[deliverable/linux.git] / include / asm-s390 / system.h

/*
 *  include/asm-s390/system.h
 *
 *  S390 version
 *    Copyright (C) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation
 *    Author(s): Martin Schwidefsky (schwidefsky@de.ibm.com),
 *
 *  Derived from "include/asm-i386/system.h"
 */

#ifndef __ASM_SYSTEM_H
#define __ASM_SYSTEM_H

#include <linux/kernel.h>
#include <asm/types.h>
#include <asm/ptrace.h>
#include <asm/setup.h>
#include <asm/processor.h>

#ifdef __KERNEL__

struct task_struct;

extern struct task_struct *__switch_to(void *, void *);

#ifdef __s390x__
#define __FLAG_SHIFT 56
#else /* ! __s390x__ */
#define __FLAG_SHIFT 24
#endif /* ! __s390x__ */

static inline void save_fp_regs(s390_fp_regs *fpregs)
{
	asm volatile (
		"   std   0,8(%1)\n"
		"   std   2,24(%1)\n"
		"   std   4,40(%1)\n"
		"   std   6,56(%1)"
		: "=m" (*fpregs) : "a" (fpregs), "m" (*fpregs) : "memory" );
	if (!MACHINE_HAS_IEEE)
		return;
	asm volatile(
		"   stfpc 0(%1)\n"
		"   std   1,16(%1)\n"
		"   std   3,32(%1)\n"
		"   std   5,48(%1)\n"
		"   std   7,64(%1)\n"
		"   std   8,72(%1)\n"
		"   std   9,80(%1)\n"
		"   std   10,88(%1)\n"
		"   std   11,96(%1)\n"
		"   std   12,104(%1)\n"
		"   std   13,112(%1)\n"
		"   std   14,120(%1)\n"
		"   std   15,128(%1)\n"
		: "=m" (*fpregs) : "a" (fpregs), "m" (*fpregs) : "memory" );
}

static inline void restore_fp_regs(s390_fp_regs *fpregs)
{
	asm volatile (
		"   ld    0,8(%0)\n"
		"   ld    2,24(%0)\n"
		"   ld    4,40(%0)\n"
		"   ld    6,56(%0)"
		: : "a" (fpregs), "m" (*fpregs) );
	if (!MACHINE_HAS_IEEE)
		return;
	asm volatile(
		"   lfpc  0(%0)\n"
		"   ld    1,16(%0)\n"
		"   ld    3,32(%0)\n"
		"   ld    5,48(%0)\n"
		"   ld    7,64(%0)\n"
		"   ld    8,72(%0)\n"
		"   ld    9,80(%0)\n"
		"   ld    10,88(%0)\n"
		"   ld    11,96(%0)\n"
		"   ld    12,104(%0)\n"
		"   ld    13,112(%0)\n"
		"   ld    14,120(%0)\n"
		"   ld    15,128(%0)\n"
		: : "a" (fpregs), "m" (*fpregs) );
}

static inline void save_access_regs(unsigned int *acrs)
{
	asm volatile ("stam 0,15,0(%0)" : : "a" (acrs) : "memory" );
}

static inline void restore_access_regs(unsigned int *acrs)
{
	asm volatile ("lam 0,15,0(%0)" : : "a" (acrs) );
}

#define switch_to(prev,next,last) do {					     \
	if (prev == next)						     \
		break;							     \
	save_fp_regs(&prev->thread.fp_regs);				     \
	restore_fp_regs(&next->thread.fp_regs);				     \
	save_access_regs(&prev->thread.acrs[0]);			     \
	restore_access_regs(&next->thread.acrs[0]);			     \
	prev = __switch_to(prev,next);					     \
} while (0)

/*
 * On SMP systems, when the scheduler does migration-cost autodetection,
 * it needs a way to flush as much of the CPU's caches as possible.
 *
 * TODO: fill this in!
 */
static inline void sched_cacheflush(void)
{
}

#ifdef CONFIG_VIRT_CPU_ACCOUNTING
extern void account_vtime(struct task_struct *);
extern void account_tick_vtime(struct task_struct *);
extern void account_system_vtime(struct task_struct *);
#else
#define account_vtime(x) do { /* empty */ } while (0)
#endif

#define finish_arch_switch(prev) do {					     \
	set_fs(current->thread.mm_segment);				     \
	account_vtime(prev);						     \
} while (0)

#define nop() __asm__ __volatile__ ("nop")

