[patch] hwpmc [1/13]
Aggelos Economopoulos
aoiko at cc.ece.ntua.gr
Sat Nov 24 13:17:45 PST 2007
Bring in some atomic primitives. Required for hwpmc.
Obtained-from: FreeBSD
Index: cpu/i386/include/atomic.h
===================================================================
retrieving revision 1.23
diff -u -p -r1.23 atomic.h
--- cpu/i386/include/atomic.h
+++ cpu/i386/include/atomic.h
@@ -70,6 +70,9 @@
#define ATOMIC_ASM(NAME, TYPE, OP, V) \
extern void atomic_##NAME##_##TYPE(volatile u_##TYPE *p, u_##TYPE v); \
extern void atomic_##NAME##_##TYPE##_nonlocked(volatile u_##TYPE *p, u_##TYPE v);
+#define ATOMIC_STORE_LOAD(TYPE, LOP, SOP) \
+u_##TYPE atomic_load_acq_##TYPE(volatile u_##TYPE *p); \
+void atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v)
#else /* !KLD_MODULE */
#if defined(SMP) || !defined(_KERNEL)
#define MPLOCKED "lock ; "
@@ -107,6 +110,61 @@ atomic_##NAME##_##TYPE##_nonlocked(volat
: "iq" (V)); \
}
+#if defined(_KERNEL) && !defined(SMP)
+
+/*
+ * We assume that a = b will do atomic loads and stores. However, on a
+ * PentiumPro or higher, reads may pass writes, so for that case we have
+ * to use a serializing instruction (i.e. with LOCK) to do the load in
+ * SMP kernels. For UP kernels, however, the cache of the single processor
+ * is always consistent, so we don't need any memory barriers.
+ */
+#define ATOMIC_STORE_LOAD(TYPE, LOP, SOP) \
+static __inline u_##TYPE \
+atomic_load_acq_##TYPE(volatile u_##TYPE *p) \
+{ \
+ return (*p); \
+} \
+ \
+static __inline void \
+atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\
+{ \
+ *p = v; \
+} \
+struct __hack
+
+#else /* !(_KERNEL && !SMP) */
+
+#define ATOMIC_STORE_LOAD(TYPE, LOP, SOP) \
+static __inline u_##TYPE \
+atomic_load_acq_##TYPE(volatile u_##TYPE *p) \
+{ \
+ u_##TYPE res; \
+ \
+ __asm __volatile(MPLOCKED LOP \
+ : "=a" (res), /* 0 */ \
+ "=m" (*p) /* 1 */ \
+ : "m" (*p) /* 2 */ \
+ : "memory"); \
+ \
+ return (res); \
+} \
+ \
+/* \
+ * The XCHG instruction asserts LOCK automagically. \
+ */ \
+static __inline void \
+atomic_store_rel_##TYPE(volatile u_##TYPE *p, u_##TYPE v)\
+{ \
+ __asm __volatile(SOP \
+ : "=m" (*p), /* 0 */ \
+ "+r" (v) /* 1 */ \
+ : "m" (*p)); /* 2 */ \
+} \
+struct __hack
+
+#endif /* _KERNEL && !SMP */
+
#endif /* KLD_MODULE */
/* egcs 1.1.2+ version */
@@ -130,6 +188,14 @@ ATOMIC_ASM(clear, long, "andl %1,%0"
ATOMIC_ASM(add, long, "addl %1,%0", v)
ATOMIC_ASM(subtract, long, "subl %1,%0", v)
+ATOMIC_STORE_LOAD(char, "cmpxchgb %b0,%1", "xchgb %b1,%0");
+ATOMIC_STORE_LOAD(short,"cmpxchgw %w0,%1", "xchgw %w1,%0");
+ATOMIC_STORE_LOAD(int, "cmpxchgl %0,%1", "xchgl %1,%0");
+ATOMIC_STORE_LOAD(long, "cmpxchgl %0,%1", "xchgl %1,%0");
+
+#undef ATOMIC_ASM
+#undef ATOMIC_STORE_LOAD
+
/*
* atomic_poll_acquire_int(P) Returns non-zero on success, 0 if the lock
* has already been acquired.
