 0x02f0 */
 
 	/* Current process. */
-	__u64	current_task;			/* 0x02e8 */
-	__u64	thread_info;			/* 0x02f0 */
-	__u64	kernel_stack;			/* 0x02f8 */
+	__u64	current_task;			/* 0x02f8 */
+	__u64	thread_info;			/* 0x0300 */
+	__u64	kernel_stack;			/* 0x0308 */
 
 	/* Interrupt and panic stack. */
-	__u64	async_stack;			/* 0x0300 */
-	__u64	panic_stack;			/* 0x0308 */
+	__u64	async_stack;			/* 0x0310 */
+	__u64	panic_stack;			/* 0x0318 */
 
 	/* Address space pointer. */
-	__u64	kernel_asce;			/* 0x0310 */
-	__u64	user_asce;			/* 0x0318 */
-	__u64	current_pid;			/* 0x0320 */
+	__u64	kernel_asce;			/* 0x0320 */
+	__u64	user_asce;			/* 0x0328 */
+	__u64	current_pid;			/* 0x0330 */
 
 	/* SMP info area */
-	__u32	cpu_nr;				/* 0x0328 */
-	__u32	softirq_pending;		/* 0x032c */
-	__u64	percpu_offset;			/* 0x0330 */
-	__u64	ext_call_fast;			/* 0x0338 */
-	__u64	int_clock;			/* 0x0340 */
-	__u64	mcck_clock;			/* 0x0348 */
-	__u64	clock_comparator;		/* 0x0350 */
-	__u64	vdso_per_cpu_data;		/* 0x0358 */
-	__u64	machine_flags;			/* 0x0360 */
-	__u64	ftrace_func;			/* 0x0368 */
-	__u64	gmap;				/* 0x0370 */
-	__u8	pad_0x0378[0x0380-0x0378];	/* 0x0378 */
+	__u32	cpu_nr;				/* 0x0338 */
+	__u32	softirq_pending;		/* 0x033c */
+	__u64	percpu_offset;			/* 0x0340 */
+	__u64	ext_call_fast;			/* 0x0348 */
+	__u64	int_clock;			/* 0x0350 */
+	__u64	mcck_clock;			/* 0x0358 */
+	__u64	clock_comparator;		/* 0x0360 */
+	__u64	vdso_per_cpu_data;		/* 0x0368 */
+	__u64	machine_flags;			/* 0x0370 */
+	__u64	ftrace_func;			/* 0x0378 */
+	__u64	gmap;				/* 0x0380 */
+	__u8	pad_0x0388[0x0400-0x0388];	/* 0x0388 */
 
 	/* Interrupt response block. */
-	__u8	irb[64];			/* 0x0380 */
+	__u8	irb[64];			/* 0x0400 */
 
 	/* Per cpu primary space access list */
-	__u32	paste[16];			/* 0x03c0 */
+	__u32	paste[16];			/* 0x0440 */
 
-	__u8	pad_0x0400[0x0e00-0x0400];	/* 0x0400 */
+	__u8	pad_0x0480[0x0e00-0x0480];	/* 0x0480 */
 
 	/*
 	 * 0xe00 contains the address of the IPL Parameter Information
diff --git a/arch/s390/kernel/asm-offsets.c b/arch/s390/kernel/asm-offsets.c
index 0717363033eb..c1a56ba5f848 100644
--- a/arch/s390/kernel/asm-offsets.c
+++ b/arch/s390/kernel/asm-offsets.c
@@ -108,7 +108,9 @@ int main(void)
 	DEFINE(__LC_PGM_NEW_PSW, offsetof(struct _lowcore, program_new_psw));
 	DEFINE(__LC_MCK_NEW_PSW, offsetof(struct _lowcore, mcck_new_psw));
 	DEFINE(__LC_IO_NEW_PSW, offsetof(struct _lowcore, io_new_psw));
-	DEFINE(__LC_SAVE_AREA, offsetof(struct _lowcore, save_area));
+	DEFINE(__LC_SAVE_AREA_SYNC, offsetof(struct _lowcore, save_area_sync));
+	DEFINE(__LC_SAVE_AREA_ASYNC, offsetof(struct _lowcore, save_area_async));
+	DEFINE(__LC_SAVE_AREA_RESTART, offsetof(struct _lowcore, save_area_restart));
 	DEFINE(__LC_RETURN_PSW, offsetof(struct _lowcore, return_psw));
 	DEFINE(__LC_RETURN_MCCK_PSW, offsetof(struct _lowcore, return_mcck_psw));
 	DEFINE(__LC_SYNC_ENTER_TIMER, offsetof(struct _lowcore, sync_enter_timer));
diff --git a/arch/s390/kernel/base.S b/arch/s390/kernel/base.S
index f8828d38fa6e..3aa4d00aaf50 100644
--- a/arch/s390/kernel/base.S
+++ b/arch/s390/kernel/base.S
@@ -33,7 +33,7 @@ s390_base_mcck_handler_fn:
 	.previous
 
 ENTRY(s390_base_ext_handler)
-	stmg	%r0,%r15,__LC_SAVE_AREA
+	stmg	%r0,%r15,__LC_SAVE_AREA_ASYNC
 	basr	%r13,0
 0:	aghi	%r15,-STACK_FRAME_OVERHEAD
 	larl	%r1,s390_base_ext_handler_fn
@@ -41,7 +41,7 @@ ENTRY(s390_base_ext_handler)
 	ltgr	%r1,%r1
 	jz	1f
 	basr	%r14,%r1
-1:	lmg	%r0,%r15,__LC_SAVE_AREA
+1:	lmg	%r0,%r15,__LC_SAVE_AREA_ASYNC
 	ni	__LC_EXT_OLD_PSW+1,0xfd	# clear wait state bit
 	lpswe	__LC_EXT_OLD_PSW
 
@@ -53,7 +53,7 @@ s390_base_ext_handler_fn:
 	.previous
 
 ENTRY(s390_base_pgm_handler)
-	stmg	%r0,%r15,__LC_SAVE_AREA
+	stmg	%r0,%r15,__LC_SAVE_AREA_SYNC
 	basr	%r13,0
 0:	aghi	%r15,-STACK_FRAME_OVERHEAD
 	larl	%r1,s390_base_pgm_handler_fn
@@ -61,7 +61,7 @@ ENTRY(s390_base_pgm_handler)
 	ltgr	%r1,%r1
 	jz	1f
 	basr	%r14,%r1
-	lmg	%r0,%r15,__LC_SAVE_AREA
+	lmg	%r0,%r15,__LC_SAVE_AREA_SYNC
 	lpswe	__LC_PGM_OLD_PSW
 1:	lpswe	disabled_wait_psw-0b(%r13)
 
@@ -142,7 +142,7 @@ s390_base_mcck_handler_fn:
 	.previous
 
 ENTRY(s390_base_ext_handler)
-	stm	%r0,%r15,__LC_SAVE_AREA
+	stm	%r0,%r15,__LC_SAVE_AREA_ASYNC
 	basr	%r13,0
 0:	ahi	%r15,-STACK_FRAME_OVERHEAD
 	l	%r1,2f-0b(%r13)
@@ -150,7 +150,7 @@ ENTRY(s390_base_ext_handler)
 	ltr	%r1,%r1
 	jz	1f
 	basr	%r14,%r1
-1:	lm	%r0,%r15,__LC_SAVE_AREA
+1:	lm	%r0,%r15,__LC_SAVE_AREA_ASYNC
 	ni	__LC_EXT_OLD_PSW+1,0xfd	# clear wait state bit
 	lpsw	__LC_EXT_OLD_PSW
 
@@ -164,7 +164,7 @@ s390_base_ext_handler_fn:
 	.previous
 
 ENTRY(s390_base_pgm_handler)
-	stm	%r0,%r15,__LC_SAVE_AREA
+	stm	%r0,%r15,__LC_SAVE_AREA_SYNC
 	basr	%r13,0
 0:	ahi	%r15,-STACK_FRAME_OVERHEAD
 	l	%r1,2f-0b(%r13)
@@ -172,7 +172,7 @@ ENTRY(s390_base_pgm_handler)
 	ltr	%r1,%r1
 	jz	1f
 	basr	%r14,%r1
-	lm	%r0,%r15,__LC_SAVE_AREA
+	lm	%r0,%r15,__LC_SAVE_AREA_SYNC
 	lpsw	__LC_PGM_OLD_PSW
 
 1:	lpsw	disabled_wait_psw-0b(%r13)
diff --git a/arch/s390/kernel/entry.S b/arch/s390/kernel/entry.S
index b13157057e02..c2773cff89c3 100644
--- a/arch/s390/kernel/entry.S
+++ b/arch/s390/kernel/entry.S
@@ -19,32 +19,22 @@
 #include <asm/unistd.h>
 #include <asm/page.h>
 
-/*
- * Stack layout for the system_call stack entry.
- * The first few entries are identical to the user_regs_struct.
- */
-SP_PTREGS    =	STACK_FRAME_OVERHEAD
-SP_ARGS      =	STACK_FRAME_OVERHEAD + __PT_ARGS
-SP_PSW	     =	STACK_FRAME_OVERHEAD + __PT_PSW
-SP_R0	     =	STACK_FRAME_OVERHEAD + __PT_GPRS
-SP_R1	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 4
-SP_R2	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 8
-SP_R3	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 12
-SP_R4	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 16
-SP_R5	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 20
-SP_R6	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 24
-SP_R7	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 28
-SP_R8	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 32
-SP_R9	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 36
-SP_R10	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 40
-SP_R11	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 44
-SP_R12	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 48
-SP_R13	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 52
-SP_R14	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 56
-SP_R15	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 60
-SP_ORIG_R2   =	STACK_FRAME_OVERHEAD + __PT_ORIG_GPR2
-SP_SVC_CODE  =	STACK_FRAME_OVERHEAD + __PT_SVC_CODE
-SP_SIZE      =	STACK_FRAME_OVERHEAD + __PT_SIZE
+__PT_R0      =	__PT_GPRS
+__PT_R1      =	__PT_GPRS + 4
+__PT_R2      =	__PT_GPRS + 8
+__PT_R3      =	__PT_GPRS + 12
+__PT_R4      =	__PT_GPRS + 16
+__PT_R5      =	__PT_GPRS + 20
+__PT_R6      =	__PT_GPRS + 24
+__PT_R7      =	__PT_GPRS + 28
+__PT_R8      =	__PT_GPRS + 32
+__PT_R9      =	__PT_GPRS + 36
+__PT_R10     =	__PT_GPRS + 40
+__PT_R11     =	__PT_GPRS + 44
+__PT_R12     =	__PT_GPRS + 48
+__PT_R13     =	__PT_GPRS + 524
+__PT_R14     =	__PT_GPRS + 56
+__PT_R15     =	__PT_GPRS + 60
 
 _TIF_WORK_SVC = (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_NEED_RESCHED | \
 		 _TIF_MCCK_PENDING | _TIF_PER_TRAP )
@@ -58,133 +48,91 @@ STACK_SIZE  = 1 << STACK_SHIFT
 
 #define BASED(name) name-system_call(%r13)
 
-#ifdef CONFIG_TRACE_IRQFLAGS
 	.macro	TRACE_IRQS_ON
+#ifdef CONFIG_TRACE_IRQFLAGS
 	basr	%r2,%r0
-	l	%r1,BASED(.Ltrace_irq_on_caller)
-	basr	%r14,%r1
+	l	%r1,BASED(.Lhardirqs_on)
+	basr	%r14,%r1		# call trace_hardirqs_on_caller
+#endif
 	.endm
 
 	.macro	TRACE_IRQS_OFF
+#ifdef CONFIG_TRACE_IRQFLAGS
 	basr	%r2,%r0
-	l	%r1,BASED(.Ltrace_irq_off_caller)
-	basr	%r14,%r1
-	.endm
-#else
-#define TRACE_IRQS_ON
-#define TRACE_IRQS_OFF
+	l	%r1,BASED(.Lhardirqs_off)
+	basr	%r14,%r1		# call trace_hardirqs_off_caller
 #endif
+	.endm
 
-#ifdef CONFIG_LOCKDEP
 	.macro	LOCKDEP_SYS_EXIT
-	tm	SP_PSW+1(%r15),0x01	# returning to user ?
-	jz	0f
+#ifdef CONFIG_LOCKDEP
+	tm	__PT_PSW+1(%r11),0x01	# returning to user ?
+	jz	.+10
 	l	%r1,BASED(.Llockdep_sys_exit)
-	basr	%r14,%r1
-0:
-	.endm
-#else
-#define LOCKDEP_SYS_EXIT
+	basr	%r14,%r1		# call lockdep_sys_exit
 #endif
-
-/*
- * Register usage in interrupt handlers:
- *    R9  - pointer to current task structure
- *    R13 - pointer to literal pool
- *    R14 - return register for function calls
- *    R15 - kernel stack pointer
- */
-
-	.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
-
-	.macro	SAVE_ALL_SVC psworg,savearea
-	stm	%r12,%r15,\savearea
-	l	%r13,__LC_SVC_NEW_PSW+4	# load &system_call to %r13
-	l	%r15,__LC_KERNEL_STACK	# problem state -> load ksp
-	s	%r15,BASED(.Lc_spsize)	# make room for registers & psw
-	.endm
-
-	.macro	SAVE_ALL_BASE savearea
-	stm	%r12,%r15,\savearea
-	l	%r13,__LC_SVC_NEW_PSW+4	# load &system_call to %r13
 	.endm
 
-	.macro	SAVE_ALL_PGM psworg,savearea
-	tm	\psworg+1,0x01		# test problem state bit
+	.macro	CHECK_STACK stacksize,savearea
 #ifdef CONFIG_CHECK_STACK
-	bnz	BASED(1f)
-	tml	%r15,STACK_SIZE - CONFIG_STACK_GUARD
-	bnz	BASED(2f)
-	la	%r12,\psworg
-	b	BASED(stack_overflow)
-#else
-	bz	BASED(2f)
+	tml	%r15,\stacksize - CONFIG_STACK_GUARD
+	la	%r14,\savearea
+	jz	stack_overflow
 #endif
-1:	l	%r15,__LC_KERNEL_STACK	# problem state -> load ksp
-2:	s	%r15,BASED(.Lc_spsize)	# make room for registers & psw
 	.endm
 
-	.macro	SAVE_ALL_ASYNC psworg,savearea
-	stm	%r12,%r15,\savearea
-	l	%r13,__LC_SVC_NEW_PSW+4	# load &system_call to %r13
-	la	%r12,\psworg
-	tm	\psworg+1,0x01		# test problem state bit
-	bnz	BASED(1f)		# from user -> load async stack
-	clc	\psworg+4(4),BASED(.Lcritical_end)
-	bhe	BASED(0f)
-	clc	\psworg+4(4),BASED(.Lcritical_start)
-	bl	BASED(0f)
-	l	%r14,BASED(.Lcleanup_critical)
-	basr	%r14,%r14
-	tm	1(%r12),0x01		# retest problem state after cleanup
-	bnz	BASED(1f)
-0:	l	%r14,__LC_ASYNC_STACK	# are we already on the async stack ?
+	.macro	SWITCH_ASYNC savearea,stack,shift
+	tmh	%r8,0x0001		# interrupting from user ?
+	jnz	1f
+	lr	%r14,%r9
+	sl	%r14,BASED(.Lcritical_start)
+	cl	%r14,BASED(.Lcritical_length)
+	jhe	0f
+	la	%r11,\savearea		# inside critical section, do cleanup
+	bras	%r14,cleanup_critical
+	tmh	%r8,0x0001		# retest problem state after cleanup
+	jnz	1f
+0:	l	%r14,\stack		# are we already on the target stack?
 	slr	%r14,%r15
-	sra	%r14,STACK_SHIFT
-#ifdef CONFIG_CHECK_STACK
-	bnz	BASED(1f)
-	tml	%r15,STACK_SIZE - CONFIG_STACK_GUARD
-	bnz	BASED(2f)
-	b	BASED(stack_overflow)
-#else
-	bz	BASED(2f)
-#endif
-1:	l	%r15,__LC_ASYNC_STACK
-2:	s	%r15,BASED(.Lc_spsize)	# make room for registers & psw
+	sra	%r14,\shift
+	jnz	1f
+	CHECK_STACK 1<<\shift,\savearea
+	j	2f
+1:	l	%r15,\stack		# load target stack
+2:	ahi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
 	.endm
 
-	.macro	CREATE_STACK_FRAME savearea
-	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15)
-	st	%r2,SP_ORIG_R2(%r15)	# store original content of gpr 2
-	mvc	SP_R12(16,%r15),\savearea # move %r12-%r15 to stack
-	stm	%r0,%r11,SP_R0(%r15)	# store gprs %r0-%r11 to kernel stack
+	.macro	ADD64 high,low,timer
+	al	\high,\timer
+	al	\low,\timer+4
+	brc	12,.+8
+	ahi	\high,1
 	.endm
 
-	.macro	RESTORE_ALL psworg,sync
-	mvc	\psworg(8),SP_PSW(%r15) # move user PSW to lowcore
-	.if !\sync
-	ni	\psworg+1,0xfd		# clear wait state bit
-	.endif
-	lm	%r0,%r15,SP_R0(%r15)	# load gprs 0-15 of user
-	stpt	__LC_EXIT_TIMER
-	lpsw	\psworg			# back to caller
+	.macro	SUB64 high,low,timer
+	sl	\high,\timer
+	sl	\low,\timer+4
+	brc	3,.+8
+	ahi	\high,-1
+	.endm
+
+	.macro	UPDATE_VTIME high,low,enter_timer
+	lm	\high,\low,__LC_EXIT_TIMER
+	SUB64	\high,\low,\enter_timer
+	ADD64	\high,\low,__LC_USER_TIMER
+	stm	\high,\low,__LC_USER_TIMER
+	lm	\high,\low,__LC_LAST_UPDATE_TIMER
+	SUB64	\high,\low,__LC_EXIT_TIMER
+	ADD64	\high,\low,__LC_SYSTEM_TIMER
+	stm	\high,\low,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),\enter_timer
 	.endm
 