#define xchg(ptr,x) \
  ((__typeof__(*(ptr)))__xchg((unsigned long)(x),(void *)(ptr),sizeof(*(ptr))))

static inline unsigned long __xchg(unsigned long x, void * ptr, int size)
{
	unsigned long addr, old;
	int shift;

        switch (size) {
	case 1:
		addr = (unsigned long) ptr;
		shift = (3 ^ (addr & 3)) << 3;
		addr ^= addr & 3;
		asm volatile(
			"    l   %0,0(%4)\n"
			"0:  lr  0,%0\n"
			"    nr  0,%3\n"
			"    or  0,%2\n"
			"    cs  %0,0,0(%4)\n"
			"    jl  0b\n"
			: "=&d" (old), "=m" (*(int *) addr)
			: "d" (x << shift), "d" (~(255 << shift)), "a" (addr),
			  "m" (*(int *) addr) : "memory", "cc", "0" );
		x = old >> shift;
		break;
	case 2:
		addr = (unsigned long) ptr;
		shift = (2 ^ (addr & 2)) << 3;
		addr ^= addr & 2;
		asm volatile(
			"    l   %0,0(%4)\n"
			"0:  lr  0,%0\n"
			"    nr  0,%3\n"
			"    or  0,%2\n"
			"    cs  %0,0,0(%4)\n"
			"    jl  0b\n"
			: "=&d" (old), "=m" (*(int *) addr)
			: "d" (x << shift), "d" (~(65535 << shift)), "a" (addr),
			  "m" (*(int *) addr) : "memory", "cc", "0" );
		x = old >> shift;
		break;
	case 4:
		asm volatile (
			"    l   %0,0(%3)\n"
			"0:  cs  %0,%2,0(%3)\n"
			"    jl  0b\n"
			: "=&d" (old), "=m" (*(int *) ptr)
			: "d" (x), "a" (ptr), "m" (*(int *) ptr)
			: "memory", "cc" );
		x = old;
		break;
#ifdef __s390x__
	case 8:
		asm volatile (
			"    lg  %0,0(%3)\n"
			"0:  csg %0,%2,0(%3)\n"
			"    jl  0b\n"
			: "=&d" (old), "=m" (*(long *) ptr)
			: "d" (x), "a" (ptr), "m" (*(long *) ptr)
			: "memory", "cc" );
		x = old;
		break;
#endif /* __s390x__ */
        }
        return x;
}

/*
 * Atomic compare and exchange.  Compare OLD with MEM, if identical,
 * store NEW in MEM.  Return the initial value in MEM.  Success is
 * indicated by comparing RETURN with OLD.
 */

#define __HAVE_ARCH_CMPXCHG 1

#define cmpxchg(ptr,o,n)\
	((__typeof__(*(ptr)))__cmpxchg((ptr),(unsigned long)(o),\
					(unsigned long)(n),sizeof(*(ptr))))

static inline unsigned long
__cmpxchg(volatile void *ptr, unsigned long old, unsigned long new, int size)
{
	unsigned long addr, prev, tmp;
	int shift;

        switch (size) {
	case 1:
		addr = (unsigned long) ptr;
		shift = (3 ^ (addr & 3)) << 3;
		addr ^= addr & 3;
		asm volatile(
			"    l   %0,0(%4)\n"
			"0:  nr  %0,%5\n"
                        "    lr  %1,%0\n"
			"    or  %0,%2\n"
			"    or  %1,%3\n"
			"    cs  %0,%1,0(%4)\n"
			"    jnl 1f\n"
			"    xr  %1,%0\n"
			"    nr  %1,%5\n"
			"    jnz 0b\n"
			"1:"
			: "=&d" (prev), "=&d" (tmp)
			: "d" (old << shift), "d" (new << shift), "a" (ptr),
			  "d" (~(255 << shift))
			: "memory", "cc" );
		return prev >> shift;
	case 2:
		addr = (unsigned long) ptr;
		shift = (2 ^ (addr & 2)) << 3;
		addr ^= addr & 2;
		asm volatile(
			"    l   %0,0(%4)\n"
			"0:  nr  %0,%5\n"
                        "    lr  %1,%0\n"
			"    or  %0,%2\n"
			"    or  %1,%3\n"
			"    cs  %0,%1,0(%4)\n"
			"    jnl 1f\n"
			"    xr  %1,%0\n"
			"    nr  %1,%5\n"
			"    jnz 0b\n"
			"1:"
			: "=&d" (prev), "=&d" (tmp)
			: "d" (old << shift), "d" (new << shift), "a" (ptr),
			  "d" (~(65535 << shift))
			: "memory", "cc" );
		return prev >> shift;
	case 4:
		asm volatile (
			"    cs  %0,%2,0(%3)\n"
			: "=&d" (prev) : "0" (old), "d" (new), "a" (ptr)
			: "memory", "cc" );
		return prev;
#ifdef __s390x__
	case 8:
		asm volatile (
			"    csg %0,%2,0(%3)\n"
			: "=&d" (prev) : "0" (old), "d" (new), "a" (ptr)
			: "memory", "cc" );
		return prev;
#endif /* __s390x__ */
        }
        return old;
}

