[linux.git] / include / asm-ia64 / spinlock.h

#ifndef _ASM_IA64_SPINLOCK_H
#define _ASM_IA64_SPINLOCK_H

/*
 * Copyright (C) 1998-2003 Hewlett-Packard Co
 *	David Mosberger-Tang <[email protected]>
 * Copyright (C) 1999 Walt Drummond <[email protected]>
 *
 * This file is used for SMP configurations only.
 */

#include <linux/compiler.h>
#include <linux/kernel.h>

#include <asm/atomic.h>
#include <asm/bitops.h>
#include <asm/intrinsics.h>
#include <asm/system.h>

#define __raw_spin_lock_init(x)			((x)->lock = 0)

#ifdef ASM_SUPPORTED
/*
 * Try to get the lock.  If we fail to get the lock, make a non-standard call to
 * ia64_spinlock_contention().  We do not use a normal call because that would force all
 * callers of __raw_spin_lock() to be non-leaf routines.  Instead, ia64_spinlock_contention() is
 * carefully coded to touch only those registers that __raw_spin_lock() marks "clobbered".
 */

#define IA64_SPINLOCK_CLOBBERS "ar.ccv", "ar.pfs", "p14", "p15", "r27", "r28", "r29", "r30", "b6", "memory"

static inline void
__raw_spin_lock_flags (raw_spinlock_t *lock, unsigned long flags)
{
	register volatile unsigned int *ptr asm ("r31") = &lock->lock;

#if (__GNUC__ == 3 && __GNUC_MINOR__ < 3)
# ifdef CONFIG_ITANIUM
	/* don't use brl on Itanium... */
	asm volatile ("{\n\t"
		      "  mov ar.ccv = r0\n\t"
		      "  mov r28 = ip\n\t"
		      "  mov r30 = 1;;\n\t"
		      "}\n\t"
		      "cmpxchg4.acq r30 = [%1], r30, ar.ccv\n\t"
		      "movl r29 = ia64_spinlock_contention_pre3_4;;\n\t"
		      "cmp4.ne p14, p0 = r30, r0\n\t"
		      "mov b6 = r29;;\n\t"
		      "mov r27=%2\n\t"
		      "(p14) br.cond.spnt.many b6"
		      : "=r"(ptr) : "r"(ptr), "r" (flags) : IA64_SPINLOCK_CLOBBERS);
# else
	asm volatile ("{\n\t"
		      "  mov ar.ccv = r0\n\t"
		      "  mov r28 = ip\n\t"
		      "  mov r30 = 1;;\n\t"
		      "}\n\t"
		      "cmpxchg4.acq r30 = [%1], r30, ar.ccv;;\n\t"
		      "cmp4.ne p14, p0 = r30, r0\n\t"
		      "mov r27=%2\n\t"
		      "(p14) brl.cond.spnt.many ia64_spinlock_contention_pre3_4;;"
		      : "=r"(ptr) : "r"(ptr), "r" (flags) : IA64_SPINLOCK_CLOBBERS);
# endif /* CONFIG_MCKINLEY */
#else
# ifdef CONFIG_ITANIUM
	/* don't use brl on Itanium... */
	/* mis-declare, so we get the entry-point, not it's function descriptor: */
	asm volatile ("mov r30 = 1\n\t"
		      "mov r27=%2\n\t"
		      "mov ar.ccv = r0;;\n\t"
		      "cmpxchg4.acq r30 = [%0], r30, ar.ccv\n\t"
		      "movl r29 = ia64_spinlock_contention;;\n\t"
		      "cmp4.ne p14, p0 = r30, r0\n\t"
		      "mov b6 = r29;;\n\t"
		      "(p14) br.call.spnt.many b6 = b6"
		      : "=r"(ptr) : "r"(ptr), "r" (flags) : IA64_SPINLOCK_CLOBBERS);
# else
	asm volatile ("mov r30 = 1\n\t"
		      "mov r27=%2\n\t"
		      "mov ar.ccv = r0;;\n\t"
		      "cmpxchg4.acq r30 = [%0], r30, ar.ccv;;\n\t"
		      "cmp4.ne p14, p0 = r30, r0\n\t"
		      "(p14) brl.call.spnt.many b6=ia64_spinlock_contention;;"
		      : "=r"(ptr) : "r"(ptr), "r" (flags) : IA64_SPINLOCK_CLOBBERS);
# endif /* CONFIG_MCKINLEY */
#endif
}