 	.macro REENABLE_IRQS
-	mvc	__SF_EMPTY(1,%r15),SP_PSW(%r15)
-	ni	__SF_EMPTY(%r15),0xbf
-	ssm	__SF_EMPTY(%r15)
+	st	%r8,__LC_RETURN_PSW
+	ni	__LC_RETURN_PSW,0xbf
+	ssm	__LC_RETURN_PSW
 	.endm
 
 	.section .kprobes.text, "ax"
@@ -197,14 +145,13 @@ STACK_SIZE  = 1 << STACK_SHIFT
  *  gpr2 = prev
  */
 ENTRY(__switch_to)
-	basr	%r1,0
-0:	l	%r4,__THREAD_info(%r2)		# get thread_info of prev
+	l	%r4,__THREAD_info(%r2)		# get thread_info of prev
 	l	%r5,__THREAD_info(%r3)		# get thread_info of next
 	tm	__TI_flags+3(%r4),_TIF_MCCK_PENDING # machine check pending?
-	bz	1f-0b(%r1)
+	jz	0f
 	ni	__TI_flags+3(%r4),255-_TIF_MCCK_PENDING	# clear flag in prev
 	oi	__TI_flags+3(%r5),_TIF_MCCK_PENDING	# set it in next
-1:	stm	%r6,%r15,__SF_GPRS(%r15)	# store gprs of prev task
+0:	stm	%r6,%r15,__SF_GPRS(%r15)	# store gprs of prev task
 	st	%r15,__THREAD_ksp(%r2)		# store kernel stack of prev
 	l	%r15,__THREAD_ksp(%r3)		# load kernel stack of next
 	lctl	%c4,%c4,__TASK_pid(%r3)		# load pid to control reg. 4
@@ -224,48 +171,55 @@ __critical_start:
 
 ENTRY(system_call)
 	stpt	__LC_SYNC_ENTER_TIMER
-sysc_saveall:
-	SAVE_ALL_SVC __LC_SVC_OLD_PSW,__LC_SAVE_AREA
-	CREATE_STACK_FRAME __LC_SAVE_AREA
-	l	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	mvc	SP_PSW(8,%r15),__LC_SVC_OLD_PSW
-	mvc	SP_SVC_CODE(4,%r15),__LC_SVC_ILC
-	oi	__TI_flags+3(%r12),_TIF_SYSCALL
+sysc_stm:
+	stm	%r8,%r15,__LC_SAVE_AREA_SYNC
+	l	%r12,__LC_THREAD_INFO
+	l	%r13,__LC_SVC_NEW_PSW+4
+sysc_per:
+	l	%r15,__LC_KERNEL_STACK
+	ahi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)	# pointer to pt_regs
 sysc_vtime:
-	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
+	UPDATE_VTIME %r8,%r9,__LC_SYNC_ENTER_TIMER
+	stm	%r0,%r7,__PT_R0(%r11)
+	mvc	__PT_R8(32,%r11),__LC_SAVE_AREA_SYNC
+	mvc	__PT_PSW(8,%r11),__LC_SVC_OLD_PSW
+	mvc	__PT_SVC_CODE(4,%r11),__LC_SVC_ILC
 sysc_do_svc:
-	xr	%r7,%r7
-	icm	%r7,3,SP_SVC_CODE+2(%r15)# load svc number and test for svc 0
-	bnz	BASED(sysc_nr_ok)	# svc number > 0
+	oi	__TI_flags+3(%r12),_TIF_SYSCALL
+	lh	%r8,__PT_SVC_CODE+2(%r11)
+	sla	%r8,2				# shift and test for svc0
+	jnz	sysc_nr_ok
 	# svc 0: system call number in %r1
 	cl	%r1,BASED(.Lnr_syscalls)
-	bnl	BASED(sysc_nr_ok)
-	sth	%r1,SP_SVC_CODE+2(%r15)
-	lr	%r7,%r1 	  # copy svc number to %r7
+	jnl	sysc_nr_ok
+	sth	%r1,__PT_SVC_CODE+2(%r11)
+	lr	%r8,%r1
+	sla	%r8,2
 sysc_nr_ok:
-	sll	%r7,2		  # svc number *4
-	l	%r10,BASED(.Lsysc_table)
+	l	%r10,BASED(.Lsys_call_table)	# 31 bit system call table
+	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15)
+	st	%r2,__PT_ORIG_GPR2(%r11)
+	st	%r7,STACK_FRAME_OVERHEAD(%r15)
+	l	%r9,0(%r8,%r10)			# get system call addr.
 	tm	__TI_flags+2(%r12),_TIF_TRACE >> 8
-	mvc	SP_ARGS(4,%r15),SP_R7(%r15)
-	l	%r8,0(%r7,%r10)	  # get system call addr.
-	bnz	BASED(sysc_tracesys)
-	basr	%r14,%r8	  # call sys_xxxx
-	st	%r2,SP_R2(%r15)   # store return value (change R2 on stack)
+	jnz	sysc_tracesys
+	basr	%r14,%r9			# call sys_xxxx
+	st	%r2,__PT_R2(%r11)		# store return value
 
 sysc_return:
 	LOCKDEP_SYS_EXIT
 sysc_tif:
-	tm	SP_PSW+1(%r15),0x01	# returning to user ?
-	bno	BASED(sysc_restore)
+	tm	__PT_PSW+1(%r11),0x01		# returning to user ?
+	jno	sysc_restore
 	tm	__TI_flags+3(%r12),_TIF_WORK_SVC
-	bnz	BASED(sysc_work)  # there is work to do (signals etc.)
+	jnz	sysc_work			# check for work
 	ni	__TI_flags+3(%r12),255-_TIF_SYSCALL
 sysc_restore:
-	RESTORE_ALL __LC_RETURN_PSW,1
+	mvc	__LC_RETURN_PSW(8),__PT_PSW(%r11)
+	stpt	__LC_EXIT_TIMER
+	lm	%r0,%r15,__PT_R0(%r11)
+	lpsw	__LC_RETURN_PSW
 sysc_done:
 
 #
@@ -273,16 +227,16 @@ sysc_done:
 #
 sysc_work:
 	tm	__TI_flags+3(%r12),_TIF_MCCK_PENDING
-	bo	BASED(sysc_mcck_pending)
+	jo	sysc_mcck_pending
 	tm	__TI_flags+3(%r12),_TIF_NEED_RESCHED
-	bo	BASED(sysc_reschedule)
+	jo	sysc_reschedule
 	tm	__TI_flags+3(%r12),_TIF_SIGPENDING
-	bo	BASED(sysc_sigpending)
+	jo	sysc_sigpending
 	tm	__TI_flags+3(%r12),_TIF_NOTIFY_RESUME
-	bo	BASED(sysc_notify_resume)
+	jo	sysc_notify_resume
 	tm	__TI_flags+3(%r12),_TIF_PER_TRAP
-	bo	BASED(sysc_singlestep)
-	b	BASED(sysc_return)	# beware of critical section cleanup
+	jo	sysc_singlestep
+	j	sysc_return		# beware of critical section cleanup
 
 #
 # _TIF_NEED_RESCHED is set, call schedule
@@ -290,13 +244,13 @@ sysc_work:
 sysc_reschedule:
 	l	%r1,BASED(.Lschedule)
 	la	%r14,BASED(sysc_return)
-	br	%r1			# call scheduler
+	br	%r1			# call schedule
 
 #
 # _TIF_MCCK_PENDING is set, call handler
 #
 sysc_mcck_pending:
-	l	%r1,BASED(.Ls390_handle_mcck)
+	l	%r1,BASED(.Lhandle_mcck)
 	la	%r14,BASED(sysc_return)
 	br	%r1			# TIF bit will be cleared by handler
 
@@ -305,23 +259,24 @@ sysc_mcck_pending:
 #
 sysc_sigpending:
 	ni	__TI_flags+3(%r12),255-_TIF_PER_TRAP # clear TIF_PER_TRAP
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
+	lr	%r2,%r11		# pass pointer to pt_regs
 	l	%r1,BASED(.Ldo_signal)
 	basr	%r14,%r1		# call do_signal
 	tm	__TI_flags+3(%r12),_TIF_SYSCALL
-	bno	BASED(sysc_return)
-	lm	%r2,%r6,SP_R2(%r15)	# load svc arguments
-	xr	%r7,%r7			# svc 0 returns -ENOSYS
-	clc	SP_SVC_CODE+2(2,%r15),BASED(.Lnr_syscalls+2)
-	bnl	BASED(sysc_nr_ok)	# invalid svc number -> do svc 0
-	icm	%r7,3,SP_SVC_CODE+2(%r15)# load new svc number
-	b	BASED(sysc_nr_ok)	# restart svc
+	jno	sysc_return
+	lm	%r2,%r7,__PT_R2(%r11)	# load svc arguments
+	xr	%r8,%r8			# svc 0 returns -ENOSYS
+	clc	__PT_SVC_CODE+2(2,%r11),BASED(.Lnr_syscalls+2)
+	jnl	sysc_nr_ok		# invalid svc number -> do svc 0
+	lh	%r8,__PT_SVC_CODE+2(%r11)	# load new svc number
+	sla	%r8,2
+	j	sysc_nr_ok		# restart svc
 
 #
 # _TIF_NOTIFY_RESUME is set, call do_notify_resume
 #
 sysc_notify_resume:
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
+	lr	%r2,%r11		# pass pointer to pt_regs
 	l	%r1,BASED(.Ldo_notify_resume)
 	la	%r14,BASED(sysc_return)
 	br	%r1			# call do_notify_resume
@@ -331,56 +286,57 @@ sysc_notify_resume:
 #
 sysc_singlestep:
 	ni	__TI_flags+3(%r12),255-(_TIF_SYSCALL | _TIF_PER_TRAP)
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
-	l	%r1,BASED(.Lhandle_per)	# load adr. of per handler
-	la	%r14,BASED(sysc_return)	# load adr. of system return
-	br	%r1			# branch to do_per_trap
+	lr	%r2,%r11		# pass pointer to pt_regs
+	l	%r1,BASED(.Ldo_per_trap)
+	la	%r14,BASED(sysc_return)
+	br	%r1			# call do_per_trap
 
 #
 # call tracehook_report_syscall_entry/tracehook_report_syscall_exit before
 # and after the system call
 #
 sysc_tracesys:
-	l	%r1,BASED(.Ltrace_entry)
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
+	l	%r1,BASED(.Ltrace_enter)
+	lr	%r2,%r11		# pass pointer to pt_regs
 	la	%r3,0
 	xr	%r0,%r0
-	icm	%r0,3,SP_SVC_CODE(%r15)
-	st	%r0,SP_R2(%r15)
-	basr	%r14,%r1
+	icm	%r0,3,__PT_SVC_CODE+2(%r11)
+	st	%r0,__PT_R2(%r11)
+	basr	%r14,%r1		# call do_syscall_trace_enter
 	cl	%r2,BASED(.Lnr_syscalls)
-	bnl	BASED(sysc_tracenogo)
-	lr	%r7,%r2
-	sll	%r7,2			# svc number *4
-	l	%r8,0(%r7,%r10)
+	jnl	sysc_tracenogo
+	lr	%r8,%r2
+	sll	%r8,2
+	l	%r9,0(%r8,%r10)
 sysc_tracego:
-	lm	%r3,%r6,SP_R3(%r15)
-	mvc	SP_ARGS(4,%r15),SP_R7(%r15)
-	l	%r2,SP_ORIG_R2(%r15)
-	basr	%r14,%r8		# call sys_xxx
-	st	%r2,SP_R2(%r15)		# store return value
+	lm	%r3,%r7,__PT_R3(%r11)
+	st	%r7,STACK_FRAME_OVERHEAD(%r15)
+	l	%r2,__PT_ORIG_GPR2(%r11)
+	basr	%r14,%r9		# call sys_xxx
+	st	%r2,__PT_R2(%r11)	# store return value
 sysc_tracenogo:
 	tm	__TI_flags+2(%r12),_TIF_TRACE >> 8
-	bz	BASED(sysc_return)
+	jz	sysc_return
 	l	%r1,BASED(.Ltrace_exit)
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
+	lr	%r2,%r11		# pass pointer to pt_regs
 	la	%r14,BASED(sysc_return)
-	br	%r1
+	br	%r1			# call do_syscall_trace_exit
 
 #
 # a new process exits the kernel with ret_from_fork
 #
 ENTRY(ret_from_fork)
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+	l	%r12,__LC_THREAD_INFO
 	l	%r13,__LC_SVC_NEW_PSW+4
-	l	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	tm	SP_PSW+1(%r15),0x01	# forking a kernel thread ?
-	bo	BASED(0f)
-	st	%r15,SP_R15(%r15)	# store stack pointer for new kthread
-0:	l	%r1,BASED(.Lschedtail)
-	basr	%r14,%r1
+	tm	__PT_PSW+1(%r11),0x01	# forking a kernel thread ?
+	jo	0f
+	st	%r15,__PT_R15(%r11)	# store stack pointer for new kthread
+0:	l	%r1,BASED(.Lschedule_tail)
+	basr	%r14,%r1		# call schedule_tail
 	TRACE_IRQS_ON
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
-	b	BASED(sysc_tracenogo)
+	ssm	__LC_SVC_NEW_PSW	# reenable interrupts
+	j	sysc_tracenogo
 
 #
 # kernel_execve function needs to deal with pt_regs that is not
@@ -390,153 +346,98 @@ ENTRY(kernel_execve)
 	stm	%r12,%r15,48(%r15)
 	lr	%r14,%r15
 	l	%r13,__LC_SVC_NEW_PSW+4
-	s	%r15,BASED(.Lc_spsize)
+	ahi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
 	st	%r14,__SF_BACKCHAIN(%r15)
-	la	%r12,SP_PTREGS(%r15)
+	la	%r12,STACK_FRAME_OVERHEAD(%r15)
 	xc	0(__PT_SIZE,%r12),0(%r12)
 	l	%r1,BASED(.Ldo_execve)
 	lr	%r5,%r12
-	basr	%r14,%r1
+	basr	%r14,%r1		# call do_execve
 	ltr	%r2,%r2
-	be	BASED(0f)
-	a	%r15,BASED(.Lc_spsize)
+	je	0f
+	ahi	%r15,(STACK_FRAME_OVERHEAD + __PT_SIZE)
 	lm	%r12,%r15,48(%r15)
 	br	%r14
 	# execve succeeded.
-0:	stnsm	__SF_EMPTY(%r15),0xfc	# disable interrupts
+0:	ssm	__LC_PGM_NEW_PSW	# disable I/O and ext. interrupts
 	l	%r15,__LC_KERNEL_STACK	# load ksp
-	s	%r15,BASED(.Lc_spsize)	# make room for registers & psw
-	mvc	SP_PTREGS(__PT_SIZE,%r15),0(%r12)	# copy pt_regs
+	ahi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+	mvc	0(__PT_SIZE,%r11),0(%r12)	# copy pt_regs
 	l	%r12,__LC_THREAD_INFO
 	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15)
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
+	ssm	__LC_SVC_NEW_PSW	# reenable interrupts
 	l	%r1,BASED(.Lexecve_tail)
-	basr	%r14,%r1
-	b	BASED(sysc_return)
+	basr	%r14,%r1		# call execve_tail
+	j	sysc_return
 
 /*
  * Program check handler routine
  */
 
 ENTRY(pgm_check_handler)
-/*
- * First we need to check for a special case:
- * Single stepping an instruction that disables the PER event mask will
- * cause a PER event AFTER the mask has been set. Example: SVC or LPSW.
- * For a single stepped SVC the program check handler gets control after
- * the SVC new PSW has been loaded. But we want to execute the SVC first and
- * then handle the PER event. Therefore we update the SVC old PSW to point
- * to the pgm_check_handler and branch to the SVC handler after we checked
- * if we have to load the kernel stack register.
- * For every other possible cause for PER event without the PER mask set
- * we just ignore the PER event (FIXME: is there anything we have to do
- * for LPSW?).
- */
 	stpt	__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_PGM __LC_PGM_OLD_PSW,__LC_SAVE_AREA
-	CREATE_STACK_FRAME __LC_SAVE_AREA
-	mvc	SP_PSW(8,%r15),__LC_PGM_OLD_PSW
-	l	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	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:
-	l	%r3,__LC_PGM_ILC	# load program interruption code
-	l	%r4,__LC_TRANS_EXC_CODE
-	REENABLE_IRQS
-	la	%r8,0x7f
-	nr	%r8,%r3
-	sll	%r8,2
-	l	%r1,BASED(.Ljump_table)
-	l	%r1,0(%r8,%r1)		# load address of handler routine
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
-	basr	%r14,%r1		# branch to interrupt-handler
-pgm_exit:
-	b	BASED(sysc_return)
-
-#
-# handle per exception
-#
-pgm_per:
-	tm	__LC_PGM_OLD_PSW,0x40	# test if per event recording is on
-	bnz	BASED(pgm_per_std)	# ok, normal per event from user space
-# ok its one of the special cases, now we need to find out which one
-	clc	__LC_PGM_OLD_PSW(8),__LC_SVC_NEW_PSW
-	be	BASED(pgm_svcper)
-# no interesting special case, ignore PER event
-	lm	%r12,%r15,__LC_SAVE_AREA
-	lpsw	0x28
-
-#
-# Normal per exception
-#
-pgm_per_std:
-	SAVE_ALL_PGM __LC_PGM_OLD_PSW,__LC_SAVE_AREA
-	CREATE_STACK_FRAME __LC_SAVE_AREA
-	mvc	SP_PSW(8,%r15),__LC_PGM_OLD_PSW
-	l	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	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:
+	stm	%r8,%r15,__LC_SAVE_AREA_SYNC
+	l	%r12,__LC_THREAD_INFO
+	l	%r13,__LC_SVC_NEW_PSW+4
+	lm	%r8,%r9,__LC_PGM_OLD_PSW
+	tmh	%r8,0x0001		# test problem state bit
+	jnz	1f			# -> fault in user space
+	tmh	%r8,0x4000		# PER bit set in old PSW ?
+	jnz	0f			# -> enabled, can't be a double fault
+	tm	__LC_PGM_ILC+3,0x80	# check for per exception
+	jnz	pgm_svcper		# -> single stepped svc
+0:	CHECK_STACK STACK_SIZE,__LC_SAVE_AREA_SYNC
+	j	2f
+1:	UPDATE_VTIME %r14,%r15,__LC_SYNC_ENTER_TIMER
+	l	%r15,__LC_KERNEL_STACK
+2:	ahi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+	stm	%r0,%r7,__PT_R0(%r11)
+	mvc	__PT_R8(32,%r11),__LC_SAVE_AREA_SYNC
+	stm	%r8,%r9,__PT_PSW(%r11)
+	tm	__LC_PGM_ILC+3,0x80	# check for per exception
+	jz	0f
 	l	%r1,__TI_task(%r12)
-	tm	SP_PSW+1(%r15),0x01	# kernel per event ?
-	bz	BASED(kernel_per)
-	mvc	__THREAD_per_cause(2,%r1),__LC_PER_CAUSE
+	tmh	%r8,0x0001		# kernel per event ?
+	jz	pgm_kprobe
+	oi	__TI_flags+3(%r12),_TIF_PER_TRAP
 	mvc	__THREAD_per_address(4,%r1),__LC_PER_ADDRESS
+	mvc	__THREAD_per_cause(2,%r1),__LC_PER_CAUSE
 	mvc	__THREAD_per_paid(1,%r1),__LC_PER_PAID
-	oi	__TI_flags+3(%r12),_TIF_PER_TRAP # set TIF_PER_TRAP
-	l	%r3,__LC_PGM_ILC	# load program interruption code
+0:	l	%r3,__LC_PGM_ILC	# load program interruption code
 	l	%r4,__LC_TRANS_EXC_CODE
 	REENABLE_IRQS
-	la	%r8,0x7f
-	nr	%r8,%r3 		# clear per-event-bit and ilc
-	be	BASED(pgm_exit2)	# only per or per+check ?
-	sll	%r8,2
+	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15)
 	l	%r1,BASED(.Ljump_table)
-	l	%r1,0(%r8,%r1)		# load address of handler routine
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
+	la	%r10,0x7f
+	nr	%r10,%r3
+	je	sysc_return
+	sll	%r10,2
+	l	%r1,0(%r10,%r1)		# load address of handler routine
+	lr	%r2,%r11		# pass pointer to pt_regs
 	basr	%r14,%r1		# branch to interrupt-handler
-pgm_exit2:
-	b	BASED(sysc_return)
+	j	sysc_return
 