/*
 * Force strict CPU ordering.
 * And yes, this is required on UP too when we're talking
 * to devices.
 *
 * This is very similar to the ppc eieio/sync instruction in that is
 * does a checkpoint syncronisation & makes sure that 
 * all memory ops have completed wrt other CPU's ( see 7-15 POP  DJB ).
 */

#define eieio()  __asm__ __volatile__ ( "bcr 15,0" : : : "memory" ) 
# define SYNC_OTHER_CORES(x)   eieio() 
#define mb()    eieio()
#define rmb()   eieio()
#define wmb()   eieio()
#define read_barrier_depends() do { } while(0)
#define smp_mb()       mb()
#define smp_rmb()      rmb()
#define smp_wmb()      wmb()
#define smp_read_barrier_depends()    read_barrier_depends()
#define smp_mb__before_clear_bit()     smp_mb()
#define smp_mb__after_clear_bit()      smp_mb()


#define set_mb(var, value)      do { var = value; mb(); } while (0)

#ifdef __s390x__

#define __ctl_load(array, low, high) ({ \
	typedef struct { char _[sizeof(array)]; } addrtype; \
	__asm__ __volatile__ ( \
		"   bras  1,0f\n" \
                "   lctlg 0,0,0(%0)\n" \
		"0: ex    %1,0(1)" \
		: : "a" (&array), "a" (((low)<<4)+(high)), \
		    "m" (*(addrtype *)(array)) : "1" ); \
	})

#define __ctl_store(array, low, high) ({ \
	typedef struct { char _[sizeof(array)]; } addrtype; \
	__asm__ __volatile__ ( \
		"   bras  1,0f\n" \
		"   stctg 0,0,0(%1)\n" \
		"0: ex    %2,0(1)" \
		: "=m" (*(addrtype *)(array)) \
		: "a" (&array), "a" (((low)<<4)+(high)) : "1" ); \
	})

#define __ctl_set_bit(cr, bit) ({ \
        __u8 __dummy[24]; \
        __asm__ __volatile__ ( \
                "    bras  1,0f\n"       /* skip indirect insns */ \
                "    stctg 0,0,0(%1)\n" \
                "    lctlg 0,0,0(%1)\n" \
                "0:  ex    %2,0(1)\n"    /* execute stctl */ \
                "    lg    0,0(%1)\n" \
                "    ogr   0,%3\n"       /* set the bit */ \
                "    stg   0,0(%1)\n" \
                "1:  ex    %2,6(1)"      /* execute lctl */ \
                : "=m" (__dummy) \
		: "a" ((((unsigned long) &__dummy) + 7) & ~7UL), \
		  "a" (cr*17), "a" (1L<<(bit)) \
                : "cc", "0", "1" ); \
        })

#define __ctl_clear_bit(cr, bit) ({ \
        __u8 __dummy[16]; \
        __asm__ __volatile__ ( \
                "    bras  1,0f\n"       /* skip indirect insns */ \
                "    stctg 0,0,0(%1)\n" \
                "    lctlg 0,0,0(%1)\n" \
                "0:  ex    %2,0(1)\n"    /* execute stctl */ \
                "    lg    0,0(%1)\n" \
                "    ngr   0,%3\n"       /* set the bit */ \
                "    stg   0,0(%1)\n" \
                "1:  ex    %2,6(1)"      /* execute lctl */ \
                : "=m" (__dummy) \
		: "a" ((((unsigned long) &__dummy) + 7) & ~7UL), \
		  "a" (cr*17), "a" (~(1L<<(bit))) \
                : "cc", "0", "1" ); \
        })

#else /* __s390x__ */

#define __ctl_load(array, low, high) ({ \
	typedef struct { char _[sizeof(array)]; } addrtype; \
	__asm__ __volatile__ ( \
		"   bras  1,0f\n" \
                "   lctl 0,0,0(%0)\n" \
		"0: ex    %1,0(1)" \
		: : "a" (&array), "a" (((low)<<4)+(high)), \
		    "m" (*(addrtype *)(array)) : "1" ); \
	})

