user_atomic.h

/*-
 * Copyright (c) 2009-2010 Brad Penoff
 * Copyright (c) 2009-2010 Humaira Kamal
 * Copyright (c) 2011-2012 Irene Ruengeler
 * Copyright (c) 2011-2012 Michael Tuexen
 *
 * All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in the
 *    documentation and/or other materials provided with the distribution.
 *
 * THIS SOFTWARE IS PROVIDED BY THE AUTHOR AND CONTRIBUTORS ``AS IS'' AND
 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
 * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
 * SUCH DAMAGE.
 */

#ifndef _USER_ATOMIC_H_
#define _USER_ATOMIC_H_

/* __Userspace__ version of sys/i386/include/atomic.h goes here */

/* TODO In the future, might want to not use i386 specific assembly.
 *    The options include:
 *       - implement them generically (but maybe not truly atomic?) in userspace
 *       - have ifdef's for __Userspace_arch_ perhaps (OS isn't enough...)
 */

#include <stdio.h>
#include <sys/types.h>

#if defined(__APPLE__) || defined(_WIN32)
#if defined(_WIN32)
#define atomic_add_int(addr, val) InterlockedExchangeAdd((LPLONG)addr, (LONG)val)
#define atomic_fetchadd_int(addr, val) InterlockedExchangeAdd((LPLONG)addr, (LONG)val)
#define atomic_subtract_int(addr, val)   InterlockedExchangeAdd((LPLONG)addr,-((LONG)val))
#define atomic_cmpset_int(dst, exp, src) InterlockedCompareExchange((LPLONG)dst, src, exp)
#define SCTP_DECREMENT_AND_CHECK_REFCOUNT(addr) (InterlockedExchangeAdd((LPLONG)addr, (-1L)) == 1)
#else
#include <libkern/OSAtomic.h>
#define atomic_add_int(addr, val) OSAtomicAdd32Barrier(val, (int32_t *)addr)
#define atomic_fetchadd_int(addr, val) OSAtomicAdd32Barrier(val, (int32_t *)addr)
#define atomic_subtract_int(addr, val) OSAtomicAdd32Barrier(-val, (int32_t *)addr)
#define atomic_cmpset_int(dst, exp, src) OSAtomicCompareAndSwapIntBarrier(exp, src, (int *)dst)
#define SCTP_DECREMENT_AND_CHECK_REFCOUNT(addr) (atomic_fetchadd_int(addr, -1) == 0)
#endif

#if defined(INVARIANTS)
#define SCTP_SAVE_ATOMIC_DECREMENT(addr, val) \
{ \
	int32_t newval; \
	newval = atomic_fetchadd_int(addr, -val); \
	if (newval < 0) { \
		panic("Counter goes negative"); \
	} \
}
#else
#define SCTP_SAVE_ATOMIC_DECREMENT(addr, val) \
{ \
	int32_t newval; \
	newval = atomic_fetchadd_int(addr, -val); \
	if (newval < 0) { \
		*addr = 0; \
	} \
}
#endif
#if defined(_WIN32)
static void atomic_init(void) {} /* empty when we are not using atomic_mtx */
#else
static inline void atomic_init(void) {} /* empty when we are not using atomic_mtx */
#endif

#else
/* Using gcc built-in functions for atomic memory operations
   Reference: http://gcc.gnu.org/onlinedocs/gcc-4.1.0/gcc/Atomic-Builtins.html
   Requires gcc version 4.1.0
   compile with -march=i486
 */

/*Atomically add V to *P.*/
#define atomic_add_int(P, V)	 (void) __sync_fetch_and_add(P, V)

/*Atomically subtrace V from *P.*/
#define atomic_subtract_int(P, V) (void) __sync_fetch_and_sub(P, V)

/*
 * Atomically add the value of v to the integer pointed to by p and return
 * the previous value of *p.
 */
#define atomic_fetchadd_int(p, v) __sync_fetch_and_add(p, v)

/* Following explanation from src/sys/i386/include/atomic.h,
 * for atomic compare and set
 *
 * if (*dst == exp) *dst = src (all 32 bit words)
 *
 * Returns 0 on failure, non-zero on success
 */

#define atomic_cmpset_int(dst, exp, src) __sync_bool_compare_and_swap(dst, exp, src)

#define SCTP_DECREMENT_AND_CHECK_REFCOUNT(addr) (atomic_fetchadd_int(addr, -1) == 1)
#if defined(INVARIANTS)
#define SCTP_SAVE_ATOMIC_DECREMENT(addr, val) \
{ \
	int32_t oldval; \
	oldval = atomic_fetchadd_int(addr, -val); \
	if (oldval < val) { \
		panic("Counter goes negative"); \
	} \
}
#else
#define SCTP_SAVE_ATOMIC_DECREMENT(addr, val) \
{ \
	int32_t oldval; \
	oldval = atomic_fetchadd_int(addr, -val); \
	if (oldval < val) { \
		*addr = 0; \
	} \
}
#endif
static inline void atomic_init(void) {} /* empty when we are not using atomic_mtx */
#endif

#if 0 /* using libatomic_ops */
#include "user_include/atomic_ops.h"

/*Atomically add incr to *P, and return the original value of *P.*/
#define atomic_add_int(P, V)	 AO_fetch_and_add((AO_t*)P, V)