 #
-# it was a single stepped SVC that is causing all the trouble
+# PER event in supervisor state, must be kprobes
 #
-pgm_svcper:
-	SAVE_ALL_PGM __LC_SVC_OLD_PSW,__LC_SAVE_AREA
-	CREATE_STACK_FRAME __LC_SAVE_AREA
-	l	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	mvc	SP_PSW(8,%r15),__LC_SVC_OLD_PSW
-	mvc	SP_SVC_CODE(4,%r15),__LC_SVC_ILC
-	oi	__TI_flags+3(%r12),(_TIF_SYSCALL | _TIF_PER_TRAP)
-	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
-	l	%r8,__TI_task(%r12)
-	mvc	__THREAD_per_cause(2,%r8),__LC_PER_CAUSE
-	mvc	__THREAD_per_address(4,%r8),__LC_PER_ADDRESS
-	mvc	__THREAD_per_paid(1,%r8),__LC_PER_PAID
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
-	lm	%r2,%r6,SP_R2(%r15)	# load svc arguments
-	b	BASED(sysc_do_svc)
+pgm_kprobe:
+	REENABLE_IRQS
+	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15)
+	l	%r1,BASED(.Ldo_per_trap)
+	lr	%r2,%r11		# pass pointer to pt_regs
+	basr	%r14,%r1		# call do_per_trap
+	j	sysc_return
 
 #
-# per was called from kernel, must be kprobes
+# single stepped system call
 #
-kernel_per:
-	REENABLE_IRQS
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
-	l	%r1,BASED(.Lhandle_per)	# load adr. of per handler
-	basr	%r14,%r1		# branch to do_single_step
-	b	BASED(pgm_exit)
+pgm_svcper:
+	oi	__TI_flags+3(%r12),_TIF_PER_TRAP
+	mvc	__LC_RETURN_PSW(4),__LC_SVC_NEW_PSW
+	mvc	__LC_RETURN_PSW+4(4),BASED(.Lsysc_per)
+	lpsw	__LC_RETURN_PSW		# branch to sysc_per and enable irqs
 
 /*
  * IO interrupt handler routine
@@ -545,28 +446,35 @@ kernel_per:
 ENTRY(io_int_handler)
 	stck	__LC_INT_CLOCK
 	stpt	__LC_ASYNC_ENTER_TIMER
-	SAVE_ALL_ASYNC __LC_IO_OLD_PSW,__LC_SAVE_AREA+16
-	CREATE_STACK_FRAME __LC_SAVE_AREA+16
-	mvc	SP_PSW(8,%r15),0(%r12)	# move user PSW to stack
-	l	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	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:
+	stm	%r8,%r15,__LC_SAVE_AREA_ASYNC
+	l	%r12,__LC_THREAD_INFO
+	l	%r13,__LC_SVC_NEW_PSW+4
+	lm	%r8,%r9,__LC_IO_OLD_PSW
+	tmh	%r8,0x0001		# interrupting from user ?
+	jz	io_skip
+	UPDATE_VTIME %r14,%r15,__LC_ASYNC_ENTER_TIMER
+io_skip:
+	SWITCH_ASYNC __LC_SAVE_AREA_ASYNC,__LC_ASYNC_STACK,STACK_SHIFT
+	stm	%r0,%r7,__PT_R0(%r11)
+	mvc	__PT_R8(32,%r11),__LC_SAVE_AREA_ASYNC
+	stm	%r8,%r9,__PT_PSW(%r11)
 	TRACE_IRQS_OFF
-	l	%r1,BASED(.Ldo_IRQ)	# load address of do_IRQ
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
-	basr	%r14,%r1		# branch to standard irq handler
+	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15)
+	l	%r1,BASED(.Ldo_IRQ)
+	lr	%r2,%r11		# pass pointer to pt_regs
+	basr	%r14,%r1		# call do_IRQ
 io_return:
 	LOCKDEP_SYS_EXIT
 	TRACE_IRQS_ON
 io_tif:
 	tm	__TI_flags+3(%r12),_TIF_WORK_INT
-	bnz	BASED(io_work)		# there is work to do (signals etc.)
+	jnz	io_work			# there is work to do (signals etc.)
 io_restore:
-	RESTORE_ALL __LC_RETURN_PSW,0
+	mvc	__LC_RETURN_PSW(8),__PT_PSW(%r11)
+	ni	__LC_RETURN_PSW+1,0xfd	# clean wait state bit
+	stpt	__LC_EXIT_TIMER
+	lm	%r0,%r15,__PT_R0(%r11)
+	lpsw	__LC_RETURN_PSW
 io_done:
 
 #
@@ -577,28 +485,29 @@ io_done:
 # Before any work can be done, a switch to the kernel stack is required.
 #
 io_work:
-	tm	SP_PSW+1(%r15),0x01	# returning to user ?
-	bo	BASED(io_work_user)	# yes -> do resched & signal
+	tm	__PT_PSW+1(%r11),0x01	# returning to user ?
+	jo	io_work_user		# yes -> do resched & signal
 #ifdef CONFIG_PREEMPT
 	# check for preemptive scheduling
 	icm	%r0,15,__TI_precount(%r12)
-	bnz	BASED(io_restore)	# preemption disabled
+	jnz	io_restore		# preemption disabled
 	tm	__TI_flags+3(%r12),_TIF_NEED_RESCHED
-	bno	BASED(io_restore)
+	jno	io_restore
 	# switch to kernel stack
-	l	%r1,SP_R15(%r15)
-	s	%r1,BASED(.Lc_spsize)
-	mvc	SP_PTREGS(__PT_SIZE,%r1),SP_PTREGS(%r15)
-	xc	__SF_BACKCHAIN(4,%r1),__SF_BACKCHAIN(%r1) # clear back chain
+	l	%r1,__PT_R15(%r11)
+	ahi	%r1,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	mvc	STACK_FRAME_OVERHEAD(__PT_SIZE,%r1),0(%r11)
+	xc	__SF_BACKCHAIN(4,%r1),__SF_BACKCHAIN(%r1)
+	la	%r11,STACK_FRAME_OVERHEAD(%r1)
 	lr	%r15,%r1
 	# TRACE_IRQS_ON already done at io_return, call
 	# TRACE_IRQS_OFF to keep things symmetrical
 	TRACE_IRQS_OFF
-	l	%r1,BASED(.Lpreempt_schedule_irq)
+	l	%r1,BASED(.Lpreempt_irq)
 	basr	%r14,%r1		# call preempt_schedule_irq
-	b	BASED(io_return)
+	j	io_return
 #else
-	b	BASED(io_restore)
+	j	io_restore
 #endif
 
 #
@@ -606,9 +515,10 @@ io_work:
 #
 io_work_user:
 	l	%r1,__LC_KERNEL_STACK
-	s	%r1,BASED(.Lc_spsize)
-	mvc	SP_PTREGS(__PT_SIZE,%r1),SP_PTREGS(%r15)
-	xc	__SF_BACKCHAIN(4,%r1),__SF_BACKCHAIN(%r1) # clear back chain
+	ahi	%r1,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	mvc	STACK_FRAME_OVERHEAD(__PT_SIZE,%r1),0(%r11)
+	xc	__SF_BACKCHAIN(4,%r1),__SF_BACKCHAIN(%r1)
+	la	%r11,STACK_FRAME_OVERHEAD(%r1)
 	lr	%r15,%r1
 
 #
@@ -618,24 +528,24 @@ io_work_user:
 #
 io_work_tif:
 	tm	__TI_flags+3(%r12),_TIF_MCCK_PENDING
-	bo	BASED(io_mcck_pending)
+	jo	io_mcck_pending
 	tm	__TI_flags+3(%r12),_TIF_NEED_RESCHED
-	bo	BASED(io_reschedule)
+	jo	io_reschedule
 	tm	__TI_flags+3(%r12),_TIF_SIGPENDING
-	bo	BASED(io_sigpending)
+	jo	io_sigpending
 	tm	__TI_flags+3(%r12),_TIF_NOTIFY_RESUME
-	bo	BASED(io_notify_resume)
-	b	BASED(io_return)	# beware of critical section cleanup
+	jo	io_notify_resume
+	j	io_return		# beware of critical section cleanup
 
 #
 # _TIF_MCCK_PENDING is set, call handler
 #
 io_mcck_pending:
 	# TRACE_IRQS_ON already done at io_return
-	l	%r1,BASED(.Ls390_handle_mcck)
+	l	%r1,BASED(.Lhandle_mcck)
 	basr	%r14,%r1		# TIF bit will be cleared by handler
 	TRACE_IRQS_OFF
-	b	BASED(io_return)
+	j	io_return
 
 #
 # _TIF_NEED_RESCHED is set, call schedule
@@ -643,37 +553,37 @@ io_mcck_pending:
 io_reschedule:
 	# TRACE_IRQS_ON already done at io_return
 	l	%r1,BASED(.Lschedule)
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
+	ssm	__LC_SVC_NEW_PSW	# reenable interrupts
 	basr	%r14,%r1		# call scheduler
-	stnsm	__SF_EMPTY(%r15),0xfc	# disable I/O and ext. interrupts
+	ssm	__LC_PGM_NEW_PSW	# disable I/O and ext. interrupts
 	TRACE_IRQS_OFF
-	b	BASED(io_return)
+	j	io_return
 
 #
 # _TIF_SIGPENDING is set, call do_signal
 #
 io_sigpending:
 	# TRACE_IRQS_ON already done at io_return
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
 	l	%r1,BASED(.Ldo_signal)
+	ssm	__LC_SVC_NEW_PSW	# reenable interrupts
+	lr	%r2,%r11		# pass pointer to pt_regs
 	basr	%r14,%r1		# call do_signal
-	stnsm	__SF_EMPTY(%r15),0xfc	# disable I/O and ext. interrupts
+	ssm	__LC_PGM_NEW_PSW	# disable I/O and ext. interrupts
 	TRACE_IRQS_OFF
-	b	BASED(io_return)
+	j	io_return
 
 #
 # _TIF_SIGPENDING is set, call do_signal
 #
 io_notify_resume:
 	# TRACE_IRQS_ON already done at io_return
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
 	l	%r1,BASED(.Ldo_notify_resume)
-	basr	%r14,%r1		# call do_signal
-	stnsm	__SF_EMPTY(%r15),0xfc	# disable I/O and ext. interrupts
+	ssm	__LC_SVC_NEW_PSW	# reenable interrupts
+	lr	%r2,%r11		# pass pointer to pt_regs
+	basr	%r14,%r1		# call do_notify_resume
+	ssm	__LC_PGM_NEW_PSW	# disable I/O and ext. interrupts
 	TRACE_IRQS_OFF
-	b	BASED(io_return)
+	j	io_return
 
 /*
  * External interrupt handler routine
@@ -682,23 +592,25 @@ io_notify_resume:
 ENTRY(ext_int_handler)
 	stck	__LC_INT_CLOCK
 	stpt	__LC_ASYNC_ENTER_TIMER
-	SAVE_ALL_ASYNC __LC_EXT_OLD_PSW,__LC_SAVE_AREA+16
-	CREATE_STACK_FRAME __LC_SAVE_AREA+16
-	mvc	SP_PSW(8,%r15),0(%r12)	# move user PSW to stack
-	l	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	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:
+	stm	%r8,%r15,__LC_SAVE_AREA_ASYNC
+	l	%r12,__LC_THREAD_INFO
+	l	%r13,__LC_SVC_NEW_PSW+4
+	lm	%r8,%r9,__LC_EXT_OLD_PSW
+	tmh	%r8,0x0001		# interrupting from user ?
+	jz	ext_skip
+	UPDATE_VTIME %r14,%r15,__LC_ASYNC_ENTER_TIMER
+ext_skip:
+	SWITCH_ASYNC __LC_SAVE_AREA_ASYNC,__LC_ASYNC_STACK,STACK_SHIFT
+	stm	%r0,%r7,__PT_R0(%r11)
+	mvc	__PT_R8(32,%r11),__LC_SAVE_AREA_ASYNC
+	stm	%r8,%r9,__PT_PSW(%r11)
 	TRACE_IRQS_OFF
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
+	lr	%r2,%r11		# pass pointer to pt_regs
 	l	%r3,__LC_CPU_ADDRESS	# get cpu address + interruption code
 	l	%r4,__LC_EXT_PARAMS	# get external parameters
 	l	%r1,BASED(.Ldo_extint)
-	basr	%r14,%r1
-	b	BASED(io_return)
+	basr	%r14,%r1		# call do_extint
+	j	io_return
 