#define __ctl_store(array, low, high) ({ \
	typedef struct { char _[sizeof(array)]; } addrtype; \
	__asm__ __volatile__ ( \
		"   bras  1,0f\n" \
		"   stctl 0,0,0(%1)\n" \
		"0: ex    %2,0(1)" \
		: "=m" (*(addrtype *)(array)) \
		: "a" (&array), "a" (((low)<<4)+(high)): "1" ); \
	})

#define __ctl_set_bit(cr, bit) ({ \
        __u8 __dummy[16]; \
        __asm__ __volatile__ ( \
                "    bras  1,0f\n"       /* skip indirect insns */ \
                "    stctl 0,0,0(%1)\n" \
                "    lctl  0,0,0(%1)\n" \
                "0:  ex    %2,0(1)\n"    /* execute stctl */ \
                "    l     0,0(%1)\n" \
                "    or    0,%3\n"       /* set the bit */ \
                "    st    0,0(%1)\n" \
                "1:  ex    %2,4(1)"      /* execute lctl */ \
                : "=m" (__dummy) \
		: "a" ((((unsigned long) &__dummy) + 7) & ~7UL), \
		  "a" (cr*17), "a" (1<<(bit)) \
                : "cc", "0", "1" ); \
        })

#define __ctl_clear_bit(cr, bit) ({ \
        __u8 __dummy[16]; \
        __asm__ __volatile__ ( \
                "    bras  1,0f\n"       /* skip indirect insns */ \
                "    stctl 0,0,0(%1)\n" \
                "    lctl  0,0,0(%1)\n" \
                "0:  ex    %2,0(1)\n"    /* execute stctl */ \
                "    l     0,0(%1)\n" \
                "    nr    0,%3\n"       /* set the bit */ \
                "    st    0,0(%1)\n" \
                "1:  ex    %2,4(1)"      /* execute lctl */ \
                : "=m" (__dummy) \
		: "a" ((((unsigned long) &__dummy) + 7) & ~7UL), \
		  "a" (cr*17), "a" (~(1<<(bit))) \
                : "cc", "0", "1" ); \
        })
#endif /* __s390x__ */

#include <linux/irqflags.h>

/*
 * Use to set psw mask except for the first byte which
 * won't be changed by this function.
 */
static inline void
__set_psw_mask(unsigned long mask)
{
	local_save_flags(mask);
	__load_psw_mask(mask);
}

#define local_mcck_enable()  __set_psw_mask(PSW_KERNEL_BITS)
#define local_mcck_disable() __set_psw_mask(PSW_KERNEL_BITS & ~PSW_MASK_MCHECK)

#ifdef CONFIG_SMP

extern void smp_ctl_set_bit(int cr, int bit);
extern void smp_ctl_clear_bit(int cr, int bit);
#define ctl_set_bit(cr, bit) smp_ctl_set_bit(cr, bit)
#define ctl_clear_bit(cr, bit) smp_ctl_clear_bit(cr, bit)

#else

#define ctl_set_bit(cr, bit) __ctl_set_bit(cr, bit)
#define ctl_clear_bit(cr, bit) __ctl_clear_bit(cr, bit)

#endif /* CONFIG_SMP */

extern void (*_machine_restart)(char *command);
extern void (*_machine_halt)(void);
extern void (*_machine_power_off)(void);

#define arch_align_stack(x) (x)

