s390: lowcore stack pointer offsets

Store the stack pointers in the lowcore for the kernel stack, the async
stack and the panic stack with the offset required for the first user.
This avoids an unnecessary add instruction on the system call path.
Signed-off-by: Martin Schwidefsky <schwidefsky@de.ibm.com>

arch/s390/kernel/entry.S

@@ -45,6 +45,7 @@ _TIF_TRACE    = (_TIF_SYSCALL_TRACE | _TIF_SYSCALL_AUDIT | _TIF_SECCOMP | \
 
 STACK_SHIFT = PAGE_SHIFT + THREAD_ORDER
 STACK_SIZE  = 1 << STACK_SHIFT
+STACK_INIT  = STACK_SIZE - STACK_FRAME_OVERHEAD - __PT_SIZE
 
 #define BASED(name) name-system_call(%r13)
 
@@ -97,10 +98,10 @@ STACK_SIZE  = 1 << STACK_SHIFT
 	sra	%r14,\shift
 	jnz	1f
 	CHECK_STACK 1<<\shift,\savearea
+	ahi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
 	j	2f
 1:	l	%r15,\stack		# load target stack
-2:	ahi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
-	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+2:	la	%r11,STACK_FRAME_OVERHEAD(%r15)
 	.endm
 
 	.macro	ADD64 high,low,timer
@@ -150,7 +151,7 @@ ENTRY(__switch_to)
 	l	%r4,__THREAD_info(%r2)		# get thread_info of prev
 	l	%r5,__THREAD_info(%r3)		# get thread_info of next
 	lr	%r15,%r5
-	ahi	%r15,STACK_SIZE			# end of kernel stack of next
+	ahi	%r15,STACK_INIT			# end of kernel stack of next
 	st	%r3,__LC_CURRENT		# store task struct of next
 	st	%r5,__LC_THREAD_INFO		# store thread info of next
 	st	%r15,__LC_KERNEL_STACK		# store end of kernel stack
@@ -178,7 +179,6 @@ sysc_stm:
 	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 %r8,%r9,__LC_SYNC_ENTER_TIMER
@@ -359,11 +359,11 @@ ENTRY(pgm_check_handler)
 	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
+	ahi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
 	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)
+2:	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)
@@ -485,7 +485,6 @@ io_work:
 #
 io_work_user:
 	l	%r1,__LC_KERNEL_STACK
-	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)
@@ -646,7 +645,6 @@ mcck_skip:
 	tm	__PT_PSW+1(%r11),0x01	# returning to user ?
 	jno	mcck_return
 	l	%r1,__LC_KERNEL_STACK	# switch to kernel stack
-	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)
@@ -674,6 +672,7 @@ mcck_panic:
 	sra	%r14,PAGE_SHIFT
 	jz	0f
 	l	%r15,__LC_PANIC_STACK
+	j	mcck_skip
 0:	ahi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
 	j	mcck_skip
 
@@ -714,12 +713,10 @@ ENTRY(restart_int_handler)
  */
 stack_overflow:
 	l	%r15,__LC_PANIC_STACK	# change to panic stack
-	ahi	%r15,-__PT_SIZE		# create pt_regs
-	stm	%r0,%r7,__PT_R0(%r15)
-	stm	%r8,%r9,__PT_PSW(%r15)
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+	stm	%r0,%r7,__PT_R0(%r11)
+	stm	%r8,%r9,__PT_PSW(%r11)
 	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
@@ -799,15 +796,14 @@ cleanup_system_call:
 	mvc	__LC_LAST_UPDATE_TIMER(8),__LC_SYNC_ENTER_TIMER
 	# set up saved register 11
 	l	%r15,__LC_KERNEL_STACK
-	ahi	%r15,-__PT_SIZE
-	st	%r15,12(%r11)		# r11 pt_regs pointer
+	la	%r9,STACK_FRAME_OVERHEAD(%r15)
+	st	%r9,12(%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_INT_CODE(4,%r15),__LC_SVC_ILC
+	mvc	__PT_R8(32,%r9),__LC_SAVE_AREA_SYNC
+	stm	%r0,%r7,__PT_R0(%r9)
+	mvc	__PT_PSW(8,%r9),__LC_SVC_OLD_PSW
+	mvc	__PT_INT_CODE(4,%r9),__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