 __critical_end:
 
@@ -710,82 +622,74 @@ ENTRY(mcck_int_handler)
 	stck	__LC_MCCK_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
-	la	%r12,__LC_MCK_OLD_PSW
+	l	%r12,__LC_THREAD_INFO
+	l	%r13,__LC_SVC_NEW_PSW+4
+	lm	%r8,%r9,__LC_MCK_OLD_PSW
 	tm	__LC_MCCK_CODE,0x80	# system damage?
-	bo	BASED(mcck_int_main)	# yes -> rest of mcck code invalid
-	mvc	__LC_MCCK_ENTER_TIMER(8),__LC_CPU_TIMER_SAVE_AREA
+	jo	mcck_panic		# yes -> rest of mcck code invalid
+	la	%r14,__LC_CPU_TIMER_SAVE_AREA
+	mvc	__LC_MCCK_ENTER_TIMER(8),0(%r14)
 	tm	__LC_MCCK_CODE+5,0x02	# stored cpu timer value valid?
-	bo	BASED(1f)
+	jo	3f
 	la	%r14,__LC_SYNC_ENTER_TIMER
 	clc	0(8,%r14),__LC_ASYNC_ENTER_TIMER
-	bl	BASED(0f)
+	jl	0f
 	la	%r14,__LC_ASYNC_ENTER_TIMER
 0:	clc	0(8,%r14),__LC_EXIT_TIMER
-	bl	BASED(0f)
+	jl	1f
 	la	%r14,__LC_EXIT_TIMER
-0:	clc	0(8,%r14),__LC_LAST_UPDATE_TIMER
-	bl	BASED(0f)
+1:	clc	0(8,%r14),__LC_LAST_UPDATE_TIMER
+	jl	2f
 	la	%r14,__LC_LAST_UPDATE_TIMER
-0:	spt	0(%r14)
+2:	spt	0(%r14)
 	mvc	__LC_MCCK_ENTER_TIMER(8),0(%r14)
-1:	tm	__LC_MCCK_CODE+2,0x09	# mwp + ia of old psw valid?
-	bno	BASED(mcck_int_main)	# no -> skip cleanup critical
-	tm	__LC_MCK_OLD_PSW+1,0x01	# test problem state bit
-	bnz	BASED(mcck_int_main)	# from user -> load async stack
-	clc	__LC_MCK_OLD_PSW+4(4),BASED(.Lcritical_end)
-	bhe	BASED(mcck_int_main)
-	clc	__LC_MCK_OLD_PSW+4(4),BASED(.Lcritical_start)
-	bl	BASED(mcck_int_main)
-	l	%r14,BASED(.Lcleanup_critical)
-	basr	%r14,%r14
-mcck_int_main:
-	l	%r14,__LC_PANIC_STACK	# are we already on the panic stack?
-	slr	%r14,%r15
-	sra	%r14,PAGE_SHIFT
-	be	BASED(0f)
-	l	%r15,__LC_PANIC_STACK	# load panic stack
-0:	s	%r15,BASED(.Lc_spsize)	# make room for registers & psw
-	CREATE_STACK_FRAME __LC_SAVE_AREA+32
-	mvc	SP_PSW(8,%r15),0(%r12)
-	l	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	tm	__LC_MCCK_CODE+2,0x08	# mwp of old psw valid?
-	bno	BASED(mcck_no_vtime)	# no -> skip cleanup critical
-	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
-	bz	BASED(mcck_no_vtime)
-	UPDATE_VTIME __LC_EXIT_TIMER,__LC_MCCK_ENTER_TIMER,__LC_USER_TIMER
-	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
-	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_MCCK_ENTER_TIMER
-mcck_no_vtime:
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
-	l	%r1,BASED(.Ls390_mcck)
-	basr	%r14,%r1		# call machine check handler
-	tm	SP_PSW+1(%r15),0x01	# returning to user ?
-	bno	BASED(mcck_return)
+3:	tm	__LC_MCCK_CODE+2,0x09	# mwp + ia of old psw valid?
+	jno	mcck_panic		# no -> skip cleanup critical
+	tm	%r8,0x0001		# interrupting from user ?
+	jz	mcck_skip
+	UPDATE_VTIME %r14,%r15,__LC_MCCK_ENTER_TIMER
+mcck_skip:
+	SWITCH_ASYNC __LC_GPREGS_SAVE_AREA+32,__LC_PANIC_STACK,PAGE_SHIFT
+	mvc	__PT_R0(64,%r11),__LC_GPREGS_SAVE_AREA
+	stm	%r8,%r9,__PT_PSW(%r11)
+	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15)
+	l	%r1,BASED(.Ldo_machine_check)
+	lr	%r2,%r11		# pass pointer to pt_regs
+	basr	%r14,%r1		# call s390_do_machine_check
+	tm	__PT_PSW+1(%r11),0x01	# returning to user ?
+	jno	mcck_return
 	l	%r1,__LC_KERNEL_STACK	# switch to kernel stack
-	s	%r1,BASED(.Lc_spsize)
-	mvc	SP_PTREGS(__PT_SIZE,%r1),SP_PTREGS(%r15)
-	xc	__SF_BACKCHAIN(4,%r1),__SF_BACKCHAIN(%r1) # clear back chain
+	ahi	%r1,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	mvc	STACK_FRAME_OVERHEAD(__PT_SIZE,%r1),0(%r11)
+	xc	__SF_BACKCHAIN(4,%r1),__SF_BACKCHAIN(%r1)
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
 	lr	%r15,%r1
-	stosm	__SF_EMPTY(%r15),0x04	# turn dat on
+	ssm	__LC_PGM_NEW_PSW	# turn dat on, keep irqs off
 	tm	__TI_flags+3(%r12),_TIF_MCCK_PENDING
-	bno	BASED(mcck_return)
+	jno	mcck_return
 	TRACE_IRQS_OFF
-	l	%r1,BASED(.Ls390_handle_mcck)
-	basr	%r14,%r1		# call machine check handler
+	l	%r1,BASED(.Lhandle_mcck)
+	basr	%r14,%r1		# call s390_handle_mcck
 	TRACE_IRQS_ON
 mcck_return:
-	mvc	__LC_RETURN_MCCK_PSW(8),SP_PSW(%r15) # move return PSW
+	mvc	__LC_RETURN_MCCK_PSW(8),__PT_PSW(%r11) # move return PSW
 	ni	__LC_RETURN_MCCK_PSW+1,0xfd # clear wait state bit
 	tm	__LC_RETURN_MCCK_PSW+1,0x01 # returning to user ?
-	bno	BASED(0f)
-	lm	%r0,%r15,SP_R0(%r15)	# load gprs 0-15
+	jno	0f
+	lm	%r0,%r15,__PT_R0(%r11)
 	stpt	__LC_EXIT_TIMER
-	lpsw	__LC_RETURN_MCCK_PSW	# back to caller
-0:	lm	%r0,%r15,SP_R0(%r15)	# load gprs 0-15
-	lpsw	__LC_RETURN_MCCK_PSW	# back to caller
+	lpsw	__LC_RETURN_MCCK_PSW
+0:	lm	%r0,%r15,__PT_R0(%r11)
+	lpsw	__LC_RETURN_MCCK_PSW
 
-	RESTORE_ALL __LC_RETURN_MCCK_PSW,0
+mcck_panic:
+	l	%r14,__LC_PANIC_STACK
+	slr	%r14,%r15
+	sra	%r14,PAGE_SHIFT
+	jz	0f
+	l	%r15,__LC_PANIC_STACK
+0:	ahi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	j	mcck_skip
 
 /*
  * Restart interruption handler, kick starter for additional CPUs
@@ -799,18 +703,18 @@ restart_base:
 	stck	__LC_LAST_UPDATE_CLOCK
 	mvc	__LC_LAST_UPDATE_TIMER(8),restart_vtime-restart_base(%r1)
 	mvc	__LC_EXIT_TIMER(8),restart_vtime-restart_base(%r1)
-	l	%r15,__LC_SAVE_AREA+60	# load ksp
+	l	%r15,__LC_GPREGS_SAVE_AREA+60 # load ksp
 	lctl	%c0,%c15,__LC_CREGS_SAVE_AREA # get new ctl regs
 	lam	%a0,%a15,__LC_AREGS_SAVE_AREA
-	lm	%r6,%r15,__SF_GPRS(%r15) # load registers from clone
+	lm	%r6,%r15,__SF_GPRS(%r15)# load registers from clone
 	l	%r1,__LC_THREAD_INFO
 	mvc	__LC_USER_TIMER(8),__TI_user_timer(%r1)
 	mvc	__LC_SYSTEM_TIMER(8),__TI_system_timer(%r1)
 	xc	__LC_STEAL_TIMER(8),__LC_STEAL_TIMER
-	stosm	__SF_EMPTY(%r15),0x04	# now we can turn dat on
+	ssm	__LC_PGM_NEW_PSW	# turn dat on, keep irqs off
 	basr	%r14,0
 	l	%r14,restart_addr-.(%r14)
-	basr	%r14,%r14		# branch to start_secondary
+	basr	%r14,%r14		# call start_secondary
 restart_addr:
 	.long	start_secondary
 	.align	8
@@ -835,19 +739,19 @@ restart_go:
 # PSW restart interrupt handler
 #
 ENTRY(psw_restart_int_handler)
-	st	%r15,__LC_SAVE_AREA+48(%r0)	# save r15
+	st	%r15,__LC_SAVE_AREA_RESTART
 	basr	%r15,0
 0:	l	%r15,.Lrestart_stack-0b(%r15)	# load restart stack
 	l	%r15,0(%r15)
-	ahi	%r15,-SP_SIZE			# make room for pt_regs
-	stm	%r0,%r14,SP_R0(%r15)		# store gprs %r0-%r14 to stack
-	mvc	SP_R15(4,%r15),__LC_SAVE_AREA+48(%r0)# store saved %r15 to stack
-	mvc	SP_PSW(8,%r15),__LC_RST_OLD_PSW(%r0) # store restart old psw
-	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15) # set backchain to 0
+	ahi	%r15,-__PT_SIZE			# create pt_regs on stack
+	stm	%r0,%r14,__PT_R0(%r15)
+	mvc	__PT_R15(4,%r15),__LC_SAVE_AREA_RESTART
+	mvc	__PT_PSW(8,%r15),__LC_RST_OLD_PSW # store restart old psw
+	ahi	%r15,-STACK_FRAME_OVERHEAD
+	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15)
 	basr	%r14,0
 1:	l	%r14,.Ldo_restart-1b(%r14)
 	basr	%r14,%r14
-
 	basr	%r14,0				# load disabled wait PSW if
 2:	lpsw	restart_psw_crash-2b(%r14)	# do_restart returns
 	.align 4
@@ -869,215 +773,174 @@ restart_psw_crash:
  */
 stack_overflow:
 	l	%r15,__LC_PANIC_STACK	# change to panic stack
-	sl	%r15,BASED(.Lc_spsize)
-	mvc	SP_PSW(8,%r15),0(%r12)	# move user PSW to stack
-	stm	%r0,%r11,SP_R0(%r15)	# store gprs %r0-%r11 to kernel stack
-	la	%r1,__LC_SAVE_AREA
-	ch	%r12,BASED(.L0x020)	# old psw addr == __LC_SVC_OLD_PSW ?
-	be	BASED(0f)
-	ch	%r12,BASED(.L0x028)	# old psw addr == __LC_PGM_OLD_PSW ?
-	be	BASED(0f)
-	la	%r1,__LC_SAVE_AREA+16
-0:	mvc	SP_R12(16,%r15),0(%r1)	# move %r12-%r15 to stack
-	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15) # clear back chain
-	l	%r1,BASED(1f)		# branch to kernel_stack_overflow
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
-	br	%r1
+	ahi	%r15,-__PT_SIZE		# create pt_regs
+	stm	%r0,%r7,__PT_R0(%r15)
+	stm	%r8,%r9,__PT_PSW(%r15)
+	mvc	__PT_R8(32,%r11),0(%r14)
+	lr	%r15,%r11
+	ahi	%r15,-STACK_FRAME_OVERHEAD
+	l	%r1,BASED(1f)
+	xc	__SF_BACKCHAIN(4,%r15),__SF_BACKCHAIN(%r15)
+	lr	%r2,%r11		# pass pointer to pt_regs
+	br	%r1			# branch to kernel_stack_overflow
 1:	.long	kernel_stack_overflow
 #endif
 
-cleanup_table_system_call:
-	.long	system_call + 0x80000000, sysc_do_svc + 0x80000000
-cleanup_table_sysc_tif:
-	.long	sysc_tif + 0x80000000, sysc_restore + 0x80000000
-cleanup_table_sysc_restore:
-	.long	sysc_restore + 0x80000000, sysc_done + 0x80000000
-cleanup_table_io_tif:
-	.long	io_tif + 0x80000000, io_restore + 0x80000000
-cleanup_table_io_restore:
-	.long	io_restore + 0x80000000, io_done + 0x80000000
+cleanup_table:
+	.long	system_call + 0x80000000
+	.long	sysc_do_svc + 0x80000000
+	.long	sysc_tif + 0x80000000
+	.long	sysc_restore + 0x80000000
+	.long	sysc_done + 0x80000000
+	.long	io_tif + 0x80000000
+	.long	io_restore + 0x80000000
+	.long	io_done + 0x80000000
 
 cleanup_critical:
-	clc	4(4,%r12),BASED(cleanup_table_system_call)
-	bl	BASED(0f)
-	clc	4(4,%r12),BASED(cleanup_table_system_call+4)
-	bl	BASED(cleanup_system_call)
-0:
-	clc	4(4,%r12),BASED(cleanup_table_sysc_tif)
-	bl	BASED(0f)
-	clc	4(4,%r12),BASED(cleanup_table_sysc_tif+4)
-	bl	BASED(cleanup_sysc_tif)
-0:
-	clc	4(4,%r12),BASED(cleanup_table_sysc_restore)
-	bl	BASED(0f)
-	clc	4(4,%r12),BASED(cleanup_table_sysc_restore+4)
-	bl	BASED(cleanup_sysc_restore)
-0:
-	clc	4(4,%r12),BASED(cleanup_table_io_tif)
-	bl	BASED(0f)
-	clc	4(4,%r12),BASED(cleanup_table_io_tif+4)
-	bl	BASED(cleanup_io_tif)
-0:
-	clc	4(4,%r12),BASED(cleanup_table_io_restore)
-	bl	BASED(0f)
-	clc	4(4,%r12),BASED(cleanup_table_io_restore+4)
-	bl	BASED(cleanup_io_restore)
-0:
-	br	%r14
+	cl	%r9,BASED(cleanup_table)	# system_call
+	jl	0f
+	cl	%r9,BASED(cleanup_table+4)	# sysc_do_svc
+	jl	cleanup_system_call
+	cl	%r9,BASED(cleanup_table+8)	# sysc_tif
+	jl	0f
+	cl	%r9,BASED(cleanup_table+12)	# sysc_restore
+	jl	cleanup_sysc_tif
+	cl	%r9,BASED(cleanup_table+16)	# sysc_done
+	jl	cleanup_sysc_restore
+	cl	%r9,BASED(cleanup_table+20)	# io_tif
+	jl	0f
+	cl	%r9,BASED(cleanup_table+24)	# io_restore
+	jl	cleanup_io_tif
+	cl	%r9,BASED(cleanup_table+28)	# io_done
+	jl	cleanup_io_restore
+0:	br	%r14
 
 cleanup_system_call:
-	mvc	__LC_RETURN_PSW(8),0(%r12)
-	clc	__LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn+4)
-	bh	BASED(0f)
-	mvc	__LC_SYNC_ENTER_TIMER(8),__LC_MCCK_ENTER_TIMER
-	c	%r12,BASED(.Lmck_old_psw)
-	be	BASED(0f)
+	# check if stpt has been executed
+	cl	%r9,BASED(cleanup_system_call_insn)
+	jh	0f
 	mvc	__LC_SYNC_ENTER_TIMER(8),__LC_ASYNC_ENTER_TIMER
-0:	c	%r12,BASED(.Lmck_old_psw)
-	la	%r12,__LC_SAVE_AREA+32
-	be	BASED(0f)
-	la	%r12,__LC_SAVE_AREA+16
-0:	clc	__LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn+8)
-	bhe	BASED(cleanup_vtime)
-	clc	__LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn)
-	bh	BASED(0f)
-	mvc	__LC_SAVE_AREA(16),0(%r12)
-0:	st	%r13,4(%r12)
-	l	%r15,__LC_KERNEL_STACK	# problem state -> load ksp
-	s	%r15,BASED(.Lc_spsize)	# make room for registers & psw
-	st	%r15,12(%r12)
-	CREATE_STACK_FRAME __LC_SAVE_AREA
-	mvc	0(4,%r12),__LC_THREAD_INFO
-	l	%r12,__LC_THREAD_INFO
-	mvc	SP_PSW(8,%r15),__LC_SVC_OLD_PSW
-	mvc	SP_SVC_CODE(4,%r15),__LC_SVC_ILC
-	oi	__TI_flags+3(%r12),_TIF_SYSCALL
-cleanup_vtime:
-	clc	__LC_RETURN_PSW+4(4),BASED(cleanup_system_call_insn+12)
-	bhe	BASED(cleanup_stime)
-	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:
+	chi	%r11,__LC_SAVE_AREA_ASYNC
+	je	0f
+	mvc	__LC_SYNC_ENTER_TIMER(8),__LC_MCCK_ENTER_TIMER
+0:	# check if stm has been executed
+	cl	%r9,BASED(cleanup_system_call_insn+4)
+	jh	0f
+	mvc	__LC_SAVE_AREA_SYNC(32),0(%r11)
+0:	# set up saved registers r12, and r13
+	st	%r12,16(%r11)		# r12 thread-info pointer
+	st	%r13,20(%r11)		# r13 literal-pool pointer
+	# check if the user time calculation has been done
+	cl	%r9,BASED(cleanup_system_call_insn+8)
+	jh	0f
+	l	%r10,__LC_EXIT_TIMER
+	l	%r15,__LC_EXIT_TIMER+4
+	SUB64	%r10,%r15,__LC_SYNC_ENTER_TIMER
+	ADD64	%r10,%r15,__LC_USER_TIMER
+	st	%r10,__LC_USER_TIMER
+	st	%r15,__LC_USER_TIMER+4
+0:	# check if the system time calculation has been done
+	cl	%r9,BASED(cleanup_system_call_insn+12)
+	jh	0f
+	l	%r10,__LC_LAST_UPDATE_TIMER
+	l	%r15,__LC_LAST_UPDATE_TIMER+4
+	SUB64	%r10,%r15,__LC_EXIT_TIMER
+	ADD64	%r10,%r15,__LC_SYSTEM_TIMER
+	st	%r10,__LC_SYSTEM_TIMER
+	st	%r15,__LC_SYSTEM_TIMER+4
+0:	# update accounting time stamp
 	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
-	mvc	__LC_RETURN_PSW+4(4),BASED(cleanup_table_system_call+4)
-	la	%r12,__LC_RETURN_PSW
+	# set up saved register 11
+	l	%r15,__LC_KERNEL_STACK
+	ahi	%r15,-__PT_SIZE
+	st	%r15,12(%r11)		# r11 pt_regs pointer
+	# fill pt_regs
+	mvc	__PT_R8(32,%r15),__LC_SAVE_AREA_SYNC
+	stm	%r0,%r7,__PT_R0(%r15)
+	mvc	__PT_PSW(8,%r15),__LC_SVC_OLD_PSW
+	mvc	__PT_SVC_CODE(4,%r15),__LC_SVC_ILC
+	# setup saved register 15
+	ahi	%r15,-STACK_FRAME_OVERHEAD
+	st	%r15,28(%r11)		# r15 stack pointer
+	# set new psw address and exit
+	l	%r9,BASED(cleanup_table+4)	# sysc_do_svc + 0x80000000
 	br	%r14
 cleanup_system_call_insn:
-	.long	sysc_saveall + 0x80000000
 	.long	system_call + 0x80000000
-	.long	sysc_vtime + 0x80000000
-	.long	sysc_stime + 0x80000000
-	.long	sysc_update + 0x80000000
+	.long	sysc_stm + 0x80000000
+	.long	sysc_vtime + 0x80000000 + 36
+	.long	sysc_vtime + 0x80000000 + 76
 
 cleanup_sysc_tif:
-	mvc	__LC_RETURN_PSW(4),0(%r12)
-	mvc	__LC_RETURN_PSW+4(4),BASED(cleanup_table_sysc_tif)
-	la	%r12,__LC_RETURN_PSW
+	l	%r9,BASED(cleanup_table+8)	# sysc_tif + 0x80000000
 	br	%r14
 