#endif /* __KERNEL__ */

#endif
Commit	Line	Data
1da177e4 LT	1	/*
	2	* include/asm-s390/system.h
	3	*
	4	* S390 version
	5	* Copyright (C) 1999 IBM Deutschland Entwicklung GmbH, IBM Corporation
	6	* Author(s): Martin Schwidefsky (schwidefsky@de.ibm.com),
	7	*
	8	* Derived from "include/asm-i386/system.h"
	9	*/
	10
	11	#ifndef __ASM_SYSTEM_H
	12	#define __ASM_SYSTEM_H
	13
1da177e4 LT	14	#include <linux/kernel.h>
	15	#include <asm/types.h>
	16	#include <asm/ptrace.h>
	17	#include <asm/setup.h>
77fa2245	18	#include <asm/processor.h>
1da177e4 LT	19
	20	#ifdef __KERNEL__
	21
	22	struct task_struct;
	23
	24	extern struct task_struct __switch_to(void , void *);
	25
	26	#ifdef __s390x__
	27	#define __FLAG_SHIFT 56
	28	#else /* ! __s390x__ */
	29	#define __FLAG_SHIFT 24
	30	#endif /* ! __s390x__ */
	31
	32	static inline void save_fp_regs(s390_fp_regs *fpregs)
	33	{
	34	asm volatile (
	35	" std 0,8(%1)\n"
	36	" std 2,24(%1)\n"
	37	" std 4,40(%1)\n"
	38	" std 6,56(%1)"
	39	: "=m" (fpregs) : "a" (fpregs), "m" (fpregs) : "memory" );
	40	if (!MACHINE_HAS_IEEE)
	41	return;
	42	asm volatile(
	43	" stfpc 0(%1)\n"
	44	" std 1,16(%1)\n"
	45	" std 3,32(%1)\n"
	46	" std 5,48(%1)\n"
	47	" std 7,64(%1)\n"
	48	" std 8,72(%1)\n"
	49	" std 9,80(%1)\n"
	50	" std 10,88(%1)\n"
	51	" std 11,96(%1)\n"
	52	" std 12,104(%1)\n"
	53	" std 13,112(%1)\n"
	54	" std 14,120(%1)\n"
	55	" std 15,128(%1)\n"
	56	: "=m" (fpregs) : "a" (fpregs), "m" (fpregs) : "memory" );
	57	}
	58
	59	static inline void restore_fp_regs(s390_fp_regs *fpregs)
	60	{
	61	asm volatile (
	62	" ld 0,8(%0)\n"
	63	" ld 2,24(%0)\n"
	64	" ld 4,40(%0)\n"
	65	" ld 6,56(%0)"
	66	: : "a" (fpregs), "m" (*fpregs) );
	67	if (!MACHINE_HAS_IEEE)
	68	return;
	69	asm volatile(
	70	" lfpc 0(%0)\n"
	71	" ld 1,16(%0)\n"
	72	" ld 3,32(%0)\n"
	73	" ld 5,48(%0)\n"
	74	" ld 7,64(%0)\n"
	75	" ld 8,72(%0)\n"
	76	" ld 9,80(%0)\n"
	77	" ld 10,88(%0)\n"
	78	" ld 11,96(%0)\n"
	79	" ld 12,104(%0)\n"
	80	" ld 13,112(%0)\n"
	81	" ld 14,120(%0)\n"
	82	" ld 15,128(%0)\n"
83	: : "a" (fpregs), "m" (*fpregs) );
84	}
85
86	static inline void save_access_regs(unsigned int *acrs)
87	{
88	asm volatile ("stam 0,15,0(%0)" : : "a" (acrs) : "memory" );
89	}
90
91	static inline void restore_access_regs(unsigned int *acrs)
92	{
93	asm volatile ("lam 0,15,0(%0)" : : "a" (acrs) );
94	}
95
96	#define switch_to(prev,next,last) do { \
97	if (prev == next) \
98	break; \
99	save_fp_regs(&prev->thread.fp_regs); \
100	restore_fp_regs(&next->thread.fp_regs); \
101	save_access_regs(&prev->thread.acrs[0]); \
102	restore_access_regs(&next->thread.acrs[0]); \
103	prev = __switch_to(prev,next); \
104	} while (0)
105
4dc7a0bb IM	106	/*
	107	* On SMP systems, when the scheduler does migration-cost autodetection,
	108	* it needs a way to flush as much of the CPU's caches as possible.
	109	*
	110	* TODO: fill this in!
	111	*/
	112	static inline void sched_cacheflush(void)
	113	{
	114	}
	115
1da177e4	116	#ifdef CONFIG_VIRT_CPU_ACCOUNTING
1f1c12af MS	117	extern void account_vtime(struct task_struct *);
1f1c12af MS	118	extern void account_tick_vtime(struct task_struct *);
1da177e4	119	extern void account_system_vtime(struct task_struct *);