#define __raw_spin_lock(lock) __raw_spin_lock_flags(lock, 0)

/* Unlock by doing an ordered store and releasing the cacheline with nta */
static inline void __raw_spin_unlock(raw_spinlock_t *x) {
	barrier();
	asm volatile ("st4.rel.nta [%0] = r0\n\t" :: "r"(x));
}

#else /* !ASM_SUPPORTED */
#define __raw_spin_lock_flags(lock, flags) __raw_spin_lock(lock)
# define __raw_spin_lock(x)								\
do {											\
	__u32 *ia64_spinlock_ptr = (__u32 *) (x);					\
	__u64 ia64_spinlock_val;							\
	ia64_spinlock_val = ia64_cmpxchg4_acq(ia64_spinlock_ptr, 1, 0);			\
	if (unlikely(ia64_spinlock_val)) {						\
		do {									\
			while (*ia64_spinlock_ptr)					\
				ia64_barrier();						\
			ia64_spinlock_val = ia64_cmpxchg4_acq(ia64_spinlock_ptr, 1, 0);	\
		} while (ia64_spinlock_val);						\
	}										\
} while (0)
#define __raw_spin_unlock(x)	do { barrier(); ((raw_spinlock_t *) x)->lock = 0; } while (0)
#endif /* !ASM_SUPPORTED */

#define __raw_spin_is_locked(x)		((x)->lock != 0)
#define __raw_spin_trylock(x)		(cmpxchg_acq(&(x)->lock, 0, 1) == 0)
#define __raw_spin_unlock_wait(lock) \
	do { while (__raw_spin_is_locked(lock)) cpu_relax(); } while (0)

#define __raw_read_can_lock(rw)		(*(volatile int *)(rw) >= 0)
#define __raw_write_can_lock(rw)	(*(volatile int *)(rw) == 0)

#define __raw_read_lock(rw)								\
do {											\
	raw_rwlock_t *__read_lock_ptr = (rw);						\
											\
	while (unlikely(ia64_fetchadd(1, (int *) __read_lock_ptr, acq) < 0)) {		\
		ia64_fetchadd(-1, (int *) __read_lock_ptr, rel);			\
		while (*(volatile int *)__read_lock_ptr < 0)				\
			cpu_relax();							\
	}										\
} while (0)

#define __raw_read_unlock(rw)					\
do {								\
	raw_rwlock_t *__read_lock_ptr = (rw);			\
	ia64_fetchadd(-1, (int *) __read_lock_ptr, rel);	\
} while (0)

#ifdef ASM_SUPPORTED
#define __raw_write_lock(rw)							\
do {										\
 	__asm__ __volatile__ (							\
		"mov ar.ccv = r0\n"						\
		"dep r29 = -1, r0, 31, 1;;\n"					\
		"1:\n"								\
		"ld4 r2 = [%0];;\n"						\
		"cmp4.eq p0,p7 = r0,r2\n"					\
		"(p7) br.cond.spnt.few 1b \n"					\
		"cmpxchg4.acq r2 = [%0], r29, ar.ccv;;\n"			\
		"cmp4.eq p0,p7 = r0, r2\n"					\
		"(p7) br.cond.spnt.few 1b;;\n"					\
		:: "r"(rw) : "ar.ccv", "p7", "r2", "r29", "memory");		\
} while(0)

#define __raw_write_trylock(rw)							\
({										\
	register long result;							\
										\
	__asm__ __volatile__ (							\
		"mov ar.ccv = r0\n"						\
		"dep r29 = -1, r0, 31, 1;;\n"					\
		"cmpxchg4.acq %0 = [%1], r29, ar.ccv\n"				\
		: "=r"(result) : "r"(rw) : "ar.ccv", "r29", "memory");		\
	(result == 0);								\
})

static inline void __raw_write_unlock(raw_rwlock_t *x)
{
	u8 *y = (u8 *)x;
	barrier();
	asm volatile ("st1.rel.nta [%0] = r0\n\t" :: "r"(y+3) : "memory" );
}

#else /* !ASM_SUPPORTED */

#define __raw_write_lock(l)								\
({											\
	__u64 ia64_val, ia64_set_val = ia64_dep_mi(-1, 0, 31, 1);			\
	__u32 *ia64_write_lock_ptr = (__u32 *) (l);					\
	do {										\
		while (*ia64_write_lock_ptr)						\
			ia64_barrier();							\
		ia64_val = ia64_cmpxchg4_acq(ia64_write_lock_ptr, ia64_set_val, 0);	\
	} while (ia64_val);								\
})

