Use C11-like atomics instead of plain memory accesses in x86 lock elision.

This uses atomic operations to access lock elision metadata that is accessed
concurrently (ie, adapt_count fields).  The size of the data is less than a
word but accessed only with atomic loads and stores; therefore, we add
support for shorter-size atomic load and stores too.

	* include/atomic.h (__atomic_check_size_ls): New.
	(atomic_load_relaxed, atomic_load_acquire, atomic_store_relaxed,
	atomic_store_release): Use it.
	* sysdeps/x86/elide.h (ACCESS_ONCE): Remove.
	(elision_adapt, ELIDE_LOCK): Use atomics.
	* sysdeps/unix/sysv/linux/x86/elision-lock.c (__lll_lock_elision): Use
	atomics and improve code comments.
	* sysdeps/unix/sysv/linux/x86/elision-trylock.c
	(__lll_trylock_elision): Likewise.

1 files changed, 20 insertions, 4 deletions

diff --git a/include/atomic.h b/include/atomic.h
index c8b4664..d14cbc5 100644
--- a/include/atomic.h
+++ b/include/atomic.h
@@ -550,6 +550,20 @@ void __atomic_link_error (void);
    if (sizeof (*mem) != 4)						      \
      __atomic_link_error ();
 # endif
+/* We additionally provide 8b and 16b atomic loads and stores; we do not yet
+   need other atomic operations of such sizes, and restricting the support to
+   loads and stores makes this easier for archs that do not have native
+   support for atomic operations to less-than-word-sized data.  */
+# if __HAVE_64B_ATOMICS == 1
+#  define __atomic_check_size_ls(mem) \
+   if ((sizeof (*mem) != 1) && (sizeof (*mem) != 2) && (sizeof (*mem) != 4)   \
+       && (sizeof (*mem) != 8))						      \
+     __atomic_link_error ();
+# else
+#  define __atomic_check_size_ls(mem) \
+   if ((sizeof (*mem) != 1) && (sizeof (*mem) != 2) && sizeof (*mem) != 4)    \
+     __atomic_link_error ();
+# endif
 
 # define atomic_thread_fence_acquire() \
   __atomic_thread_fence (__ATOMIC_ACQUIRE)
@@ -559,18 +573,20 @@ void __atomic_link_error (void);
   __atomic_thread_fence (__ATOMIC_SEQ_CST)
 
 # define atomic_load_relaxed(mem) \
-  ({ __atomic_check_size((mem)); __atomic_load_n ((mem), __ATOMIC_RELAXED); })
+  ({ __atomic_check_size_ls((mem));					      \
+     __atomic_load_n ((mem), __ATOMIC_RELAXED); })
 # define atomic_load_acquire(mem) \
-  ({ __atomic_check_size((mem)); __atomic_load_n ((mem), __ATOMIC_ACQUIRE); })
+  ({ __atomic_check_size_ls((mem));					      \
+     __atomic_load_n ((mem), __ATOMIC_ACQUIRE); })
 
 # define atomic_store_relaxed(mem, val) \
   do {									      \
-    __atomic_check_size((mem));						      \
+    __atomic_check_size_ls((mem));					      \
     __atomic_store_n ((mem), (val), __ATOMIC_RELAXED);			      \
   } while (0)
 # define atomic_store_release(mem, val) \
   do {									      \
-    __atomic_check_size((mem));						      \
+    __atomic_check_size_ls((mem));					      \
     __atomic_store_n ((mem), (val), __ATOMIC_RELEASE);			      \
   } while (0)