 cleanup_sysc_restore:
-	clc	4(4,%r12),BASED(cleanup_sysc_restore_insn)
-	be	BASED(2f)
-	mvc	__LC_EXIT_TIMER(8),__LC_MCCK_ENTER_TIMER
-	c	%r12,BASED(.Lmck_old_psw)
-	be	BASED(0f)
-	mvc	__LC_EXIT_TIMER(8),__LC_ASYNC_ENTER_TIMER
-0:	clc	4(4,%r12),BASED(cleanup_sysc_restore_insn+4)
-	be	BASED(2f)
-	mvc	__LC_RETURN_PSW(8),SP_PSW(%r15)
-	c	%r12,BASED(.Lmck_old_psw)
-	la	%r12,__LC_SAVE_AREA+32
-	be	BASED(1f)
-	la	%r12,__LC_SAVE_AREA+16
-1:	mvc	0(16,%r12),SP_R12(%r15)
-	lm	%r0,%r11,SP_R0(%r15)
-	l	%r15,SP_R15(%r15)
-2:	la	%r12,__LC_RETURN_PSW
+	cl	%r9,BASED(cleanup_sysc_restore_insn)
+	jhe	0f
+	l	%r9,12(%r11)		# get saved pointer to pt_regs
+	mvc	__LC_RETURN_PSW(8),__PT_PSW(%r9)
+	mvc	0(32,%r11),__PT_R8(%r9)
+	lm	%r0,%r7,__PT_R0(%r9)
+0:	lm	%r8,%r9,__LC_RETURN_PSW
 	br	%r14
 cleanup_sysc_restore_insn:
 	.long	sysc_done - 4 + 0x80000000
-	.long	sysc_done - 8 + 0x80000000
 
 cleanup_io_tif:
-	mvc	__LC_RETURN_PSW(4),0(%r12)
-	mvc	__LC_RETURN_PSW+4(4),BASED(cleanup_table_io_tif)
-	la	%r12,__LC_RETURN_PSW
+	l	%r9,BASED(cleanup_table+20)	# io_tif + 0x80000000
 	br	%r14
 
 cleanup_io_restore:
-	clc	4(4,%r12),BASED(cleanup_io_restore_insn)
-	be	BASED(1f)
-	mvc	__LC_EXIT_TIMER(8),__LC_MCCK_ENTER_TIMER
-	clc	4(4,%r12),BASED(cleanup_io_restore_insn+4)
-	be	BASED(1f)
-	mvc	__LC_RETURN_PSW(8),SP_PSW(%r15)
-	mvc	__LC_SAVE_AREA+32(16),SP_R12(%r15)
-	lm	%r0,%r11,SP_R0(%r15)
-	l	%r15,SP_R15(%r15)
-1:	la	%r12,__LC_RETURN_PSW
+	cl	%r9,BASED(cleanup_io_restore_insn)
+	jhe	0f
+	l	%r9,12(%r11)		# get saved r11 pointer to pt_regs
+	mvc	__LC_RETURN_PSW(8),__PT_PSW(%r9)
+	ni	__LC_RETURN_PSW+1,0xfd	# clear wait state bit
+	mvc	0(32,%r11),__PT_R8(%r9)
+	lm	%r0,%r7,__PT_R0(%r9)
+0:	lm	%r8,%r9,__LC_RETURN_PSW
 	br	%r14
 cleanup_io_restore_insn:
 	.long	io_done - 4 + 0x80000000
-	.long	io_done - 8 + 0x80000000
 
 /*
  * Integer constants
  */
-		.align	4
-.Lc_spsize:	.long	SP_SIZE
-.Lc_overhead:	.long	STACK_FRAME_OVERHEAD
-.Lnr_syscalls:	.long	NR_syscalls
-.L0x018:	.short	0x018
-.L0x020:	.short	0x020
-.L0x028:	.short	0x028
-.L0x030:	.short	0x030
-.L0x038:	.short	0x038
-.Lc_1:		.long	1
+	.align	4
+.Lnr_syscalls:		.long	NR_syscalls
 
 /*
  * Symbol constants
  */
-.Ls390_mcck:	.long	s390_do_machine_check
-.Ls390_handle_mcck:
-		.long	s390_handle_mcck
-.Lmck_old_psw:	.long	__LC_MCK_OLD_PSW
-.Ldo_IRQ:	.long	do_IRQ
-.Ldo_extint:	.long	do_extint
-.Ldo_signal:	.long	do_signal
-.Ldo_notify_resume:
-		.long	do_notify_resume
-.Lhandle_per:	.long	do_per_trap
-.Ldo_execve:	.long	do_execve
-.Lexecve_tail:	.long	execve_tail
-.Ljump_table:	.long	pgm_check_table
-.Lschedule:	.long	schedule
+.Ldo_machine_check:	.long	s390_do_machine_check
+.Lhandle_mcck:		.long	s390_handle_mcck
+.Ldo_IRQ:		.long	do_IRQ
+.Ldo_extint:		.long	do_extint
+.Ldo_signal:		.long	do_signal
+.Ldo_notify_resume:	.long	do_notify_resume
+.Ldo_per_trap:		.long	do_per_trap
+.Ldo_execve:		.long	do_execve
+.Lexecve_tail:		.long	execve_tail
+.Ljump_table:		.long	pgm_check_table
+.Lschedule:		.long	schedule
 #ifdef CONFIG_PREEMPT
-.Lpreempt_schedule_irq:
-		.long	preempt_schedule_irq
+.Lpreempt_irq:		.long	preempt_schedule_irq
 #endif
-.Ltrace_entry:	.long	do_syscall_trace_enter
-.Ltrace_exit:	.long	do_syscall_trace_exit
-.Lschedtail:	.long	schedule_tail
-.Lsysc_table:	.long	sys_call_table
+.Ltrace_enter:		.long	do_syscall_trace_enter
+.Ltrace_exit:		.long	do_syscall_trace_exit
+.Lschedule_tail:	.long	schedule_tail
+.Lsys_call_table:	.long	sys_call_table
+.Lsysc_per:		.long	sysc_per + 0x80000000
 #ifdef CONFIG_TRACE_IRQFLAGS
-.Ltrace_irq_on_caller:
-		.long	trace_hardirqs_on_caller
-.Ltrace_irq_off_caller:
-		.long	trace_hardirqs_off_caller
+.Lhardirqs_on:		.long	trace_hardirqs_on_caller
+.Lhardirqs_off:		.long	trace_hardirqs_off_caller
 #endif
 #ifdef CONFIG_LOCKDEP
-.Llockdep_sys_exit:
-		.long	lockdep_sys_exit
+.Llockdep_sys_exit:	.long	lockdep_sys_exit
 #endif
-.Lcritical_start:
-		.long	__critical_start + 0x80000000
-.Lcritical_end:
-		.long	__critical_end + 0x80000000
-.Lcleanup_critical:
-		.long	cleanup_critical
+.Lcritical_start:	.long	__critical_start + 0x80000000
+.Lcritical_length:	.long	__critical_end - __critical_start
 
 		.section .rodata, "a"
 #define SYSCALL(esa,esame,emu)	.long esa
diff --git a/arch/s390/kernel/entry64.S b/arch/s390/kernel/entry64.S
index 130fb02305c1..73845a9e587c 100644
--- a/arch/s390/kernel/entry64.S
+++ b/arch/s390/kernel/entry64.S
@@ -1,3 +1,4 @@
+
 /*
  *  arch/s390/kernel/entry64.S
  *    S390 low-level entry points.
@@ -19,32 +20,22 @@
 #include <asm/unistd.h>
 #include <asm/page.h>
 
-/*
- * Stack layout for the system_call stack entry.
- * The first few entries are identical to the user_regs_struct.
- */
-SP_PTREGS    =	STACK_FRAME_OVERHEAD
-SP_ARGS      =	STACK_FRAME_OVERHEAD + __PT_ARGS
-SP_PSW	     =	STACK_FRAME_OVERHEAD + __PT_PSW
-SP_R0	     =	STACK_FRAME_OVERHEAD + __PT_GPRS
-SP_R1	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 8
-SP_R2	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 16
-SP_R3	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 24
-SP_R4	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 32
-SP_R5	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 40
-SP_R6	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 48
-SP_R7	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 56
-SP_R8	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 64
-SP_R9	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 72
-SP_R10	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 80
-SP_R11	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 88
-SP_R12	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 96
-SP_R13	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 104
-SP_R14	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 112
-SP_R15	     =	STACK_FRAME_OVERHEAD + __PT_GPRS + 120
-SP_ORIG_R2   =	STACK_FRAME_OVERHEAD + __PT_ORIG_GPR2
-SP_SVC_CODE  =	STACK_FRAME_OVERHEAD + __PT_SVC_CODE
-SP_SIZE      =	STACK_FRAME_OVERHEAD + __PT_SIZE
+__PT_R0      =	__PT_GPRS
+__PT_R1      =	__PT_GPRS + 8
+__PT_R2      =	__PT_GPRS + 16
+__PT_R3      =	__PT_GPRS + 24
+__PT_R4      =	__PT_GPRS + 32
+__PT_R5      =	__PT_GPRS + 40
+__PT_R6      =	__PT_GPRS + 48
+__PT_R7      =	__PT_GPRS + 56
+__PT_R8      =	__PT_GPRS + 64
+__PT_R9      =	__PT_GPRS + 72
+__PT_R10     =	__PT_GPRS + 80
+__PT_R11     =	__PT_GPRS + 88
+__PT_R12     =	__PT_GPRS + 96
+__PT_R13     =	__PT_GPRS + 104
+__PT_R14     =	__PT_GPRS + 112
+__PT_R15     =	__PT_GPRS + 120
 
 STACK_SHIFT = PAGE_SHIFT + THREAD_ORDER
 STACK_SIZE  = 1 << STACK_SHIFT
@@ -59,154 +50,103 @@ _TIF_EXIT_SIE = (_TIF_SIGPENDING | _TIF_NEED_RESCHED | _TIF_MCCK_PENDING)
 
 #define BASED(name) name-system_call(%r13)
 
-	.macro SPP newpp
-#if defined(CONFIG_KVM) || defined(CONFIG_KVM_MODULE)
-	tm	__LC_MACHINE_FLAGS+6,0x20	# MACHINE_FLAG_SPP
-	jz	.+8
-	.insn	s,0xb2800000,\newpp
-#endif
-	.endm
-
-	.macro	HANDLE_SIE_INTERCEPT
-#if defined(CONFIG_KVM) || defined(CONFIG_KVM_MODULE)
-	tm	__TI_flags+6(%r12),_TIF_SIE>>8
-	jz	0f
-	SPP	BASED(.Lhost_id)		# set host id
-	clc	SP_PSW+8(8,%r15),BASED(.Lsie_loop)
-	jl	0f
-	clc	SP_PSW+8(8,%r15),BASED(.Lsie_done)
-	jhe	0f
-	mvc	SP_PSW+8(8,%r15),BASED(.Lsie_loop)
-0:
-#endif
-	.endm
-
-#ifdef CONFIG_TRACE_IRQFLAGS
 	.macro	TRACE_IRQS_ON
+#ifdef CONFIG_TRACE_IRQFLAGS
 	basr	%r2,%r0
 	brasl	%r14,trace_hardirqs_on_caller
+#endif
 	.endm
 
 	.macro	TRACE_IRQS_OFF
+#ifdef CONFIG_TRACE_IRQFLAGS
 	basr	%r2,%r0
 	brasl	%r14,trace_hardirqs_off_caller
-	.endm
-#else
-#define TRACE_IRQS_ON
-#define TRACE_IRQS_OFF
 #endif
+	.endm
 
-#ifdef CONFIG_LOCKDEP
 	.macro	LOCKDEP_SYS_EXIT
-	tm	SP_PSW+1(%r15),0x01	# returning to user ?
-	jz	0f
+#ifdef CONFIG_LOCKDEP
+	tm	__PT_PSW+1(%r11),0x01	# returning to user ?
+	jz	.+10
 	brasl	%r14,lockdep_sys_exit
-0:
-	.endm
-#else
-#define LOCKDEP_SYS_EXIT
 #endif
-
-	.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
 
-/*
- * Register usage in interrupt handlers:
- *    R9  - pointer to current task structure
- *    R13 - pointer to literal pool
- *    R14 - return register for function calls
- *    R15 - kernel stack pointer
- */
+	.macro SPP newpp
+#if defined(CONFIG_KVM) || defined(CONFIG_KVM_MODULE)
+	tm	__LC_MACHINE_FLAGS+6,0x20	# MACHINE_FLAG_SPP
+	jz	.+8
+	.insn	s,0xb2800000,\newpp
+#endif
+	.endm
 
-	.macro	SAVE_ALL_SVC psworg,savearea
-	stmg	%r11,%r15,\savearea
-	lg	%r15,__LC_KERNEL_STACK	# problem state -> load ksp
-	aghi	%r15,-SP_SIZE		# make room for registers & psw
-	lg	%r11,__LC_LAST_BREAK
+	.macro	HANDLE_SIE_INTERCEPT scratch
+#if defined(CONFIG_KVM) || defined(CONFIG_KVM_MODULE)
+	tm	__TI_flags+6(%r12),_TIF_SIE>>8
+	jz	.+42
+	tm	__LC_MACHINE_FLAGS+6,0x20	# MACHINE_FLAG_SPP
+	jz	.+8
+	.insn	s,0xb2800000,BASED(.Lhost_id)	# set host id
+	lgr	\scratch,%r9
+	slg	\scratch,BASED(.Lsie_loop)
+	clg	\scratch,BASED(.Lsie_length)
+	jhe	.+10
+	lg	%r9,BASED(.Lsie_loop)
+#endif
 	.endm
 
-	.macro	SAVE_ALL_PGM psworg,savearea
-	stmg	%r11,%r15,\savearea
-	tm	\psworg+1,0x01		# test problem state bit
+	.macro	CHECK_STACK stacksize,savearea
 #ifdef CONFIG_CHECK_STACK
-	jnz	1f
-	tml	%r15,STACK_SIZE - CONFIG_STACK_GUARD
-	jnz	2f
-	la	%r12,\psworg
-	j	stack_overflow
-#else
-	jz	2f
+	tml	%r15,\stacksize - CONFIG_STACK_GUARD
+	lghi	%r14,\savearea
+	jz	stack_overflow
 #endif
-1:	lg	%r15,__LC_KERNEL_STACK	# problem state -> load ksp
-2:	aghi	%r15,-SP_SIZE		# make room for registers & psw
-	larl	%r13,system_call
-	lg	%r11,__LC_LAST_BREAK
 	.endm
 
-	.macro	SAVE_ALL_ASYNC psworg,savearea
-	stmg	%r11,%r15,\savearea
-	larl	%r13,system_call
-	lg	%r11,__LC_LAST_BREAK
-	la	%r12,\psworg
-	tm	\psworg+1,0x01		# test problem state bit
-	jnz	1f			# from user -> load kernel stack
-	clc	\psworg+8(8),BASED(.Lcritical_end)
+	.macro	SWITCH_ASYNC savearea,stack,shift
+	tmhh	%r8,0x0001		# interrupting from user ?
+	jnz	1f
+	lgr	%r14,%r9
+	slg	%r14,BASED(.Lcritical_start)
+	clg	%r14,BASED(.Lcritical_length)
 	jhe	0f
-	clc	\psworg+8(8),BASED(.Lcritical_start)
-	jl	0f
+	lghi	%r11,\savearea		# inside critical section, do cleanup
 	brasl	%r14,cleanup_critical
-	tm	1(%r12),0x01		# retest problem state after cleanup
+	tmhh	%r8,0x0001		# retest problem state after cleanup
 	jnz	1f
-0:	lg	%r14,__LC_ASYNC_STACK	# are we already on the async. stack ?
+0:	lg	%r14,\stack		# are we already on the target stack?
 	slgr	%r14,%r15
-	srag	%r14,%r14,STACK_SHIFT
-#ifdef CONFIG_CHECK_STACK
+	srag	%r14,%r14,\shift
 	jnz	1f
-	tml	%r15,STACK_SIZE - CONFIG_STACK_GUARD
-	jnz	2f
-	j	stack_overflow
-#else
-	jz	2f
-#endif
-1:	lg	%r15,__LC_ASYNC_STACK	# load async stack
-2:	aghi	%r15,-SP_SIZE		# make room for registers & psw
-	.endm
-
-	.macro	CREATE_STACK_FRAME savearea
-	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
-	stg	%r2,SP_ORIG_R2(%r15)	# store original content of gpr 2
-	mvc	SP_R11(40,%r15),\savearea # move %r11-%r15 to stack
-	stmg	%r0,%r10,SP_R0(%r15)	# store gprs %r0-%r10 to kernel stack
+	CHECK_STACK 1<<\shift,\savearea
+	j	2f
+1:	lg	%r15,\stack		# load target stack
+2:	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
 	.endm
 
-	.macro	RESTORE_ALL psworg,sync
-	mvc	\psworg(16),SP_PSW(%r15) # move user PSW to lowcore
-	.if !\sync
-	ni	\psworg+1,0xfd		# clear wait state bit
-	.endif
-	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
+	.macro UPDATE_VTIME scratch,enter_timer
+	lg	\scratch,__LC_EXIT_TIMER
+	slg	\scratch,\enter_timer
+	alg	\scratch,__LC_USER_TIMER
+	stg	\scratch,__LC_USER_TIMER
+	lg	\scratch,__LC_LAST_UPDATE_TIMER
+	slg	\scratch,__LC_EXIT_TIMER
+	alg	\scratch,__LC_SYSTEM_TIMER
+	stg	\scratch,__LC_SYSTEM_TIMER
+	mvc	__LC_LAST_UPDATE_TIMER(8),\enter_timer
 	.endm
 
-	.macro	LAST_BREAK
-	srag	%r10,%r11,23
-	jz	0f
-	stg	%r11,__TI_last_break(%r12)
-0:
+	.macro	LAST_BREAK scratch
+	srag	\scratch,%r10,23
+	jz	.+10
+	stg	%r10,__TI_last_break(%r12)
 	.endm
 
 	.macro REENABLE_IRQS
-	mvc	__SF_EMPTY(1,%r15),SP_PSW(%r15)
-	ni	__SF_EMPTY(%r15),0xbf
-	ssm	__SF_EMPTY(%r15)
+	stg	%r8,__LC_RETURN_PSW
+	ni	__LC_RETURN_PSW,0xbf
+	ssm	__LC_RETURN_PSW
 	.endm
 
 	.section .kprobes.text, "ax"
@@ -245,55 +185,66 @@ __critical_start:
 