#define __raw_write_trylock(rw)						\
({									\
	__u64 ia64_val;							\
	__u64 ia64_set_val = ia64_dep_mi(-1, 0, 31,1);			\
	ia64_val = ia64_cmpxchg4_acq((__u32 *)(rw), ia64_set_val, 0);	\
	(ia64_val == 0);						\
})

static inline void __raw_write_unlock(raw_rwlock_t *x)
{
	barrier();
	x->write_lock = 0;
}

#endif /* !ASM_SUPPORTED */

static inline int __raw_read_trylock(raw_rwlock_t *x)
{
	union {
		raw_rwlock_t lock;
		__u32 word;
	} old, new;
	old.lock = new.lock = *x;
	old.lock.write_lock = new.lock.write_lock = 0;
	++new.lock.read_counter;
	return (u32)ia64_cmpxchg4_acq((__u32 *)(x), new.word, old.word) == old.word;
}

#define _raw_spin_relax(lock)	cpu_relax()
#define _raw_read_relax(lock)	cpu_relax()
#define _raw_write_relax(lock)	cpu_relax()

#endif /*  _ASM_IA64_SPINLOCK_H */
Commit	Line	Data
1da177e4 LT	1	#ifndef _ASM_IA64_SPINLOCK_H
	2	#define _ASM_IA64_SPINLOCK_H
	3
	4	/*
	5	* Copyright (C) 1998-2003 Hewlett-Packard Co
	6	* David Mosberger-Tang <[email protected]>
	7	* Copyright (C) 1999 Walt Drummond <[email protected]>
	8	*
	9	* This file is used for SMP configurations only.
	10	*/
	11
	12	#include <linux/compiler.h>
	13	#include <linux/kernel.h>
	14
	15	#include <asm/atomic.h>
	16	#include <asm/bitops.h>
	17	#include <asm/intrinsics.h>
	18	#include <asm/system.h>
	19
fb1c8f93	20	#define __raw_spin_lock_init(x) ((x)->lock = 0)
1da177e4 LT	21
	22	#ifdef ASM_SUPPORTED
	23	/*
	24	* Try to get the lock. If we fail to get the lock, make a non-standard call to
	25	* ia64_spinlock_contention(). We do not use a normal call because that would force all
fb1c8f93 IM	26	* callers of __raw_spin_lock() to be non-leaf routines. Instead, ia64_spinlock_contention() is
fb1c8f93 IM	27	* carefully coded to touch only those registers that __raw_spin_lock() marks "clobbered".
1da177e4 LT	28	*/
	29
	30	#define IA64_SPINLOCK_CLOBBERS "ar.ccv", "ar.pfs", "p14", "p15", "r27", "r28", "r29", "r30", "b6", "memory"
	31
	32	static inline void
fb1c8f93	33	__raw_spin_lock_flags (raw_spinlock_t *lock, unsigned long flags)
1da177e4 LT	34	{
	35	register volatile unsigned int *ptr asm ("r31") = &lock->lock;
	36
a1365647	37	#if (__GNUC__ == 3 && __GNUC_MINOR__ < 3)
1da177e4 LT	38	# ifdef CONFIG_ITANIUM
	39	/* don't use brl on Itanium... */
	40	asm volatile ("{\n\t"
	41	" mov ar.ccv = r0\n\t"
	42	" mov r28 = ip\n\t"
	43	" mov r30 = 1;;\n\t"
	44	"}\n\t"
	45	"cmpxchg4.acq r30 = [%1], r30, ar.ccv\n\t"
	46	"movl r29 = ia64_spinlock_contention_pre3_4;;\n\t"
	47	"cmp4.ne p14, p0 = r30, r0\n\t"
	48	"mov b6 = r29;;\n\t"
	49	"mov r27=%2\n\t"
	50	"(p14) br.cond.spnt.many b6"
	51	: "=r"(ptr) : "r"(ptr), "r" (flags) : IA64_SPINLOCK_CLOBBERS);
	52	# else
	53	asm volatile ("{\n\t"
	54	" mov ar.ccv = r0\n\t"
	55	" mov r28 = ip\n\t"
	56	" mov r30 = 1;;\n\t"
	57	"}\n\t"
	58	"cmpxchg4.acq r30 = [%1], r30, ar.ccv;;\n\t"