#define atomic_subtract_int(P, V) AO_fetch_and_add((AO_t*)P, -(V))

/*
 * Atomically add the value of v to the integer pointed to by p and return
 * the previous value of *p.
 */
#define atomic_fetchadd_int(p, v) AO_fetch_and_add((AO_t*)p, v)

/* Atomically compare *addr to old_val, and replace *addr by new_val
   if the first comparison succeeds.  Returns nonzero if the comparison
   succeeded and *addr was updated.
*/
/* Following Explanation from src/sys/i386/include/atomic.h, which
   matches that of AO_compare_and_swap above.
 * Atomic compare and set, used by the mutex functions
 *
 * if (*dst == exp) *dst = src (all 32 bit words)
 *
 * Returns 0 on failure, non-zero on success
 */

#define atomic_cmpset_int(dst, exp, src) AO_compare_and_swap((AO_t*)dst, exp, src)

static inline void atomic_init() {} /* empty when we are not using atomic_mtx */
#endif /* closing #if for libatomic */

#if 0 /* using atomic_mtx */

#include <pthread.h>

extern userland_mutex_t atomic_mtx;

#if defined(_WIN32)
static inline void atomic_init() {
	InitializeCriticalSection(&atomic_mtx);
}
static inline void atomic_destroy() {
	DeleteCriticalSection(&atomic_mtx);
}
static inline void atomic_lock() {
	EnterCriticalSection(&atomic_mtx);
}
static inline void atomic_unlock() {
	LeaveCriticalSection(&atomic_mtx);
}
#else
static inline void atomic_init() {
	pthread_mutexattr_t mutex_attr;

	pthread_mutexattr_init(&mutex_attr);
#ifdef INVARIANTS
	pthread_mutexattr_settype(&mutex_attr, PTHREAD_MUTEX_ERRORCHECK);
#endif
	pthread_mutex_init(&accept_mtx, &mutex_attr);
	pthread_mutexattr_destroy(&mutex_attr);
}
static inline void atomic_destroy() {
	(void)pthread_mutex_destroy(&atomic_mtx);
}
static inline void atomic_lock() {
#ifdef INVARIANTS
	KASSERT(pthread_mutex_lock(&atomic_mtx) == 0, ("atomic_lock: atomic_mtx already locked"))
#else
	(void)pthread_mutex_lock(&atomic_mtx);
#endif
}
static inline void atomic_unlock() {
#ifdef INVARIANTS
	KASSERT(pthread_mutex_unlock(&atomic_mtx) == 0, ("atomic_unlock: atomic_mtx not locked"))
#else
	(void)pthread_mutex_unlock(&atomic_mtx);
#endif
}
#endif
/*
 * For userland, always use lock prefixes so that the binaries will run
 * on both SMP and !SMP systems.
 */

#define	MPLOCKED	"lock ; "

/*
 * Atomically add the value of v to the integer pointed to by p and return
 * the previous value of *p.
 */
static __inline u_int
atomic_fetchadd_int(volatile void *n, u_int v)
{
	int *p = (int *) n;
	atomic_lock();
	__asm __volatile(
	"	" MPLOCKED "		"
	"	xaddl	%0, %1 ;	"
	"# atomic_fetchadd_int"
	: "+r" (v),			/* 0 (result) */
	  "=m" (*p)			/* 1 */
	: "m" (*p));			/* 2 */
	atomic_unlock();

	return (v);
}


#ifdef CPU_DISABLE_CMPXCHG

static __inline int
atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src)
{
	u_char res;

	atomic_lock();
	__asm __volatile(
	"	pushfl ;		"
	"	cli ;			"
	"	cmpl	%3,%4 ;		"
	"	jne	1f ;		"
	"	movl	%2,%1 ;		"
	"1:				"
	"       sete	%0 ;		"
	"	popfl ;			"
	"# atomic_cmpset_int"
	: "=q" (res),			/* 0 */
	  "=m" (*dst)			/* 1 */
	: "r" (src),			/* 2 */
	  "r" (exp),			/* 3 */
	  "m" (*dst)			/* 4 */
	: "memory");
	atomic_unlock();

	return (res);
}

#else /* !CPU_DISABLE_CMPXCHG */

static __inline int
atomic_cmpset_int(volatile u_int *dst, u_int exp, u_int src)
{
	atomic_lock();
	u_char res;

	__asm __volatile(
	"	" MPLOCKED "		"
	"	cmpxchgl %2,%1 ;	"
	"	sete	%0 ;		"
	"1:				"
	"# atomic_cmpset_int"
	: "=a" (res),			/* 0 */
	  "=m" (*dst)			/* 1 */
	: "r" (src),			/* 2 */
	  "a" (exp),			/* 3 */
	  "m" (*dst)			/* 4 */
	: "memory");
	atomic_unlock();

	return (res);
}

#endif /* CPU_DISABLE_CMPXCHG */

#define atomic_add_int(P, V)	 do {   \
		atomic_lock();          \
		(*(u_int *)(P) += (V)); \
		atomic_unlock();        \
} while(0)
#define atomic_subtract_int(P, V)  do {   \
		atomic_lock();            \
		(*(u_int *)(P) -= (V));   \
		atomic_unlock();          \
} while(0)

#endif
#endif