arch/s390/kernel/entry64.S

@@ -39,6 +39,7 @@ __PT_R15     =	__PT_GPRS + 120
 
 STACK_SHIFT = PAGE_SHIFT + THREAD_ORDER
 STACK_SIZE  = 1 << STACK_SHIFT
+STACK_INIT = STACK_SIZE - STACK_FRAME_OVERHEAD - __PT_SIZE
 
 _TIF_WORK_SVC = (_TIF_SIGPENDING | _TIF_NOTIFY_RESUME | _TIF_NEED_RESCHED | \
 		 _TIF_MCCK_PENDING | _TIF_PER_TRAP )
@@ -124,10 +125,10 @@ _TIF_EXIT_SIE = (_TIF_SIGPENDING | _TIF_NEED_RESCHED | _TIF_MCCK_PENDING)
 	srag	%r14,%r14,\shift
 	jnz	1f
 	CHECK_STACK 1<<\shift,\savearea
+	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
 	j	2f
 1:	lg	%r15,\stack		# load target stack
-2:	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
-	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+2:	la	%r11,STACK_FRAME_OVERHEAD(%r15)
 	.endm
 
 	.macro UPDATE_VTIME scratch,enter_timer
@@ -177,7 +178,7 @@ ENTRY(__switch_to)
 	lg	%r4,__THREAD_info(%r2)		# get thread_info of prev
 	lg	%r5,__THREAD_info(%r3)		# get thread_info of next
 	lgr	%r15,%r5
-	aghi	%r15,STACK_SIZE			# end of kernel stack of next
+	aghi	%r15,STACK_INIT			# end of kernel stack of next
 	stg	%r3,__LC_CURRENT		# store task struct of next
 	stg	%r5,__LC_THREAD_INFO		# store thread info of next
 	stg	%r15,__LC_KERNEL_STACK		# store end of kernel stack
@@ -203,10 +204,8 @@ 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 %r13,__LC_SYNC_ENTER_TIMER
@@ -389,6 +388,7 @@ ENTRY(pgm_check_handler)
 	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
+	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
 	j	2f
 1:	UPDATE_VTIME %r14,__LC_SYNC_ENTER_TIMER
 	LAST_BREAK %r14
@@ -398,8 +398,7 @@ ENTRY(pgm_check_handler)
 	tm	__LC_PGM_ILC+2,0x02	# check for transaction abort
 	jz	2f
 	mvc	__THREAD_trap_tdb(256,%r14),0(%r13)
-2:	aghi	%r15,-(STACK_FRAME_OVERHEAD + __PT_SIZE)
-	la	%r11,STACK_FRAME_OVERHEAD(%r15)
+2:	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)
@@ -526,7 +525,6 @@ io_work:
 #
 io_work_user:
 	lg	%r1,__LC_KERNEL_STACK
-	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)
@@ -688,7 +686,6 @@ mcck_skip:
 	tm	__PT_PSW+1(%r11),0x01	# returning to user ?
 	jno	mcck_return
 	lg	%r1,__LC_KERNEL_STACK	# switch to kernel stack
-	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)
@@ -755,14 +752,12 @@ ENTRY(restart_int_handler)
  * Setup a pt_regs so that show_trace can provide a good call trace.
  */
 stack_overflow:
-	lg	%r11,__LC_PANIC_STACK	# change to panic stack
-	aghi	%r11,-__PT_SIZE		# create pt_regs
+	lg	%r15,__LC_PANIC_STACK	# change to panic stack
+	la	%r11,STACK_FRAME_OVERHEAD(%r15)
 	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
@@ -846,15 +841,14 @@ cleanup_system_call:
 	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
+	la	%r9,STACK_FRAME_OVERHEAD(%r15)
+	stg	%r9,24(%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_INT_CODE(4,%r15),__LC_SVC_ILC
+	mvc	__PT_R8(64,%r9),__LC_SAVE_AREA_SYNC
+	stmg	%r0,%r7,__PT_R0(%r9)
+	mvc	__PT_PSW(16,%r9),__LC_SVC_OLD_PSW
+	mvc	__PT_INT_CODE(4,%r9),__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