 ENTRY(system_call)
 	stpt	__LC_SYNC_ENTER_TIMER
-sysc_saveall:
-	SAVE_ALL_SVC __LC_SVC_OLD_PSW,__LC_SAVE_AREA
-	CREATE_STACK_FRAME __LC_SAVE_AREA
-	lg	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	mvc	SP_PSW(16,%r15),__LC_SVC_OLD_PSW
-	mvc	SP_SVC_CODE(4,%r15),__LC_SVC_ILC
-	oi	__TI_flags+7(%r12),_TIF_SYSCALL
+sysc_stmg:
+	stmg	%r8,%r15,__LC_SAVE_AREA_SYNC
+	lg	%r10,__LC_LAST_BREAK
+	lg	%r12,__LC_THREAD_INFO
+	larl	%r13,system_call
+sysc_per:
+	lg	%r15,__LC_KERNEL_STACK
+	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)	# pointer to pt_regs
 sysc_vtime:
-	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
-	LAST_BREAK
+	UPDATE_VTIME %r13,__LC_SYNC_ENTER_TIMER
+	LAST_BREAK %r13
+	stmg	%r0,%r7,__PT_R0(%r11)
+	mvc	__PT_R8(64,%r11),__LC_SAVE_AREA_SYNC
+	mvc	__PT_PSW(16,%r11),__LC_SVC_OLD_PSW
+	mvc	__PT_SVC_CODE(4,%r11),__LC_SVC_ILC
 sysc_do_svc:
-	llgh	%r7,SP_SVC_CODE+2(%r15)
-	slag	%r7,%r7,2	# shift and test for svc 0
+	oi	__TI_flags+7(%r12),_TIF_SYSCALL
+	llgh	%r8,__PT_SVC_CODE+2(%r11)
+	slag	%r8,%r8,2			# shift and test for svc 0
 	jnz	sysc_nr_ok
 	# svc 0: system call number in %r1
-	llgfr	%r1,%r1		# clear high word in r1
+	llgfr	%r1,%r1				# clear high word in r1
 	cghi	%r1,NR_syscalls
 	jnl	sysc_nr_ok
-	sth	%r1,SP_SVC_CODE+2(%r15)
-	slag	%r7,%r1,2	# shift and test for svc 0
+	sth	%r1,__PT_SVC_CODE+2(%r11)
+	slag	%r8,%r1,2
 sysc_nr_ok:
-	larl	%r10,sys_call_table
+	larl	%r10,sys_call_table		# 64 bit system call table
 #ifdef CONFIG_COMPAT
-	tm	__TI_flags+5(%r12),(_TIF_31BIT>>16)  # running in 31 bit mode ?
+	tm	__TI_flags+5(%r12),(_TIF_31BIT>>16)
 	jno	sysc_noemu
-	larl	%r10,sys_call_table_emu  # use 31 bit emulation system calls
+	larl	%r10,sys_call_table_emu		# 31 bit system call table
 sysc_noemu:
 #endif
+	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
+	stg	%r2,__PT_ORIG_GPR2(%r11)
+	stg	%r7,STACK_FRAME_OVERHEAD(%r15)
+	lgf	%r9,0(%r8,%r10)			# get system call add.
 	tm	__TI_flags+6(%r12),_TIF_TRACE >> 8
-	mvc	SP_ARGS(8,%r15),SP_R7(%r15)
-	lgf	%r8,0(%r7,%r10) # load address of system call routine
 	jnz	sysc_tracesys
-	basr	%r14,%r8	# call sys_xxxx
-	stg	%r2,SP_R2(%r15) # store return value (change R2 on stack)
+	basr	%r14,%r9			# call sys_xxxx
+	stg	%r2,__PT_R2(%r11)		# store return value
 
 sysc_return:
 	LOCKDEP_SYS_EXIT
 sysc_tif:
-	tm	SP_PSW+1(%r15),0x01	# returning to user ?
+	tm	__PT_PSW+1(%r11),0x01		# returning to user ?
 	jno	sysc_restore
 	tm	__TI_flags+7(%r12),_TIF_WORK_SVC
-	jnz	sysc_work	# there is work to do (signals etc.)
+	jnz	sysc_work			# check for work
 	ni	__TI_flags+7(%r12),255-_TIF_SYSCALL
 sysc_restore:
-	RESTORE_ALL __LC_RETURN_PSW,1
+	lg	%r14,__LC_VDSO_PER_CPU
+	lmg	%r0,%r10,__PT_R0(%r11)
+	mvc	__LC_RETURN_PSW(16),__PT_PSW(%r11)
+	stpt	__LC_EXIT_TIMER
+	mvc	__VDSO_ECTG_BASE(16,%r14),__LC_EXIT_TIMER
+	lmg	%r11,%r15,__PT_R11(%r11)
+	lpswe	__LC_RETURN_PSW
 sysc_done:
 
 #
@@ -317,7 +268,7 @@ sysc_work:
 #
 sysc_reschedule:
 	larl	%r14,sysc_return
-	jg	schedule		# return point is sysc_return
+	jg	schedule
 
 #
 # _TIF_MCCK_PENDING is set, call handler
@@ -331,33 +282,33 @@ sysc_mcck_pending:
 #
 sysc_sigpending:
 	ni	__TI_flags+7(%r12),255-_TIF_PER_TRAP # clear TIF_PER_TRAP
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
-	brasl	%r14,do_signal		# call do_signal
+	lgr	%r2,%r11		# pass pointer to pt_regs
+	brasl	%r14,do_signal
 	tm	__TI_flags+7(%r12),_TIF_SYSCALL
 	jno	sysc_return
-	lmg	%r2,%r6,SP_R2(%r15)	# load svc arguments
-	lghi	%r7,0			# svc 0 returns -ENOSYS
-	lh	%r1,SP_SVC_CODE+2(%r15)	# load new svc number
+	lmg	%r2,%r7,__PT_R2(%r11)	# load svc arguments
+	lghi	%r8,0			# svc 0 returns -ENOSYS
+	lh	%r1,__PT_SVC_CODE+2(%r11)	# load new svc number
 	cghi	%r1,NR_syscalls
 	jnl	sysc_nr_ok		# invalid svc number -> do svc 0
-	slag	%r7,%r1,2
+	slag	%r8,%r1,2
 	j	sysc_nr_ok		# restart svc
 
 #
 # _TIF_NOTIFY_RESUME is set, call do_notify_resume
 #
 sysc_notify_resume:
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
+	lgr	%r2,%r11		# pass pointer to pt_regs
 	larl	%r14,sysc_return
-	jg	do_notify_resume	# call do_notify_resume
+	jg	do_notify_resume
 
 #
 # _TIF_PER_TRAP is set, call do_per_trap
 #
 sysc_singlestep:
 	ni	__TI_flags+7(%r12),255-(_TIF_SYSCALL | _TIF_PER_TRAP)
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
-	larl	%r14,sysc_return	# load adr. of system return
+	lgr	%r2,%r11		# pass pointer to pt_regs
+	larl	%r14,sysc_return
 	jg	do_per_trap
 
 #
@@ -365,41 +316,41 @@ sysc_singlestep:
 # and after the system call
 #
 sysc_tracesys:
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
+	lgr	%r2,%r11		# pass pointer to pt_regs
 	la	%r3,0
-	llgh	%r0,SP_SVC_CODE+2(%r15)
-	stg	%r0,SP_R2(%r15)
+	llgh	%r0,__PT_SVC_CODE+2(%r11)
+	stg	%r0,__PT_R2(%r11)
 	brasl	%r14,do_syscall_trace_enter
 	lghi	%r0,NR_syscalls
 	clgr	%r0,%r2
 	jnh	sysc_tracenogo
-	sllg	%r7,%r2,2		# svc number *4
-	lgf	%r8,0(%r7,%r10)
+	sllg	%r8,%r2,2
+	lgf	%r9,0(%r8,%r10)
 sysc_tracego:
-	lmg	%r3,%r6,SP_R3(%r15)
-	mvc	SP_ARGS(8,%r15),SP_R7(%r15)
-	lg	%r2,SP_ORIG_R2(%r15)
-	basr	%r14,%r8		# call sys_xxx
-	stg	%r2,SP_R2(%r15)		# store return value
+	lmg	%r3,%r7,__PT_R3(%r11)
+	stg	%r7,STACK_FRAME_OVERHEAD(%r15)
+	lg	%r2,__PT_ORIG_GPR2(%r11)
+	basr	%r14,%r9		# call sys_xxx
+	stg	%r2,__PT_R2(%r11)	# store return value
 sysc_tracenogo:
 	tm	__TI_flags+6(%r12),_TIF_TRACE >> 8
 	jz	sysc_return
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
-	larl	%r14,sysc_return	# return point is sysc_return
+	lgr	%r2,%r11		# pass pointer to pt_regs
+	larl	%r14,sysc_return
 	jg	do_syscall_trace_exit
 
 #
 # a new process exits the kernel with ret_from_fork
 #
 ENTRY(ret_from_fork)
-	lg	%r13,__LC_SVC_NEW_PSW+8
-	lg	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	tm	SP_PSW+1(%r15),0x01	# forking a kernel thread ?
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+	lg	%r12,__LC_THREAD_INFO
+	tm	__PT_PSW+1(%r11),0x01	# forking a kernel thread ?
 	jo	0f
-	stg	%r15,SP_R15(%r15)	# store stack pointer for new kthread
+	stg	%r15,__PT_R15(%r11)	# store stack pointer for new kthread
 0:	brasl	%r14,schedule_tail
 	TRACE_IRQS_ON
-	stosm	24(%r15),0x03		# reenable interrupts
+	ssm	__LC_SVC_NEW_PSW	# reenable interrupts
 	j	sysc_tracenogo
 
 #
@@ -409,26 +360,26 @@ ENTRY(ret_from_fork)
 ENTRY(kernel_execve)
 	stmg	%r12,%r15,96(%r15)
 	lgr	%r14,%r15
-	aghi	%r15,-SP_SIZE
+	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
 	stg	%r14,__SF_BACKCHAIN(%r15)
-	la	%r12,SP_PTREGS(%r15)
+	la	%r12,STACK_FRAME_OVERHEAD(%r15)
 	xc	0(__PT_SIZE,%r12),0(%r12)
 	lgr	%r5,%r12
 	brasl	%r14,do_execve
 	ltgfr	%r2,%r2
 	je	0f
-	aghi	%r15,SP_SIZE
+	aghi	%r15,(STACK_FRAME_OVERHEAD + __PT_SIZE)
 	lmg	%r12,%r15,96(%r15)
 	br	%r14
 	# execve succeeded.
-0:	stnsm	__SF_EMPTY(%r15),0xfc	# disable interrupts
+0:	ssm	__LC_PGM_NEW_PSW	# disable I/O and ext. interrupts
 	lg	%r15,__LC_KERNEL_STACK	# load ksp
-	aghi	%r15,-SP_SIZE		# make room for registers & psw
-	lg	%r13,__LC_SVC_NEW_PSW+8
-	mvc	SP_PTREGS(__PT_SIZE,%r15),0(%r12)	# copy pt_regs
+	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+	mvc	0(__PT_SIZE,%r11),0(%r12)	# copy pt_regs
 	lg	%r12,__LC_THREAD_INFO
 	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
+	ssm	__LC_SVC_NEW_PSW	# reenable interrupts
 	brasl	%r14,execve_tail
 	j	sysc_return
 
@@ -437,127 +388,72 @@ ENTRY(kernel_execve)
  */
 
 ENTRY(pgm_check_handler)
-/*
- * First we need to check for a special case:
- * Single stepping an instruction that disables the PER event mask will
- * cause a PER event AFTER the mask has been set. Example: SVC or LPSW.
- * For a single stepped SVC the program check handler gets control after
- * the SVC new PSW has been loaded. But we want to execute the SVC first and
- * then handle the PER event. Therefore we update the SVC old PSW to point
- * to the pgm_check_handler and branch to the SVC handler after we checked
- * if we have to load the kernel stack register.
- * For every other possible cause for PER event without the PER mask set
- * we just ignore the PER event (FIXME: is there anything we have to do
- * for LPSW?).
- */
 	stpt	__LC_SYNC_ENTER_TIMER
-	tm	__LC_PGM_INT_CODE+1,0x80 # check whether we got a per exception
-	jnz	pgm_per 		 # got per exception -> special case
-	SAVE_ALL_PGM __LC_PGM_OLD_PSW,__LC_SAVE_AREA
-	CREATE_STACK_FRAME __LC_SAVE_AREA
-	mvc	SP_PSW(16,%r15),__LC_PGM_OLD_PSW
-	lg	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	HANDLE_SIE_INTERCEPT
-	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
-	LAST_BREAK
-pgm_no_vtime:
-	stg	%r11,SP_ARGS(%r15)
-	lgf	%r3,__LC_PGM_ILC	# load program interruption code
-	lg	%r4,__LC_TRANS_EXC_CODE
-	REENABLE_IRQS
-	lghi	%r8,0x7f
-	ngr	%r8,%r3
-	sll	%r8,3
-	larl	%r1,pgm_check_table
-	lg	%r1,0(%r8,%r1)		# load address of handler routine
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
-	basr	%r14,%r1		# branch to interrupt-handler
-pgm_exit:
-	j	sysc_return
-
-#
-# handle per exception
-#
-pgm_per:
-	tm	__LC_PGM_OLD_PSW,0x40	# test if per event recording is on
-	jnz	pgm_per_std		# ok, normal per event from user space
-# ok its one of the special cases, now we need to find out which one
-	clc	__LC_PGM_OLD_PSW(16),__LC_SVC_NEW_PSW
-	je	pgm_svcper
-# no interesting special case, ignore PER event
-	lpswe	__LC_PGM_OLD_PSW
-
-#
-# Normal per exception
-#
-pgm_per_std:
-	SAVE_ALL_PGM __LC_PGM_OLD_PSW,__LC_SAVE_AREA
-	CREATE_STACK_FRAME __LC_SAVE_AREA
-	mvc	SP_PSW(16,%r15),__LC_PGM_OLD_PSW
-	lg	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	HANDLE_SIE_INTERCEPT
-	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
-	LAST_BREAK
-pgm_no_vtime2:
+	stmg	%r8,%r15,__LC_SAVE_AREA_SYNC
+	lg	%r10,__LC_LAST_BREAK
+	lg	%r12,__LC_THREAD_INFO
+	larl	%r13,system_call
+	lmg	%r8,%r9,__LC_PGM_OLD_PSW
+	HANDLE_SIE_INTERCEPT %r14
+	tmhh	%r8,0x0001		# test problem state bit
+	jnz	1f			# -> fault in user space
+	tmhh	%r8,0x4000		# PER bit set in old PSW ?
+	jnz	0f			# -> enabled, can't be a double fault
+	tm	__LC_PGM_ILC+3,0x80	# check for per exception
+	jnz	pgm_svcper		# -> single stepped svc
+0:	CHECK_STACK STACK_SIZE,__LC_SAVE_AREA_SYNC
+	j	2f
+1:	UPDATE_VTIME %r14,__LC_SYNC_ENTER_TIMER
+	LAST_BREAK %r14
+	lg	%r15,__LC_KERNEL_STACK
+2:	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+	stmg	%r0,%r7,__PT_R0(%r11)
+	mvc	__PT_R8(64,%r11),__LC_SAVE_AREA_SYNC
+	stmg	%r8,%r9,__PT_PSW(%r11)
+	stg	%r10,__PT_ARGS(%r11)
+	tm	__LC_PGM_ILC+3,0x80	# check for per exception
+	jz	0f
 	lg	%r1,__TI_task(%r12)
-	tm	SP_PSW+1(%r15),0x01	# kernel per event ?
-	jz	kernel_per
-	mvc	__THREAD_per_cause(2,%r1),__LC_PER_CAUSE
+	tmhh	%r8,0x0001		# kernel per event ?
+	jz	pgm_kprobe
+	oi	__TI_flags+7(%r12),_TIF_PER_TRAP
 	mvc	__THREAD_per_address(8,%r1),__LC_PER_ADDRESS
+	mvc	__THREAD_per_cause(2,%r1),__LC_PER_CAUSE
 	mvc	__THREAD_per_paid(1,%r1),__LC_PER_PAID
-	oi	__TI_flags+7(%r12),_TIF_PER_TRAP # set TIF_PER_TRAP
-	lgf	%r3,__LC_PGM_ILC	# load program interruption code
+0:	lgf	%r3,__LC_PGM_ILC	# load program interruption code
 	lg	%r4,__LC_TRANS_EXC_CODE
 	REENABLE_IRQS
-	lghi	%r8,0x7f
-	ngr	%r8,%r3			# clear per-event-bit and ilc
-	je	pgm_exit2
-	sll	%r8,3
+	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
+	lghi	%r10,0x7f
+	ngr	%r10,%r3
+	je	sysc_return
+	sll	%r10,3
 	larl	%r1,pgm_check_table
-	lg	%r1,0(%r8,%r1)		# load address of handler routine
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
+	lg	%r1,0(%r10,%r1)		# load address of handler routine
+	lgr	%r2,%r11		# pass pointer to pt_regs
 	basr	%r14,%r1		# branch to interrupt-handler
-pgm_exit2:
 	j	sysc_return
 
 #
-# it was a single stepped SVC that is causing all the trouble
+# PER event in supervisor state, must be kprobes
 #
-pgm_svcper:
-	SAVE_ALL_PGM __LC_SVC_OLD_PSW,__LC_SAVE_AREA
-	CREATE_STACK_FRAME __LC_SAVE_AREA
-	lg	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	mvc	SP_PSW(16,%r15),__LC_SVC_OLD_PSW
-	mvc	SP_SVC_CODE(4,%r15),__LC_SVC_ILC
-	oi	__TI_flags+7(%r12),(_TIF_SYSCALL | _TIF_PER_TRAP)
-	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
-	LAST_BREAK
-	lg	%r8,__TI_task(%r12)
-	mvc	__THREAD_per_cause(2,%r8),__LC_PER_CAUSE
-	mvc	__THREAD_per_address(8,%r8),__LC_PER_ADDRESS
-	mvc	__THREAD_per_paid(1,%r8),__LC_PER_PAID
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
-	lmg	%r2,%r6,SP_R2(%r15)	# load svc arguments
-	j	sysc_do_svc
+pgm_kprobe:
+	REENABLE_IRQS
+	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
+	lgr	%r2,%r11		# pass pointer to pt_regs
+	brasl	%r14,do_per_trap
+	j	sysc_return
 