	59	"cmp4.ne p14, p0 = r30, r0\n\t"
	60	"mov r27=%2\n\t"
	61	"(p14) brl.cond.spnt.many ia64_spinlock_contention_pre3_4;;"
	62	: "=r"(ptr) : "r"(ptr), "r" (flags) : IA64_SPINLOCK_CLOBBERS);
	63	# endif /* CONFIG_MCKINLEY */
	64	#else
	65	# ifdef CONFIG_ITANIUM
	66	/* don't use brl on Itanium... */
	67	/* mis-declare, so we get the entry-point, not it's function descriptor: */
	68	asm volatile ("mov r30 = 1\n\t"
	69	"mov r27=%2\n\t"
	70	"mov ar.ccv = r0;;\n\t"
	71	"cmpxchg4.acq r30 = [%0], r30, ar.ccv\n\t"
	72	"movl r29 = ia64_spinlock_contention;;\n\t"
	73	"cmp4.ne p14, p0 = r30, r0\n\t"
	74	"mov b6 = r29;;\n\t"
	75	"(p14) br.call.spnt.many b6 = b6"
	76	: "=r"(ptr) : "r"(ptr), "r" (flags) : IA64_SPINLOCK_CLOBBERS);
	77	# else
	78	asm volatile ("mov r30 = 1\n\t"
	79	"mov r27=%2\n\t"
	80	"mov ar.ccv = r0;;\n\t"
	81	"cmpxchg4.acq r30 = [%0], r30, ar.ccv;;\n\t"
	82	"cmp4.ne p14, p0 = r30, r0\n\t"
	83	"(p14) brl.call.spnt.many b6=ia64_spinlock_contention;;"
	84	: "=r"(ptr) : "r"(ptr), "r" (flags) : IA64_SPINLOCK_CLOBBERS);
	85	# endif /* CONFIG_MCKINLEY */
	86	#endif
	87	}
f5210891	88
fb1c8f93	89	#define __raw_spin_lock(lock) __raw_spin_lock_flags(lock, 0)
f5210891 CL	90
f5210891 CL	91	/* Unlock by doing an ordered store and releasing the cacheline with nta */
fb1c8f93	92	static inline void __raw_spin_unlock(raw_spinlock_t *x) {
f5210891 CL	93	barrier();
	94	asm volatile ("st4.rel.nta [%0] = r0\n\t" :: "r"(x));
	95	}
	96
1da177e4	97	#else /* !ASM_SUPPORTED */
fb1c8f93 IM	98	#define __raw_spin_lock_flags(lock, flags) __raw_spin_lock(lock)
fb1c8f93 IM	99	# define __raw_spin_lock(x) \
1da177e4 LT	100	do { \
	101	__u32 ia64_spinlock_ptr = (__u32 ) (x); \
	102	__u64 ia64_spinlock_val; \
	103	ia64_spinlock_val = ia64_cmpxchg4_acq(ia64_spinlock_ptr, 1, 0); \
	104	if (unlikely(ia64_spinlock_val)) { \
	105	do { \
	106	while (*ia64_spinlock_ptr) \
	107	ia64_barrier(); \
	108	ia64_spinlock_val = ia64_cmpxchg4_acq(ia64_spinlock_ptr, 1, 0); \
	109	} while (ia64_spinlock_val); \
	110	} \
	111	} while (0)
fb1c8f93	112	#define __raw_spin_unlock(x) do { barrier(); ((raw_spinlock_t *) x)->lock = 0; } while (0)
1da177e4 LT	113	#endif /* !ASM_SUPPORTED */
1da177e4 LT	114
fb1c8f93 IM	115	#define __raw_spin_is_locked(x) ((x)->lock != 0)
	116	#define __raw_spin_trylock(x) (cmpxchg_acq(&(x)->lock, 0, 1) == 0)
	117	#define __raw_spin_unlock_wait(lock) \
	118	do { while (__raw_spin_is_locked(lock)) cpu_relax(); } while (0)
1da177e4	119
fb1c8f93 IM	120	#define __raw_read_can_lock(rw) ((volatile int )(rw) >= 0)
fb1c8f93 IM	121	#define __raw_write_can_lock(rw) ((volatile int )(rw) == 0)
1da177e4	122
fb1c8f93	123	#define __raw_read_lock(rw) \
1da177e4	124	do { \
fb1c8f93	125	raw_rwlock_t *__read_lock_ptr = (rw); \
1da177e4 LT	126	\
	127	while (unlikely(ia64_fetchadd(1, (int *) __read_lock_ptr, acq) < 0)) { \