5c833890 JB	120	#else
5c833890 JB	121	#define account_vtime(x) do { /* empty */ } while (0)
4866cde0	122	#endif
1da177e4	123
5ee24d95	124	#define finish_arch_switch(prev) do { \
1da177e4	125	set_fs(current->thread.mm_segment); \
1f1c12af	126	account_vtime(prev); \
1da177e4 LT	127	} while (0)
1da177e4 LT	128
1da177e4 LT	129	#define nop() __asm__ __volatile__ ("nop")
	130
	131	#define xchg(ptr,x) \
	132	((__typeof__((ptr)))__xchg((unsigned long)(x),(void )(ptr),sizeof(*(ptr))))
	133
	134	static inline unsigned long __xchg(unsigned long x, void * ptr, int size)
	135	{
	136	unsigned long addr, old;
	137	int shift;
	138
	139	switch (size) {
	140	case 1:
	141	addr = (unsigned long) ptr;
	142	shift = (3 ^ (addr & 3)) << 3;
	143	addr ^= addr & 3;
	144	asm volatile(
	145	" l %0,0(%4)\n"
	146	"0: lr 0,%0\n"
	147	" nr 0,%3\n"
	148	" or 0,%2\n"
	149	" cs %0,0,0(%4)\n"
	150	" jl 0b\n"
	151	: "=&d" (old), "=m" ((int ) addr)
	152	: "d" (x << shift), "d" (~(255 << shift)), "a" (addr),
	153	"m" ((int ) addr) : "memory", "cc", "0" );
	154	x = old >> shift;
	155	break;
	156	case 2:
	157	addr = (unsigned long) ptr;
	158	shift = (2 ^ (addr & 2)) << 3;
	159	addr ^= addr & 2;
	160	asm volatile(
	161	" l %0,0(%4)\n"
	162	"0: lr 0,%0\n"
	163	" nr 0,%3\n"
	164	" or 0,%2\n"
	165	" cs %0,0,0(%4)\n"
	166	" jl 0b\n"
	167	: "=&d" (old), "=m" ((int ) addr)
	168	: "d" (x << shift), "d" (~(65535 << shift)), "a" (addr),
	169	"m" ((int ) addr) : "memory", "cc", "0" );
	170	x = old >> shift;
	171	break;
	172	case 4:
	173	asm volatile (
	174	" l %0,0(%3)\n"
	175	"0: cs %0,%2,0(%3)\n"
	176	" jl 0b\n"
	177	: "=&d" (old), "=m" ((int ) ptr)
	178	: "d" (x), "a" (ptr), "m" ((int ) ptr)
	179	: "memory", "cc" );
	180	x = old;
	181	break;
	182	#ifdef __s390x__
	183	case 8:
	184	asm volatile (
	185	" lg %0,0(%3)\n"
	186	"0: csg %0,%2,0(%3)\n"
	187	" jl 0b\n"
	188	: "=&d" (old), "=m" ((long ) ptr)
	189	: "d" (x), "a" (ptr), "m" ((long ) ptr)
	190	: "memory", "cc" );
	191	x = old;
	192	break;
193	#endif /* __s390x__ */
194	}
195	return x;
196	}
197
198	/*
199	* Atomic compare and exchange. Compare OLD with MEM, if identical,
200	* store NEW in MEM. Return the initial value in MEM. Success is
201	* indicated by comparing RETURN with OLD.
202	*/
203
204	#define __HAVE_ARCH_CMPXCHG 1
205
206	#define cmpxchg(ptr,o,n)\
207	((__typeof__(*(ptr)))__cmpxchg((ptr),(unsigned long)(o),\
208	(unsigned long)(n),sizeof(*(ptr))))
209
210	static inline unsigned long
211	__cmpxchg(volatile void *ptr, unsigned long old, unsigned long new, int size)
212	{
213	unsigned long addr, prev, tmp;
214	int shift;
215
216	switch (size) {
217	case 1:
218	addr = (unsigned long) ptr;
219	shift = (3 ^ (addr & 3)) << 3;
220	addr ^= addr & 3;
221	asm volatile(
222	" l %0,0(%4)\n"
223	"0: nr %0,%5\n"
224	" lr %1,%0\n"
225	" or %0,%2\n"
226	" or %1,%3\n"
227	" cs %0,%1,0(%4)\n"
228	" jnl 1f\n"
229	" xr %1,%0\n"
230	" nr %1,%5\n"
231	" jnz 0b\n"
232	"1:"
233	: "=&d" (prev), "=&d" (tmp)
234	: "d" (old << shift), "d" (new << shift), "a" (ptr),
235	"d" (~(255 << shift))
236	: "memory", "cc" );
237	return prev >> shift;
238	case 2:
239	addr = (unsigned long) ptr;
240	shift = (2 ^ (addr & 2)) << 3;
