sys/arch/x86/include/lock.h

   1 /*      $NetBSD: lock.h,v 1.24 2008/04/28 20:23:40 martin Exp $ */
   2
   3 /*-
   4  * Copyright (c) 2000, 2006 The NetBSD Foundation, Inc.
   5  * All rights reserved.
   6  *
   7  * This code is derived from software contributed to The NetBSD Foundation
   8  * by Jason R. Thorpe and Andrew Doran.
   9  *
  10  * Redistribution and use in source and binary forms, with or without
  11  * modification, are permitted provided that the following conditions
  12  * are met:
  13  * 1. Redistributions of source code must retain the above copyright
  14  *    notice, this list of conditions and the following disclaimer.
  15  * 2. Redistributions in binary form must reproduce the above copyright
  16  *    notice, this list of conditions and the following disclaimer in the
  17  *    documentation and/or other materials provided with the distribution.
  18  *
  19  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
  20  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
  21  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  22  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
  23  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  24  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  25  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  26  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  27  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  28  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
  29  * POSSIBILITY OF SUCH DAMAGE.
  30  */
  31
  32 /*
  33  * Machine-dependent spin lock operations.
  34  */
  35
  36 #ifndef _X86_LOCK_H_
  37 #define _X86_LOCK_H_
  38
  39 #include <sys/param.h>
  40
  41 static __inline int
  42 __SIMPLELOCK_LOCKED_P(__cpu_simple_lock_t *__ptr)
  43 {
  44         return *__ptr == __SIMPLELOCK_LOCKED;
  45 }
  46
  47 static __inline int
  48 __SIMPLELOCK_UNLOCKED_P(__cpu_simple_lock_t *__ptr)
  49 {
  50         return *__ptr == __SIMPLELOCK_UNLOCKED;
  51 }
  52
  53 static __inline void
  54 __cpu_simple_lock_set(__cpu_simple_lock_t *__ptr)
  55 {
  56
  57         *__ptr = __SIMPLELOCK_LOCKED;
  58 }
  59
  60 static __inline void
  61 __cpu_simple_lock_clear(__cpu_simple_lock_t *__ptr)
  62 {
  63
  64         *__ptr = __SIMPLELOCK_UNLOCKED;
  65 }
  66
  67 #ifdef _HARDKERNEL
  68
  69 #include <machine/cpufunc.h>
  70
  71 void    __cpu_simple_lock_init(__cpu_simple_lock_t *);
  72 void    __cpu_simple_lock(__cpu_simple_lock_t *);
  73 int     __cpu_simple_lock_try(__cpu_simple_lock_t *);
  74 void    __cpu_simple_unlock(__cpu_simple_lock_t *);
  75
  76 #define SPINLOCK_SPIN_HOOK      /* nothing */
  77
  78 #ifdef SPINLOCK_BACKOFF_HOOK
  79 #undef SPINLOCK_BACKOFF_HOOK
  80 #endif
  81 #define SPINLOCK_BACKOFF_HOOK   x86_pause()
  82
  83 #else
  84
  85 static __inline void __cpu_simple_lock_init(__cpu_simple_lock_t *)
  86         __unused;
  87 static __inline void __cpu_simple_lock(__cpu_simple_lock_t *)
  88         __unused;
  89 static __inline int __cpu_simple_lock_try(__cpu_simple_lock_t *)
  90         __unused;
  91 static __inline void __cpu_simple_unlock(__cpu_simple_lock_t *)
  92         __unused;
  93
  94 static __inline void
  95 __cpu_simple_lock_init(__cpu_simple_lock_t *lockp)
  96 {
  97
  98         *lockp = __SIMPLELOCK_UNLOCKED;
  99         __insn_barrier();
 100 }
 101
 102 static __inline int
 103 __cpu_simple_lock_try(__cpu_simple_lock_t *lockp)
 104 {
 105         uint8_t val;
 106
 107         val = __SIMPLELOCK_LOCKED;
 108         __asm volatile ("xchgb %0,(%2)" :
 109             "=r" (val)
 110             :"0" (val), "r" (lockp));
 111         __insn_barrier();
 112         return val == __SIMPLELOCK_UNLOCKED;
 113 }
 114
 115 static __inline void
 116 __cpu_simple_lock(__cpu_simple_lock_t *lockp)
 117 {
 118
 119         while (!__cpu_simple_lock_try(lockp))
 120                 /* nothing */;
 121         __insn_barrier();
 122 }
 123
 124 /*
 125  * Note on x86 memory ordering
 126  *
 127  * When releasing a lock we must ensure that no stores or loads from within
 128  * the critical section are re-ordered by the CPU to occur outside of it:
 129  * they must have completed and be visible to other processors once the lock
 130  * has been released.
 131  *
 132  * NetBSD usually runs with the kernel mapped (via MTRR) in a WB (write
 133  * back) memory region.  In that case, memory ordering on x86 platforms
 134  * looks like this:
 135  *
 136  * i386         All loads/stores occur in instruction sequence.
 137  *
 138  * i486         All loads/stores occur in instruction sequence.  In
 139  * Pentium      exceptional circumstances, loads can be re-ordered around
 140  *              stores, but for the purposes of releasing a lock it does
 141  *              not matter.  Stores may not be immediately visible to other
 142  *              processors as they can be buffered.  However, since the
 143  *              stores are buffered in order the lock release will always be
 144  *              the last operation in the critical section that becomes
 145  *              visible to other CPUs.
 146  *
 147  * Pentium Pro  The "Intel 64 and IA-32 Architectures Software Developer's
 148  * onwards      Manual" volume 3A (order number 248966) says that (1) "Reads
 149  *              can be carried out speculatively and in any order" and (2)
 150  *              "Reads can pass buffered stores, but the processor is
 151  *              self-consistent.".  This would be a problem for the below,
 152  *              and would mandate a locked instruction cycle or load fence
 153  *              before releasing the simple lock.
 154  *
 155  *              The "Intel Pentium 4 Processor Optimization" guide (order
 156  *              number 253668-022US) says: "Loads can be moved before stores
 157  *              that occurred earlier in the program if they are not
 158  *              predicted to load from the same linear address.".  This is
 159  *              not a problem since the only loads that can be re-ordered
 160  *              take place once the lock has been released via a store.
 161  *
 162  *              The above two documents seem to contradict each other,
 163  *              however with the exception of early steppings of the Pentium
 164  *              Pro, the second document is closer to the truth: a store
 165  *              will always act as a load fence for all loads that precede
 166  *              the store in instruction order.
 167  *
 168  *              Again, note that stores can be buffered and will not always
 169  *              become immediately visible to other CPUs: they are however
 170  *              buffered in order.
 171  *
 172  * AMD64        Stores occur in order and are buffered.  Loads can be
 173  *              reordered, however stores act as load fences, meaning that
 174  *              loads can not be reordered around stores.
 175  */
 176 static __inline void
 177 __cpu_simple_unlock(__cpu_simple_lock_t *lockp)
 178 {
 179
 180         __insn_barrier();
 181         *lockp = __SIMPLELOCK_UNLOCKED;
 182 }
 183
 184 #endif  /* _HARDKERNEL */
 185
 186 #endif /* _X86_LOCK_H_ */