 #
-# per was called from kernel, must be kprobes
+# single stepped system call
 #
-kernel_per:
-	REENABLE_IRQS
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
-	brasl	%r14,do_per_trap
-	j	pgm_exit
+pgm_svcper:
+	oi	__TI_flags+7(%r12),_TIF_PER_TRAP
+	mvc	__LC_RETURN_PSW(8),__LC_SVC_NEW_PSW
+	larl	%r14,sysc_per
+	stg	%r14,__LC_RETURN_PSW+8
+	lpswe	__LC_RETURN_PSW		# branch to sysc_per and enable irqs
 
 /*
  * IO interrupt handler routine
@@ -565,21 +461,25 @@ kernel_per:
 ENTRY(io_int_handler)
 	stck	__LC_INT_CLOCK
 	stpt	__LC_ASYNC_ENTER_TIMER
-	SAVE_ALL_ASYNC __LC_IO_OLD_PSW,__LC_SAVE_AREA+40
-	CREATE_STACK_FRAME __LC_SAVE_AREA+40
-	mvc	SP_PSW(16,%r15),0(%r12)	# move user PSW to stack
-	lg	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	HANDLE_SIE_INTERCEPT
-	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
-	LAST_BREAK
-io_no_vtime:
+	stmg	%r8,%r15,__LC_SAVE_AREA_ASYNC
+	lg	%r10,__LC_LAST_BREAK
+	lg	%r12,__LC_THREAD_INFO
+	larl	%r13,system_call
+	lmg	%r8,%r9,__LC_IO_OLD_PSW
+	HANDLE_SIE_INTERCEPT %r14
+	SWITCH_ASYNC __LC_SAVE_AREA_ASYNC,__LC_ASYNC_STACK,STACK_SHIFT
+	tmhh	%r8,0x0001		# interrupting from user?
+	jz	io_skip
+	UPDATE_VTIME %r14,__LC_ASYNC_ENTER_TIMER
+	LAST_BREAK %r14
+io_skip:
+	stmg	%r0,%r7,__PT_R0(%r11)
+	mvc	__PT_R8(64,%r11),__LC_SAVE_AREA_ASYNC
+	stmg	%r8,%r9,__PT_PSW(%r11)
 	TRACE_IRQS_OFF
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
-	brasl	%r14,do_IRQ		# call standard irq handler
+	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
+	lgr	%r2,%r11		# pass pointer to pt_regs
+	brasl	%r14,do_IRQ
 io_return:
 	LOCKDEP_SYS_EXIT
 	TRACE_IRQS_ON
@@ -587,7 +487,14 @@ io_tif:
 	tm	__TI_flags+7(%r12),_TIF_WORK_INT
 	jnz	io_work 		# there is work to do (signals etc.)
 io_restore:
-	RESTORE_ALL __LC_RETURN_PSW,0
+	lg	%r14,__LC_VDSO_PER_CPU
+	lmg	%r0,%r10,__PT_R0(%r11)
+	mvc	__LC_RETURN_PSW(16),__PT_PSW(%r11)
+	ni	__LC_RETURN_PSW+1,0xfd	# clear wait state bit
+	stpt	__LC_EXIT_TIMER
+	mvc	__VDSO_ECTG_BASE(16,%r14),__LC_EXIT_TIMER
+	lmg	%r11,%r15,__PT_R11(%r11)
+	lpswe	__LC_RETURN_PSW
 io_done:
 
 #
@@ -600,7 +507,7 @@ io_done:
 # Before any work can be done, a switch to the kernel stack is required.
 #
 io_work:
-	tm	SP_PSW+1(%r15),0x01	# returning to user ?
+	tm	__PT_PSW+1(%r11),0x01	# returning to user ?
 	jo	io_work_user		# yes -> do resched & signal
 #ifdef CONFIG_PREEMPT
 	# check for preemptive scheduling
@@ -609,10 +516,11 @@ io_work:
 	tm	__TI_flags+7(%r12),_TIF_NEED_RESCHED
 	jno	io_restore
 	# switch to kernel stack
-	lg	%r1,SP_R15(%r15)
-	aghi	%r1,-SP_SIZE
-	mvc	SP_PTREGS(__PT_SIZE,%r1),SP_PTREGS(%r15)
-	xc	__SF_BACKCHAIN(8,%r1),__SF_BACKCHAIN(%r1) # clear back chain
+	lg	%r1,__PT_R15(%r11)
+	aghi	%r1,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	mvc	STACK_FRAME_OVERHEAD(__PT_SIZE,%r1),0(%r11)
+	xc	__SF_BACKCHAIN(8,%r1),__SF_BACKCHAIN(%r1)
+	la	%r11,STACK_FRAME_OVERHEAD(%r1)
 	lgr	%r15,%r1
 	# TRACE_IRQS_ON already done at io_return, call
 	# TRACE_IRQS_OFF to keep things symmetrical
@@ -628,9 +536,10 @@ io_work:
 #
 io_work_user:
 	lg	%r1,__LC_KERNEL_STACK
-	aghi	%r1,-SP_SIZE
-	mvc	SP_PTREGS(__PT_SIZE,%r1),SP_PTREGS(%r15)
-	xc	__SF_BACKCHAIN(8,%r1),__SF_BACKCHAIN(%r1) # clear back chain
+	aghi	%r1,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	mvc	STACK_FRAME_OVERHEAD(__PT_SIZE,%r1),0(%r11)
+	xc	__SF_BACKCHAIN(8,%r1),__SF_BACKCHAIN(%r1)
+	la	%r11,STACK_FRAME_OVERHEAD(%r1)
 	lgr	%r15,%r1
 
 #
@@ -663,9 +572,9 @@ io_mcck_pending:
 #
 io_reschedule:
 	# TRACE_IRQS_ON already done at io_return
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
+	ssm	__LC_SVC_NEW_PSW	# reenable interrupts
 	brasl	%r14,schedule		# call scheduler
-	stnsm	__SF_EMPTY(%r15),0xfc	# disable I/O and ext. interrupts
+	ssm	__LC_PGM_NEW_PSW	# disable I/O and ext. interrupts
 	TRACE_IRQS_OFF
 	j	io_return
 
@@ -674,10 +583,10 @@ io_reschedule:
 #
 io_sigpending:
 	# TRACE_IRQS_ON already done at io_return
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
-	brasl	%r14,do_signal		# call do_signal
-	stnsm	__SF_EMPTY(%r15),0xfc	# disable I/O and ext. interrupts
+	ssm	__LC_SVC_NEW_PSW	# reenable interrupts
+	lgr	%r2,%r11		# pass pointer to pt_regs
+	brasl	%r14,do_signal
+	ssm	__LC_PGM_NEW_PSW	# disable I/O and ext. interrupts
 	TRACE_IRQS_OFF
 	j	io_return
 
@@ -686,10 +595,10 @@ io_sigpending:
 #
 io_notify_resume:
 	# TRACE_IRQS_ON already done at io_return
-	stosm	__SF_EMPTY(%r15),0x03	# reenable interrupts
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
-	brasl	%r14,do_notify_resume	# call do_notify_resume
-	stnsm	__SF_EMPTY(%r15),0xfc	# disable I/O and ext. interrupts
+	ssm	__LC_SVC_NEW_PSW	# reenable interrupts
+	lgr	%r2,%r11		# pass pointer to pt_regs
+	brasl	%r14,do_notify_resume
+	ssm	__LC_PGM_NEW_PSW	# disable I/O and ext. interrupts
 	TRACE_IRQS_OFF
 	j	io_return
 
@@ -699,21 +608,24 @@ io_notify_resume:
 ENTRY(ext_int_handler)
 	stck	__LC_INT_CLOCK
 	stpt	__LC_ASYNC_ENTER_TIMER
-	SAVE_ALL_ASYNC __LC_EXT_OLD_PSW,__LC_SAVE_AREA+40
-	CREATE_STACK_FRAME __LC_SAVE_AREA+40
-	mvc	SP_PSW(16,%r15),0(%r12)	# move user PSW to stack
-	lg	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	HANDLE_SIE_INTERCEPT
-	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
-	LAST_BREAK
-ext_no_vtime:
+	stmg	%r8,%r15,__LC_SAVE_AREA_ASYNC
+	lg	%r10,__LC_LAST_BREAK
+	lg	%r12,__LC_THREAD_INFO
+	larl	%r13,system_call
+	lmg	%r8,%r9,__LC_EXT_OLD_PSW
+	HANDLE_SIE_INTERCEPT %r14
+	SWITCH_ASYNC __LC_SAVE_AREA_ASYNC,__LC_ASYNC_STACK,STACK_SHIFT
+	tmhh	%r8,0x0001		# interrupting from user ?
+	jz	ext_skip
+	UPDATE_VTIME %r14,__LC_ASYNC_ENTER_TIMER
+	LAST_BREAK %r14
+ext_skip:
+	stmg	%r0,%r7,__PT_R0(%r11)
+	mvc	__PT_R8(64,%r11),__LC_SAVE_AREA_ASYNC
+	stmg	%r8,%r9,__PT_PSW(%r11)
 	TRACE_IRQS_OFF
 	lghi	%r1,4096
-	la	%r2,SP_PTREGS(%r15)	# address of register-save area
+	lgr	%r2,%r11		# pass pointer to pt_regs
 	llgf	%r3,__LC_CPU_ADDRESS	# get cpu address + interruption code
 	llgf	%r4,__LC_EXT_PARAMS	# get external parameter
 	lg	%r5,__LC_EXT_PARAMS2-4096(%r1)	# get 64 bit external parameter
@@ -730,81 +642,77 @@ ENTRY(mcck_int_handler)
 	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
-	stmg	%r11,%r15,__LC_SAVE_AREA+80
+	lg	%r10,__LC_LAST_BREAK
+	lg	%r12,__LC_THREAD_INFO
 	larl	%r13,system_call
-	lg	%r11,__LC_LAST_BREAK
-	la	%r12,__LC_MCK_OLD_PSW
+	lmg	%r8,%r9,__LC_MCK_OLD_PSW
+	HANDLE_SIE_INTERCEPT %r14
 	tm	__LC_MCCK_CODE,0x80	# system damage?
-	jo	mcck_int_main		# yes -> rest of mcck code invalid
-	la	%r14,4095
-	mvc	__LC_MCCK_ENTER_TIMER(8),__LC_CPU_TIMER_SAVE_AREA-4095(%r14)
+	jo	mcck_panic		# yes -> rest of mcck code invalid
+	lghi	%r14,__LC_CPU_TIMER_SAVE_AREA
+	mvc	__LC_MCCK_ENTER_TIMER(8),0(%r14)
 	tm	__LC_MCCK_CODE+5,0x02	# stored cpu timer value valid?
-	jo	1f
+	jo	3f
 	la	%r14,__LC_SYNC_ENTER_TIMER
 	clc	0(8,%r14),__LC_ASYNC_ENTER_TIMER
 	jl	0f
 	la	%r14,__LC_ASYNC_ENTER_TIMER
 0:	clc	0(8,%r14),__LC_EXIT_TIMER
-	jl	0f
+	jl	1f
 	la	%r14,__LC_EXIT_TIMER
-0:	clc	0(8,%r14),__LC_LAST_UPDATE_TIMER
-	jl	0f
+1:	clc	0(8,%r14),__LC_LAST_UPDATE_TIMER
+	jl	2f
 	la	%r14,__LC_LAST_UPDATE_TIMER
-0:	spt	0(%r14)
+2:	spt	0(%r14)
 	mvc	__LC_MCCK_ENTER_TIMER(8),0(%r14)
-1:	tm	__LC_MCCK_CODE+2,0x09	# mwp + ia of old psw valid?
-	jno	mcck_int_main		# no -> skip cleanup critical
-	tm	__LC_MCK_OLD_PSW+1,0x01 # test problem state bit
-	jnz	mcck_int_main		# from user -> load kernel stack
-	clc	__LC_MCK_OLD_PSW+8(8),BASED(.Lcritical_end)
-	jhe	mcck_int_main
-	clc	__LC_MCK_OLD_PSW+8(8),BASED(.Lcritical_start)
-	jl	mcck_int_main
-	brasl	%r14,cleanup_critical
-mcck_int_main:
-	lg	%r14,__LC_PANIC_STACK	# are we already on the panic stack?
-	slgr	%r14,%r15
-	srag	%r14,%r14,PAGE_SHIFT
-	jz	0f
-	lg	%r15,__LC_PANIC_STACK	# load panic stack
-0:	aghi	%r15,-SP_SIZE		# make room for registers & psw
-	CREATE_STACK_FRAME __LC_SAVE_AREA+80
-	mvc	SP_PSW(16,%r15),0(%r12)
-	lg	%r12,__LC_THREAD_INFO	# load pointer to thread_info struct
-	tm	__LC_MCCK_CODE+2,0x08	# mwp of old psw valid?
-	jno	mcck_no_vtime		# no -> no timer update
-	HANDLE_SIE_INTERCEPT
-	tm	SP_PSW+1(%r15),0x01	# interrupting from user ?
-	jz	mcck_no_vtime
-	UPDATE_VTIME __LC_EXIT_TIMER,__LC_MCCK_ENTER_TIMER,__LC_USER_TIMER
-	UPDATE_VTIME __LC_LAST_UPDATE_TIMER,__LC_EXIT_TIMER,__LC_SYSTEM_TIMER
-	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_MCCK_ENTER_TIMER
-	LAST_BREAK
-mcck_no_vtime:
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
+3:	tm	__LC_MCCK_CODE+2,0x09	# mwp + ia of old psw valid?
+	jno	mcck_panic		# no -> skip cleanup critical
+	SWITCH_ASYNC __LC_GPREGS_SAVE_AREA+64,__LC_PANIC_STACK,PAGE_SHIFT
+	tm	%r8,0x0001		# interrupting from user ?
+	jz	mcck_skip
+	UPDATE_VTIME %r14,__LC_MCCK_ENTER_TIMER
+	LAST_BREAK %r14
+mcck_skip:
+	lghi	%r14,__LC_GPREGS_SAVE_AREA
+	mvc	__PT_R0(128,%r11),0(%r14)
+	stmg	%r8,%r9,__PT_PSW(%r11)
+	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
+	lgr	%r2,%r11		# pass pointer to pt_regs
 	brasl	%r14,s390_do_machine_check
-	tm	SP_PSW+1(%r15),0x01	# returning to user ?
+	tm	__PT_PSW+1(%r11),0x01	# returning to user ?
 	jno	mcck_return
 	lg	%r1,__LC_KERNEL_STACK	# switch to kernel stack
-	aghi	%r1,-SP_SIZE
-	mvc	SP_PTREGS(__PT_SIZE,%r1),SP_PTREGS(%r15)
-	xc	__SF_BACKCHAIN(8,%r1),__SF_BACKCHAIN(%r1) # clear back chain
+	aghi	%r1,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	mvc	STACK_FRAME_OVERHEAD(__PT_SIZE,%r1),0(%r11)
+	xc	__SF_BACKCHAIN(8,%r1),__SF_BACKCHAIN(%r1)
+	la	%r11,STACK_FRAME_OVERHEAD(%r1)
 	lgr	%r15,%r1
-	stosm	__SF_EMPTY(%r15),0x04	# turn dat on
+	ssm	__LC_PGM_NEW_PSW	# turn dat on, keep irqs off
 	tm	__TI_flags+7(%r12),_TIF_MCCK_PENDING
 	jno	mcck_return
 	TRACE_IRQS_OFF
 	brasl	%r14,s390_handle_mcck
 	TRACE_IRQS_ON
 mcck_return:
-	mvc	__LC_RETURN_MCCK_PSW(16),SP_PSW(%r15) # move return PSW
+	lg	%r14,__LC_VDSO_PER_CPU
+	lmg	%r0,%r10,__PT_R0(%r11)
+	mvc	__LC_RETURN_MCCK_PSW(16),__PT_PSW(%r11) # move return PSW
 	ni	__LC_RETURN_MCCK_PSW+1,0xfd # clear wait state bit
-	lmg	%r0,%r15,SP_R0(%r15)	# load gprs 0-15
 	tm	__LC_RETURN_MCCK_PSW+1,0x01 # returning to user ?
 	jno	0f
 	stpt	__LC_EXIT_TIMER
-0:	lpswe	__LC_RETURN_MCCK_PSW	# back to caller
-mcck_done:
+	mvc	__VDSO_ECTG_BASE(16,%r14),__LC_EXIT_TIMER
+0:	lmg	%r11,%r15,__PT_R11(%r11)
+	lpswe	__LC_RETURN_MCCK_PSW
+
+mcck_panic:
+	lg	%r14,__LC_PANIC_STACK
+	slgr	%r14,%r15
+	srag	%r14,%r14,PAGE_SHIFT
+	jz	0f
+	lg	%r15,__LC_PANIC_STACK
+0:	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
+	j	mcck_skip
 