241	addr ^= addr & 2;
242	asm volatile(
243	" l %0,0(%4)\n"
244	"0: nr %0,%5\n"
245	" lr %1,%0\n"
246	" or %0,%2\n"
247	" or %1,%3\n"
248	" cs %0,%1,0(%4)\n"
249	" jnl 1f\n"
250	" xr %1,%0\n"
251	" nr %1,%5\n"
252	" jnz 0b\n"
253	"1:"
254	: "=&d" (prev), "=&d" (tmp)
255	: "d" (old << shift), "d" (new << shift), "a" (ptr),
256	"d" (~(65535 << shift))
257	: "memory", "cc" );
258	return prev >> shift;
259	case 4:
260	asm volatile (
261	" cs %0,%2,0(%3)\n"
262	: "=&d" (prev) : "0" (old), "d" (new), "a" (ptr)
263	: "memory", "cc" );
264	return prev;
265	#ifdef __s390x__
266	case 8:
267	asm volatile (
268	" csg %0,%2,0(%3)\n"
269	: "=&d" (prev) : "0" (old), "d" (new), "a" (ptr)
270	: "memory", "cc" );
271	return prev;
272	#endif /* __s390x__ */
273	}
274	return old;
275	}
276
277	/*
278	* Force strict CPU ordering.
279	* And yes, this is required on UP too when we're talking
280	* to devices.
281	*
282	* This is very similar to the ppc eieio/sync instruction in that is
283	* does a checkpoint syncronisation & makes sure that
284	* all memory ops have completed wrt other CPU's ( see 7-15 POP DJB ).
285	*/
286
287	#define eieio() __asm__ __volatile__ ( "bcr 15,0" : : : "memory" )
288	# define SYNC_OTHER_CORES(x) eieio()
289	#define mb() eieio()
290	#define rmb() eieio()
291	#define wmb() eieio()
292	#define read_barrier_depends() do { } while(0)
293	#define smp_mb() mb()
294	#define smp_rmb() rmb()
295	#define smp_wmb() wmb()
296	#define smp_read_barrier_depends() read_barrier_depends()
297	#define smp_mb__before_clear_bit() smp_mb()
298	#define smp_mb__after_clear_bit() smp_mb()
299
300
301	#define set_mb(var, value) do { var = value; mb(); } while (0)
1da177e4	302
1da177e4 LT	303	#ifdef __s390x__
1da177e4 LT	304
1da177e4 LT	305	#define __ctl_load(array, low, high) ({ \
	306	typedef struct { char _[sizeof(array)]; } addrtype; \
	307	__asm__ __volatile__ ( \
	308	" bras 1,0f\n" \
	309	" lctlg 0,0,0(%0)\n" \
	310	"0: ex %1,0(1)" \
	311	: : "a" (&array), "a" (((low)<<4)+(high)), \
	312	"m" ((addrtype )(array)) : "1" ); \
	313	})
	314
	315	#define __ctl_store(array, low, high) ({ \
	316	typedef struct { char _[sizeof(array)]; } addrtype; \
	317	__asm__ __volatile__ ( \
	318	" bras 1,0f\n" \
	319	" stctg 0,0,0(%1)\n" \
	320	"0: ex %2,0(1)" \
	321	: "=m" ((addrtype )(array)) \
	322	: "a" (&array), "a" (((low)<<4)+(high)) : "1" ); \
	323	})
	324
	325	#define __ctl_set_bit(cr, bit) ({ \
	326	__u8 __dummy[24]; \
	327	__asm__ __volatile__ ( \
	328	" bras 1,0f\n" /* skip indirect insns */ \
	329	" stctg 0,0,0(%1)\n" \
	330	" lctlg 0,0,0(%1)\n" \
	331	"0: ex %2,0(1)\n" /* execute stctl */ \
	332	" lg 0,0(%1)\n" \
	333	" ogr 0,%3\n" /* set the bit */ \
	334	" stg 0,0(%1)\n" \
	335	"1: ex %2,6(1)" /* execute lctl */ \
	336	: "=m" (__dummy) \
	337	: "a" ((((unsigned long) &__dummy) + 7) & ~7UL), \
	338	"a" (cr*17), "a" (1L<<(bit)) \
	339	: "cc", "0", "1" ); \
	340	})
	341
	342	#define __ctl_clear_bit(cr, bit) ({ \
	343	__u8 __dummy[16]; \
	344	__asm__ __volatile__ ( \
	345	" bras 1,0f\n" /* skip indirect insns */ \
	346	" stctg 0,0,0(%1)\n" \
	347	" lctlg 0,0,0(%1)\n" \
	348	"0: ex %2,0(1)\n" /* execute stctl */ \
	349	" lg 0,0(%1)\n" \