arch/s390/kernel/setup.c

@@ -377,11 +377,14 @@ static void __init setup_lowcore(void)
 		PSW_MASK_DAT | PSW_MASK_MCHECK;
 	lc->io_new_psw.addr = PSW_ADDR_AMODE | (unsigned long) io_int_handler;
 	lc->clock_comparator = -1ULL;
-	lc->kernel_stack = ((unsigned long) &init_thread_union) + THREAD_SIZE;
+	lc->kernel_stack = ((unsigned long) &init_thread_union)
+		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 	lc->async_stack = (unsigned long)
-		__alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0) + ASYNC_SIZE;
+		__alloc_bootmem(ASYNC_SIZE, ASYNC_SIZE, 0)
+		+ ASYNC_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 	lc->panic_stack = (unsigned long)
-		__alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0) + PAGE_SIZE;
+		__alloc_bootmem(PAGE_SIZE, PAGE_SIZE, 0)
+		+ PAGE_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 	lc->current_task = (unsigned long) init_thread_union.thread_info.task;
 	lc->thread_info = (unsigned long) &init_thread_union;
 	lc->machine_flags = S390_lowcore.machine_flags;

arch/s390/kernel/smp.c

@@ -181,8 +181,10 @@ static int __cpuinit pcpu_alloc_lowcore(struct pcpu *pcpu, int cpu)
 	lc = pcpu->lowcore;
 	memcpy(lc, &S390_lowcore, 512);
 	memset((char *) lc + 512, 0, sizeof(*lc) - 512);
-	lc->async_stack = pcpu->async_stack + ASYNC_SIZE;
-	lc->panic_stack = pcpu->panic_stack + PAGE_SIZE;
+	lc->async_stack = pcpu->async_stack + ASYNC_SIZE
+		- STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
+	lc->panic_stack = pcpu->panic_stack + PAGE_SIZE
+		- STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 	lc->cpu_nr = cpu;
 #ifndef CONFIG_64BIT
 	if (MACHINE_HAS_IEEE) {
@@ -253,7 +255,8 @@ static void pcpu_attach_task(struct pcpu *pcpu, struct task_struct *tsk)
 	struct _lowcore *lc = pcpu->lowcore;
 	struct thread_info *ti = task_thread_info(tsk);
 
-	lc->kernel_stack = (unsigned long) task_stack_page(tsk) + THREAD_SIZE;
+	lc->kernel_stack = (unsigned long) task_stack_page(tsk)
+		+ THREAD_SIZE - STACK_FRAME_OVERHEAD - sizeof(struct pt_regs);
 	lc->thread_info = (unsigned long) task_thread_info(tsk);
 	lc->current_task = (unsigned long) tsk;
 	lc->user_timer = ti->user_timer;
@@ -810,8 +813,10 @@ void __init smp_prepare_boot_cpu(void)
 	pcpu->state = CPU_STATE_CONFIGURED;
 	pcpu->address = boot_cpu_address;
 	pcpu->lowcore = (struct _lowcore *)(unsigned long) store_prefix();
-	pcpu->async_stack = S390_lowcore.async_stack - ASYNC_SIZE;
-	pcpu->panic_stack = S390_lowcore.panic_stack - PAGE_SIZE;
+	pcpu->async_stack = S390_lowcore.async_stack - ASYNC_SIZE
+		+ STACK_FRAME_OVERHEAD + sizeof(struct pt_regs);
+	pcpu->panic_stack = S390_lowcore.panic_stack - PAGE_SIZE
+		+ STACK_FRAME_OVERHEAD + sizeof(struct pt_regs);
 	S390_lowcore.percpu_offset = __per_cpu_offset[0];
 	smp_cpu_set_polarization(0, POLARIZATION_UNKNOWN);
 	set_cpu_present(0, true);