 /*
  * Restart interruption handler, kick starter for additional CPUs
@@ -818,17 +726,18 @@ restart_base:
 	stck	__LC_LAST_UPDATE_CLOCK
 	mvc	__LC_LAST_UPDATE_TIMER(8),restart_vtime-restart_base(%r1)
 	mvc	__LC_EXIT_TIMER(8),restart_vtime-restart_base(%r1)
-	lg	%r15,__LC_SAVE_AREA+120 # load ksp
+	lghi	%r10,__LC_GPREGS_SAVE_AREA
+	lg	%r15,120(%r10)		# load ksp
 	lghi	%r10,__LC_CREGS_SAVE_AREA
-	lctlg	%c0,%c15,0(%r10) # get new ctl regs
+	lctlg	%c0,%c15,0(%r10)	# get new ctl regs
 	lghi	%r10,__LC_AREGS_SAVE_AREA
 	lam	%a0,%a15,0(%r10)
-	lmg	%r6,%r15,__SF_GPRS(%r15) # load registers from clone
+	lmg	%r6,%r15,__SF_GPRS(%r15)# load registers from clone
 	lg	%r1,__LC_THREAD_INFO
 	mvc	__LC_USER_TIMER(8),__TI_user_timer(%r1)
 	mvc	__LC_SYSTEM_TIMER(8),__TI_system_timer(%r1)
 	xc	__LC_STEAL_TIMER(8),__LC_STEAL_TIMER
-	stosm	__SF_EMPTY(%r15),0x04	# now we can turn dat on
+	ssm	__LC_PGM_NEW_PSW	# turn dat on, keep irqs off
 	brasl	%r14,start_secondary
 	.align	8
 restart_vtime:
@@ -852,16 +761,16 @@ restart_go:
 # PSW restart interrupt handler
 #
 ENTRY(psw_restart_int_handler)
-	stg	%r15,__LC_SAVE_AREA+120(%r0)	# save r15
+	stg	%r15,__LC_SAVE_AREA_RESTART
 	larl	%r15,restart_stack		# load restart stack
 	lg	%r15,0(%r15)
-	aghi	%r15,-SP_SIZE			# make room for pt_regs
-	stmg	%r0,%r14,SP_R0(%r15)		# store gprs %r0-%r14 to stack
-	mvc	SP_R15(8,%r15),__LC_SAVE_AREA+120(%r0)# store saved %r15 to stack
-	mvc	SP_PSW(16,%r15),__LC_RST_OLD_PSW(%r0)# store restart old psw
-	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15) # set backchain to 0
+	aghi	%r15,-__PT_SIZE			# create pt_regs on stack
+	stmg	%r0,%r14,__PT_R0(%r15)
+	mvc	__PT_R15(8,%r15),__LC_SAVE_AREA_RESTART
+	mvc	__PT_PSW(16,%r15),__LC_RST_OLD_PSW # store restart old psw
+	aghi	%r15,-STACK_FRAME_OVERHEAD
+	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
 	brasl	%r14,do_restart
-
 	larl	%r14,restart_psw_crash		# load disabled wait PSW if
 	lpswe	0(%r14)				# do_restart returns
 	.align 8
@@ -877,172 +786,153 @@ restart_psw_crash:
  * Setup a pt_regs so that show_trace can provide a good call trace.
  */
 stack_overflow:
-	lg	%r15,__LC_PANIC_STACK	# change to panic stack
-	aghi	%r15,-SP_SIZE
-	mvc	SP_PSW(16,%r15),0(%r12)	# move user PSW to stack
-	stmg	%r0,%r10,SP_R0(%r15)	# store gprs %r0-%r10 to kernel stack
-	la	%r1,__LC_SAVE_AREA
-	chi	%r12,__LC_SVC_OLD_PSW
-	je	0f
-	chi	%r12,__LC_PGM_OLD_PSW
-	je	0f
-	la	%r1,__LC_SAVE_AREA+40
-0:	mvc	SP_R11(40,%r15),0(%r1)	# move %r11-%r15 to stack
-	mvc	SP_ARGS(8,%r15),__LC_LAST_BREAK
-	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15) # clear back chain
-	la	%r2,SP_PTREGS(%r15)	# load pt_regs
+	lg	%r11,__LC_PANIC_STACK	# change to panic stack
+	aghi	%r11,-__PT_SIZE		# create pt_regs
+	stmg	%r0,%r7,__PT_R0(%r11)
+	stmg	%r8,%r9,__PT_PSW(%r11)
+	mvc	__PT_R8(64,%r11),0(%r14)
+	stg	%r10,__PT_ORIG_GPR2(%r11) # store last break to orig_gpr2
+	lgr	%r15,%r11
+	aghi	%r15,-STACK_FRAME_OVERHEAD
+	xc	__SF_BACKCHAIN(8,%r15),__SF_BACKCHAIN(%r15)
+	lgr	%r2,%r11		# pass pointer to pt_regs
 	jg	kernel_stack_overflow
 #endif
 
-cleanup_table_system_call:
-	.quad	system_call, sysc_do_svc
-cleanup_table_sysc_tif:
-	.quad	sysc_tif, sysc_restore
-cleanup_table_sysc_restore:
-	.quad	sysc_restore, sysc_done
-cleanup_table_io_tif:
-	.quad	io_tif, io_restore
-cleanup_table_io_restore:
-	.quad	io_restore, io_done
+	.align	8
+cleanup_table:
+	.quad	system_call
+	.quad	sysc_do_svc
+	.quad	sysc_tif
+	.quad	sysc_restore
+	.quad	sysc_done
+	.quad	io_tif
+	.quad	io_restore
+	.quad	io_done
 
 cleanup_critical:
-	clc	8(8,%r12),BASED(cleanup_table_system_call)
+	clg	%r9,BASED(cleanup_table)	# system_call
 	jl	0f
-	clc	8(8,%r12),BASED(cleanup_table_system_call+8)
+	clg	%r9,BASED(cleanup_table+8)	# sysc_do_svc
 	jl	cleanup_system_call
-0:
-	clc	8(8,%r12),BASED(cleanup_table_sysc_tif)
+	clg	%r9,BASED(cleanup_table+16)	# sysc_tif
 	jl	0f
-	clc	8(8,%r12),BASED(cleanup_table_sysc_tif+8)
+	clg	%r9,BASED(cleanup_table+24)	# sysc_restore
 	jl	cleanup_sysc_tif
-0:
-	clc	8(8,%r12),BASED(cleanup_table_sysc_restore)
-	jl	0f
-	clc	8(8,%r12),BASED(cleanup_table_sysc_restore+8)
+	clg	%r9,BASED(cleanup_table+32)	# sysc_done
 	jl	cleanup_sysc_restore
-0:
-	clc	8(8,%r12),BASED(cleanup_table_io_tif)
+	clg	%r9,BASED(cleanup_table+40)	# io_tif
 	jl	0f
-	clc	8(8,%r12),BASED(cleanup_table_io_tif+8)
+	clg	%r9,BASED(cleanup_table+48)	# io_restore
 	jl	cleanup_io_tif
-0:
-	clc	8(8,%r12),BASED(cleanup_table_io_restore)
-	jl	0f
-	clc	8(8,%r12),BASED(cleanup_table_io_restore+8)
+	clg	%r9,BASED(cleanup_table+56)	# io_done
 	jl	cleanup_io_restore
-0:
-	br	%r14
+0:	br	%r14
+
 
 cleanup_system_call:
-	mvc	__LC_RETURN_PSW(16),0(%r12)
-	clc	__LC_RETURN_PSW+8(8),BASED(cleanup_system_call_insn+8)
+	# check if stpt has been executed
+	clg	%r9,BASED(cleanup_system_call_insn)
 	jh	0f
-	mvc	__LC_SYNC_ENTER_TIMER(8),__LC_MCCK_ENTER_TIMER
-	cghi	%r12,__LC_MCK_OLD_PSW
-	je	0f
 	mvc	__LC_SYNC_ENTER_TIMER(8),__LC_ASYNC_ENTER_TIMER
-0:	cghi	%r12,__LC_MCK_OLD_PSW
-	la	%r12,__LC_SAVE_AREA+80
+	cghi	%r11,__LC_SAVE_AREA_ASYNC
 	je	0f
-	la	%r12,__LC_SAVE_AREA+40
-0:	clc	__LC_RETURN_PSW+8(8),BASED(cleanup_system_call_insn+16)
-	jhe	cleanup_vtime
-	clc	__LC_RETURN_PSW+8(8),BASED(cleanup_system_call_insn)
+	mvc	__LC_SYNC_ENTER_TIMER(8),__LC_MCCK_ENTER_TIMER
+0:	# check if stmg has been executed
+	clg	%r9,BASED(cleanup_system_call_insn+8)
 	jh	0f
-	mvc	__LC_SAVE_AREA(40),0(%r12)
-0:	lg	%r15,__LC_KERNEL_STACK	# problem state -> load ksp
-	aghi	%r15,-SP_SIZE		# make room for registers & psw
-	stg	%r15,32(%r12)
-	stg	%r11,0(%r12)
-	CREATE_STACK_FRAME __LC_SAVE_AREA
-	mvc	8(8,%r12),__LC_THREAD_INFO
-	lg	%r12,__LC_THREAD_INFO
-	mvc	SP_PSW(16,%r15),__LC_SVC_OLD_PSW
-	mvc	SP_SVC_CODE(4,%r15),__LC_SVC_ILC
-	oi	__TI_flags+7(%r12),_TIF_SYSCALL
-cleanup_vtime:
-	clc	__LC_RETURN_PSW+8(8),BASED(cleanup_system_call_insn+24)
-	jhe	cleanup_stime
-	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_SAVE_AREA_SYNC(64),0(%r11)
+0:	# check if base register setup + TIF bit load has been done
+	clg	%r9,BASED(cleanup_system_call_insn+16)
+	jhe	0f
+	# set up saved registers r10 and r12
+	stg	%r10,16(%r11)		# r10 last break
+	stg	%r12,32(%r11)		# r12 thread-info pointer
+0:	# check if the user time update has been done
+	clg	%r9,BASED(cleanup_system_call_insn+24)
+	jh	0f
+	lg	%r15,__LC_EXIT_TIMER
+	slg	%r15,__LC_SYNC_ENTER_TIMER
+	alg	%r15,__LC_USER_TIMER
+	stg	%r15,__LC_USER_TIMER
+0:	# check if the system time update has been done
+	clg	%r9,BASED(cleanup_system_call_insn+32)
+	jh	0f
+	lg	%r15,__LC_LAST_UPDATE_TIMER
+	slg	%r15,__LC_EXIT_TIMER
+	alg	%r15,__LC_SYSTEM_TIMER
+	stg	%r15,__LC_SYSTEM_TIMER
+0:	# update accounting time stamp
 	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
-	srag	%r12,%r11,23
-	lg	%r12,__LC_THREAD_INFO
+	# do LAST_BREAK
+	lg	%r9,16(%r11)
+	srag	%r9,%r9,23
 	jz	0f
-	stg	%r11,__TI_last_break(%r12)
-0:	mvc	__LC_RETURN_PSW+8(8),BASED(cleanup_table_system_call+8)
-	la	%r12,__LC_RETURN_PSW
+	mvc	__TI_last_break(8,%r12),16(%r11)
+0:	# set up saved register r11
+	lg	%r15,__LC_KERNEL_STACK
+	aghi	%r15,-__PT_SIZE
+	stg	%r15,24(%r11)		# r11 pt_regs pointer
+	# fill pt_regs
+	mvc	__PT_R8(64,%r15),__LC_SAVE_AREA_SYNC
+	stmg	%r0,%r7,__PT_R0(%r15)
+	mvc	__PT_PSW(16,%r15),__LC_SVC_OLD_PSW
+	mvc	__PT_SVC_CODE(4,%r15),__LC_SVC_ILC
+	# setup saved register r15
+	aghi	%r15,-STACK_FRAME_OVERHEAD
+	stg	%r15,56(%r11)		# r15 stack pointer
+	# set new psw address and exit
+	larl	%r9,sysc_do_svc
 	br	%r14
 cleanup_system_call_insn:
-	.quad	sysc_saveall
 	.quad	system_call
-	.quad	sysc_vtime
-	.quad	sysc_stime
-	.quad	sysc_update
+	.quad	sysc_stmg
+	.quad	sysc_per
+	.quad	sysc_vtime+18
+	.quad	sysc_vtime+42
 
 cleanup_sysc_tif:
-	mvc	__LC_RETURN_PSW(8),0(%r12)
-	mvc	__LC_RETURN_PSW+8(8),BASED(cleanup_table_sysc_tif)
-	la	%r12,__LC_RETURN_PSW
+	larl	%r9,sysc_tif
 	br	%r14
 
 cleanup_sysc_restore:
-	clc	8(8,%r12),BASED(cleanup_sysc_restore_insn)
-	je	2f
-	clc	8(8,%r12),BASED(cleanup_sysc_restore_insn+8)
-	jhe	0f
-	mvc	__LC_EXIT_TIMER(8),__LC_MCCK_ENTER_TIMER
-	cghi	%r12,__LC_MCK_OLD_PSW
+	clg	%r9,BASED(cleanup_sysc_restore_insn)
 	je	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
-	la	%r12,__LC_SAVE_AREA+80
-	je	1f
-	la	%r12,__LC_SAVE_AREA+40
-1:	mvc	0(40,%r12),SP_R11(%r15)
-	lmg	%r0,%r10,SP_R0(%r15)
-	lg	%r15,SP_R15(%r15)
-2:	la	%r12,__LC_RETURN_PSW
+	lg	%r9,24(%r11)		# get saved pointer to pt_regs
+	mvc	__LC_RETURN_PSW(16),__PT_PSW(%r9)
+	mvc	0(64,%r11),__PT_R8(%r9)
+	lmg	%r0,%r7,__PT_R0(%r9)
+0:	lmg	%r8,%r9,__LC_RETURN_PSW
 	br	%r14
 cleanup_sysc_restore_insn:
 	.quad	sysc_done - 4
-	.quad	sysc_done - 16
 
 cleanup_io_tif:
-	mvc	__LC_RETURN_PSW(8),0(%r12)
-	mvc	__LC_RETURN_PSW+8(8),BASED(cleanup_table_io_tif)
-	la	%r12,__LC_RETURN_PSW
+	larl	%r9,io_tif
 	br	%r14
 
 cleanup_io_restore:
-	clc	8(8,%r12),BASED(cleanup_io_restore_insn)
-	je	1f
-	clc	8(8,%r12),BASED(cleanup_io_restore_insn+8)
-	jhe	0f
-	mvc	__LC_EXIT_TIMER(8),__LC_MCCK_ENTER_TIMER
-0:	mvc	__LC_RETURN_PSW(16),SP_PSW(%r15)
-	mvc	__LC_SAVE_AREA+80(40),SP_R11(%r15)
-	lmg	%r0,%r10,SP_R0(%r15)
-	lg	%r15,SP_R15(%r15)
-1:	la	%r12,__LC_RETURN_PSW
+	clg	%r9,BASED(cleanup_io_restore_insn)
+	je	0f
+	lg	%r9,24(%r11)		# get saved r11 pointer to pt_regs
+	mvc	__LC_RETURN_PSW(16),__PT_PSW(%r9)
+	ni	__LC_RETURN_PSW+1,0xfd	# clear wait state bit
+	mvc	0(64,%r11),__PT_R8(%r9)
+	lmg	%r0,%r7,__PT_R0(%r9)
+0:	lmg	%r8,%r9,__LC_RETURN_PSW
 	br	%r14
 cleanup_io_restore_insn:
 	.quad	io_done - 4
-	.quad	io_done - 16
 
 /*
  * Integer constants
  */
-		.align	4
+	.align	8
 .Lcritical_start:
-		.quad	__critical_start
-.Lcritical_end:
-		.quad	__critical_end
+	.quad	__critical_start
+.Lcritical_length:
+	.quad	__critical_end - __critical_start
+
 
 #if defined(CONFIG_KVM) || defined(CONFIG_KVM_MODULE)
 /*
@@ -1094,8 +984,8 @@ sie_fault:
 	.align	8
 .Lsie_loop:
 	.quad	sie_loop
-.Lsie_done:
-	.quad	sie_done
+.Lsie_length:
+	.quad	sie_done - sie_loop
 .Lhost_id:
 	.quad	0
 
diff --git a/arch/s390/kernel/head.S b/arch/s390/kernel/head.S
index 900068d2bf92..c27a0727f930 100644
--- a/arch/s390/kernel/head.S
+++ b/arch/s390/kernel/head.S
@@ -329,8 +329,8 @@ iplstart:
 #
 # reset files in VM reader
 #
-	stidp	__LC_SAVE_AREA		# store cpuid
-	tm	__LC_SAVE_AREA,0xff	# running VM ?
+	stidp	__LC_SAVE_AREA_SYNC	# store cpuid
+	tm	__LC_SAVE_AREA_SYNC,0xff# running VM ?
 	bno	.Lnoreset
 	la	%r2,.Lreset
 	lhi	%r3,26
diff --git a/arch/s390/kernel/reipl64.S b/arch/s390/kernel/reipl64.S
index 732a793ec53a..36b32658fb24 100644
--- a/arch/s390/kernel/reipl64.S
+++ b/arch/s390/kernel/reipl64.S
@@ -17,11 +17,11 @@
 #
 ENTRY(store_status)
 	/* Save register one and load save area base */
-	stg	%r1,__LC_SAVE_AREA+120(%r0)
+	stg	%r1,__LC_SAVE_AREA_RESTART
 	lghi	%r1,SAVE_AREA_BASE
 	/* General purpose registers */
 	stmg	%r0,%r15,__LC_GPREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
-	lg	%r2,__LC_SAVE_AREA+120(%r0)
+	lg	%r2,__LC_SAVE_AREA_RESTART
 	stg	%r2,__LC_GPREGS_SAVE_AREA-SAVE_AREA_BASE+8(%r1)
 	/* Control registers */
 	stctg	%c0,%c15,__LC_CREGS_SAVE_AREA-SAVE_AREA_BASE(%r1)
diff --git a/arch/s390/kernel/smp.c b/arch/s390/kernel/smp.c
index 9cf01e455e50..109e7422bb20 100644
--- a/arch/s390/kernel/smp.c
+++ b/arch/s390/kernel/smp.c
@@ -654,7 +654,7 @@ int __cpuinit __cpu_up(unsigned int cpu)
 				     - sizeof(struct stack_frame));
 	memset(sf, 0, sizeof(struct stack_frame));
 	sf->gprs[9] = (unsigned long) sf;
-	cpu_lowcore->save_area[15] = (unsigned long) sf;
+	cpu_lowcore->gpregs_save_area[15] = (unsigned long) sf;
 	__ctl_store(cpu_lowcore->cregs_save_area, 0, 15);
 	atomic_inc(&init_mm.context.attach_count);
 	asm volatile(