	350	" ngr 0,%3\n" /* set the bit */ \
	351	" stg 0,0(%1)\n" \
	352	"1: ex %2,6(1)" /* execute lctl */ \
	353	: "=m" (__dummy) \
	354	: "a" ((((unsigned long) &__dummy) + 7) & ~7UL), \
	355	"a" (cr*17), "a" (~(1L<<(bit))) \
	356	: "cc", "0", "1" ); \
	357	})
	358
	359	#else /* __s390x__ */
	360
1da177e4 LT	361	#define __ctl_load(array, low, high) ({ \
	362	typedef struct { char _[sizeof(array)]; } addrtype; \
	363	__asm__ __volatile__ ( \
	364	" bras 1,0f\n" \
	365	" lctl 0,0,0(%0)\n" \
	366	"0: ex %1,0(1)" \
	367	: : "a" (&array), "a" (((low)<<4)+(high)), \
	368	"m" ((addrtype )(array)) : "1" ); \
	369	})
	370
	371	#define __ctl_store(array, low, high) ({ \
	372	typedef struct { char _[sizeof(array)]; } addrtype; \
	373	__asm__ __volatile__ ( \
	374	" bras 1,0f\n" \
	375	" stctl 0,0,0(%1)\n" \
	376	"0: ex %2,0(1)" \
	377	: "=m" ((addrtype )(array)) \
	378	: "a" (&array), "a" (((low)<<4)+(high)): "1" ); \
	379	})
	380
	381	#define __ctl_set_bit(cr, bit) ({ \
	382	__u8 __dummy[16]; \
	383	__asm__ __volatile__ ( \
	384	" bras 1,0f\n" /* skip indirect insns */ \
	385	" stctl 0,0,0(%1)\n" \
	386	" lctl 0,0,0(%1)\n" \
	387	"0: ex %2,0(1)\n" /* execute stctl */ \
	388	" l 0,0(%1)\n" \
	389	" or 0,%3\n" /* set the bit */ \
	390	" st 0,0(%1)\n" \
	391	"1: ex %2,4(1)" /* execute lctl */ \
	392	: "=m" (__dummy) \
	393	: "a" ((((unsigned long) &__dummy) + 7) & ~7UL), \
	394	"a" (cr*17), "a" (1<<(bit)) \
	395	: "cc", "0", "1" ); \
	396	})
	397
	398	#define __ctl_clear_bit(cr, bit) ({ \
	399	__u8 __dummy[16]; \
	400	__asm__ __volatile__ ( \
	401	" bras 1,0f\n" /* skip indirect insns */ \
	402	" stctl 0,0,0(%1)\n" \
	403	" lctl 0,0,0(%1)\n" \
	404	"0: ex %2,0(1)\n" /* execute stctl */ \
	405	" l 0,0(%1)\n" \
	406	" nr 0,%3\n" /* set the bit */ \
	407	" st 0,0(%1)\n" \
	408	"1: ex %2,4(1)" /* execute lctl */ \
	409	: "=m" (__dummy) \
	410	: "a" ((((unsigned long) &__dummy) + 7) & ~7UL), \
	411	"a" (cr*17), "a" (~(1<<(bit))) \
	412	: "cc", "0", "1" ); \
	413	})
	414	#endif /* __s390x__ */
	415
1f194a4c	416	#include <linux/irqflags.h>
1da177e4	417
77fa2245 HC	418	/*
	419	* Use to set psw mask except for the first byte which
	420	* won't be changed by this function.
	421	*/
	422	static inline void
	423	__set_psw_mask(unsigned long mask)
	424	{
	425	local_save_flags(mask);
	426	__load_psw_mask(mask);
	427	}
	428
	429	#define local_mcck_enable() __set_psw_mask(PSW_KERNEL_BITS)
	430	#define local_mcck_disable() __set_psw_mask(PSW_KERNEL_BITS & ~PSW_MASK_MCHECK)
	431
1da177e4 LT	432	#ifdef CONFIG_SMP
	433
	434	extern void smp_ctl_set_bit(int cr, int bit);
	435	extern void smp_ctl_clear_bit(int cr, int bit);
	436	#define ctl_set_bit(cr, bit) smp_ctl_set_bit(cr, bit)
	437	#define ctl_clear_bit(cr, bit) smp_ctl_clear_bit(cr, bit)
	438
	439	#else
	440
	441	#define ctl_set_bit(cr, bit) __ctl_set_bit(cr, bit)
	442	#define ctl_clear_bit(cr, bit) __ctl_clear_bit(cr, bit)
	443
	444	#endif /* CONFIG_SMP */
	445
	446	extern void (_machine_restart)(char command);
	447	extern void (*_machine_halt)(void);
	448	extern void (*_machine_power_off)(void);
	449
	450	#define arch_align_stack(x) (x)
	451
	452	#endif /* __KERNEL__ */
	453
	454	#endif