[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_PTREG