	128	ia64_fetchadd(-1, (int *) __read_lock_ptr, rel); \
	129	while ((volatile int )__read_lock_ptr < 0) \
	130	cpu_relax(); \
	131	} \
	132	} while (0)
	133
fb1c8f93	134	#define __raw_read_unlock(rw) \
1da177e4	135	do { \
fb1c8f93	136	raw_rwlock_t *__read_lock_ptr = (rw); \
1da177e4 LT	137	ia64_fetchadd(-1, (int *) __read_lock_ptr, rel); \
	138	} while (0)
	139
	140	#ifdef ASM_SUPPORTED
fb1c8f93	141	#define __raw_write_lock(rw) \
1da177e4 LT	142	do { \
	143	__asm__ __volatile__ ( \
	144	"mov ar.ccv = r0\n" \
	145	"dep r29 = -1, r0, 31, 1;;\n" \
	146	"1:\n" \
	147	"ld4 r2 = [%0];;\n" \
	148	"cmp4.eq p0,p7 = r0,r2\n" \
	149	"(p7) br.cond.spnt.few 1b \n" \
	150	"cmpxchg4.acq r2 = [%0], r29, ar.ccv;;\n" \
	151	"cmp4.eq p0,p7 = r0, r2\n" \
	152	"(p7) br.cond.spnt.few 1b;;\n" \
	153	:: "r"(rw) : "ar.ccv", "p7", "r2", "r29", "memory"); \
	154	} while(0)
	155
fb1c8f93	156	#define __raw_write_trylock(rw) \
1da177e4 LT	157	({ \
	158	register long result; \
	159	\
	160	__asm__ __volatile__ ( \
	161	"mov ar.ccv = r0\n" \
	162	"dep r29 = -1, r0, 31, 1;;\n" \
	163	"cmpxchg4.acq %0 = [%1], r29, ar.ccv\n" \
	164	: "=r"(result) : "r"(rw) : "ar.ccv", "r29", "memory"); \
	165	(result == 0); \
	166	})
	167
fb1c8f93	168	static inline void __raw_write_unlock(raw_rwlock_t *x)
f5210891 CL	169	{
	170	u8 y = (u8 )x;
	171	barrier();
	172	asm volatile ("st1.rel.nta [%0] = r0\n\t" :: "r"(y+3) : "memory" );
	173	}
	174
1da177e4 LT	175	#else /* !ASM_SUPPORTED */
1da177e4 LT	176
fb1c8f93	177	#define __raw_write_lock(l) \
1da177e4 LT	178	({ \
	179	__u64 ia64_val, ia64_set_val = ia64_dep_mi(-1, 0, 31, 1); \
	180	__u32 ia64_write_lock_ptr = (__u32 ) (l); \
	181	do { \
	182	while (*ia64_write_lock_ptr) \
	183	ia64_barrier(); \
	184	ia64_val = ia64_cmpxchg4_acq(ia64_write_lock_ptr, ia64_set_val, 0); \
	185	} while (ia64_val); \
	186	})
	187
fb1c8f93	188	#define __raw_write_trylock(rw) \
1da177e4 LT	189	({ \
	190	__u64 ia64_val; \
	191	__u64 ia64_set_val = ia64_dep_mi(-1, 0, 31,1); \
	192	ia64_val = ia64_cmpxchg4_acq((__u32 *)(rw), ia64_set_val, 0); \
	193	(ia64_val == 0); \
	194	})
	195
fb1c8f93	196	static inline void __raw_write_unlock(raw_rwlock_t *x)
f5210891 CL	197	{
	198	barrier();
	199	x->write_lock = 0;
	200	}
	201
1da177e4 LT	202	#endif /* !ASM_SUPPORTED */
1da177e4 LT	203
bf7ececa KO	204	static inline int __raw_read_trylock(raw_rwlock_t *x)
	205	{
	206	union {
	207	raw_rwlock_t lock;
	208	__u32 word;
	209	} old, new;
	210	old.lock = new.lock = *x;
	211	old.lock.write_lock = new.lock.write_lock = 0;
	212	++new.lock.read_counter;
	213	return (u32)ia64_cmpxchg4_acq((__u32 *)(x), new.word, old.word) == old.word;
	214	}
1da177e4	215
ef6edc97 MS	216	#define _raw_spin_relax(lock) cpu_relax()
	217	#define _raw_read_relax(lock) cpu_relax()
	218	#define _raw_write_relax(lock) cpu_relax()
	219
1da177e4	220	#endif /* _ASM_IA64_SPINLOCK_H */