@@ -367,4 +433,138 @@ atomic_cmpset_int(volatile u_int *_dst,
}
#endif /* KLD_MODULE */
+/* Acquire and release variants are identical to the normal ones. */
+#define atomic_set_acq_char atomic_set_char
+#define atomic_set_rel_char atomic_set_char
+#define atomic_clear_acq_char atomic_clear_char
+#define atomic_clear_rel_char atomic_clear_char
+#define atomic_add_acq_char atomic_add_char
+#define atomic_add_rel_char atomic_add_char
+#define atomic_subtract_acq_char atomic_subtract_char
+#define atomic_subtract_rel_char atomic_subtract_char
+
+#define atomic_set_acq_short atomic_set_short
+#define atomic_set_rel_short atomic_set_short
+#define atomic_clear_acq_short atomic_clear_short
+#define atomic_clear_rel_short atomic_clear_short
+#define atomic_add_acq_short atomic_add_short
+#define atomic_add_rel_short atomic_add_short
+#define atomic_subtract_acq_short atomic_subtract_short
+#define atomic_subtract_rel_short atomic_subtract_short
+
+#define atomic_set_acq_int atomic_set_int
+#define atomic_set_rel_int atomic_set_int
+#define atomic_clear_acq_int atomic_clear_int
+#define atomic_clear_rel_int atomic_clear_int
+#define atomic_add_acq_int atomic_add_int
+#define atomic_add_rel_int atomic_add_int
+#define atomic_subtract_acq_int atomic_subtract_int
+#define atomic_subtract_rel_int atomic_subtract_int
+#define atomic_cmpset_acq_int atomic_cmpset_int
+#define atomic_cmpset_rel_int atomic_cmpset_int
+
+#define atomic_set_acq_long atomic_set_long
+#define atomic_set_rel_long atomic_set_long
+#define atomic_clear_acq_long atomic_clear_long
+#define atomic_clear_rel_long atomic_clear_long
+#define atomic_add_acq_long atomic_add_long
+#define atomic_add_rel_long atomic_add_long
+#define atomic_subtract_acq_long atomic_subtract_long
+#define atomic_subtract_rel_long atomic_subtract_long
+#define atomic_cmpset_acq_long atomic_cmpset_long
+#define atomic_cmpset_rel_long atomic_cmpset_long
+
+/* Operations on 8-bit bytes. */
+#define atomic_set_8 atomic_set_char
+#define atomic_set_acq_8 atomic_set_acq_char
+#define atomic_set_rel_8 atomic_set_rel_char
+#define atomic_clear_8 atomic_clear_char
+#define atomic_clear_acq_8 atomic_clear_acq_char
+#define atomic_clear_rel_8 atomic_clear_rel_char
+#define atomic_add_8 atomic_add_char
+#define atomic_add_acq_8 atomic_add_acq_char
+#define atomic_add_rel_8 atomic_add_rel_char
+#define atomic_subtract_8 atomic_subtract_char
+#define atomic_subtract_acq_8 atomic_subtract_acq_char
+#define atomic_subtract_rel_8 atomic_subtract_rel_char
+#define atomic_load_acq_8 atomic_load_acq_char
+#define atomic_store_rel_8 atomic_store_rel_char
+
+/* Operations on 16-bit words. */
+#define atomic_set_16 atomic_set_short
+#define atomic_set_acq_16 atomic_set_acq_short
+#define atomic_set_rel_16 atomic_set_rel_short
+#define atomic_clear_16 atomic_clear_short
+#define atomic_clear_acq_16 atomic_clear_acq_short
+#define atomic_clear_rel_16 atomic_clear_rel_short
+#define atomic_add_16 atomic_add_short
+#define atomic_add_acq_16 atomic_add_acq_short
+#define atomic_add_rel_16 atomic_add_rel_short
+#define atomic_subtract_16 atomic_subtract_short
+#define atomic_subtract_acq_16 atomic_subtract_acq_short
+#define atomic_subtract_rel_16 atomic_subtract_rel_short
+#define atomic_load_acq_16 atomic_load_acq_short
+#define atomic_store_rel_16 atomic_store_rel_short
+
+/* Operations on 32-bit double words. */
+#define atomic_set_32 atomic_set_int
+#define atomic_set_acq_32 atomic_set_acq_int
+#define atomic_set_rel_32 atomic_set_rel_int
+#define atomic_clear_32 atomic_clear_int
+#define atomic_clear_acq_32 atomic_clear_acq_int
+#define atomic_clear_rel_32 atomic_clear_rel_int
+#define atomic_add_32 atomic_add_int
+#define atomic_add_acq_32 atomic_add_acq_int
+#define atomic_add_rel_32 atomic_add_rel_int
+#define atomic_subtract_32 atomic_subtract_int
+#define atomic_subtract_acq_32 atomic_subtract_acq_int
+#define atomic_subtract_rel_32 atomic_subtract_rel_int
+#define atomic_load_acq_32 atomic_load_acq_int
+#define atomic_store_rel_32 atomic_store_rel_int
+#define atomic_cmpset_32 atomic_cmpset_int
+#define atomic_cmpset_acq_32 atomic_cmpset_acq_int
+#define atomic_cmpset_rel_32 atomic_cmpset_rel_int
+#define atomic_readandclear_32 atomic_readandclear_int
+#define atomic_fetchadd_32 atomic_fetchadd_int
+
+/* Operations on pointers. */
+#define atomic_set_ptr(p, v) \
+ atomic_set_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_set_acq_ptr(p, v) \
+ atomic_set_acq_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_set_rel_ptr(p, v) \
+ atomic_set_rel_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_clear_ptr(p, v) \
+ atomic_clear_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_clear_acq_ptr(p, v) \
+ atomic_clear_acq_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_clear_rel_ptr(p, v) \
+ atomic_clear_rel_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_add_ptr(p, v) \
+ atomic_add_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_add_acq_ptr(p, v) \
+ atomic_add_acq_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_add_rel_ptr(p, v) \
+ atomic_add_rel_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_subtract_ptr(p, v) \
+ atomic_subtract_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_subtract_acq_ptr(p, v) \
+ atomic_subtract_acq_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_subtract_rel_ptr(p, v) \
+ atomic_subtract_rel_int((volatile u_int *)(p), (u_int)(v))
+#define atomic_load_acq_ptr(p) \
+ atomic_load_acq_int((volatile u_int *)(p))
+#define atomic_store_rel_ptr(p, v) \
+ atomic_store_rel_int((volatile u_int *)(p), (v))
+#define atomic_cmpset_ptr(dst, old, new) \
+ atomic_cmpset_int((volatile u_int *)(dst), (u_int)(old), (u_int)(new))
+#define atomic_cmpset_acq_ptr(dst, old, new) \
+ atomic_cmpset_acq_int((volatile u_int *)(dst), (u_int)(old), \
+ (u_int)(new))
+#define atomic_cmpset_rel_ptr(dst, old, new) \
+ atomic_cmpset_rel_int((volatile u_int *)(dst), (u_int)(old), \
+ (u_int)(new))
+#define atomic_readandclear_ptr(p) \
+ atomic_readandclear_int((volatile u_int *)(p))
+
#endif /* ! _CPU_ATOMIC_H_ */
More information about the Submit
mailing list