liburcu-bp: Use membarrier private expedited when available

For the liburcu-bp flavor, use the membarrier private expedited command
when available. It is faster than the shared expedited command, but has
only been introduced in 4.14 Linux kernels.

When configured with --disable-sys-membarrier-fallback, liburcu-bp
will abort if running on a kernel that do not provide the membarrier
private expedited command (e.g. CONFIG_MEMBARRIER=n or kernel version
below 4.14).

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>

diff --git a/src/urcu-bp.c b/src/urcu-bp.c
index ebe96c57da9630d3e43168ee32fa96da95277bfa..5d07f040f38c2ef0b33fce285c49de8e14188528 100644 (file)
--- a/src/urcu-bp.c
+++ b/src/urcu-bp.c
@@ -33,6 +33,7 @@
 #include <errno.h>
 #include <poll.h>
 #include <unistd.h>
+#include <stdbool.h>
 #include <sys/mman.h>
 
 #include "urcu/arch.h"
@@ -102,8 +103,12 @@ int rcu_bp_refcount;
 #endif
 
 enum membarrier_cmd {
-       MEMBARRIER_CMD_QUERY = 0,
-       MEMBARRIER_CMD_SHARED = (1 << 0),
+       MEMBARRIER_CMD_QUERY                            = 0,
+       MEMBARRIER_CMD_SHARED                           = (1 << 0),
+       /* reserved for MEMBARRIER_CMD_SHARED_EXPEDITED (1 << 1) */
+       /* reserved for MEMBARRIER_CMD_PRIVATE (1 << 2) */
+       MEMBARRIER_CMD_PRIVATE_EXPEDITED                = (1 << 3),
+       MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED       = (1 << 4),
 };
 
 static
@@ -192,10 +197,12 @@ static void mutex_unlock(pthread_mutex_t *mutex)
 
 static void smp_mb_master(void)
 {
-       if (caa_likely(urcu_bp_has_sys_membarrier))
-               (void) membarrier(MEMBARRIER_CMD_SHARED, 0);
-       else
+       if (caa_likely(urcu_bp_has_sys_membarrier)) {
+               if (membarrier(MEMBARRIER_CMD_PRIVATE_EXPEDITED, 0))
+                       urcu_die(errno);
+       } else {
                cmm_smp_mb();
+       }
 }
 
 /*
@@ -584,25 +591,38 @@ void urcu_bp_thread_exit_notifier(void *rcu_key)
 
 #ifdef CONFIG_RCU_FORCE_SYS_MEMBARRIER
 static
-void rcu_sys_membarrier_status(int available)
+void rcu_sys_membarrier_status(bool available)
 {
        if (!available)
                abort();
 }
 #else
 static
-void rcu_sys_membarrier_status(int available)
+void rcu_sys_membarrier_status(bool available)
 {
-       /*
-        * membarrier has blocking behavior, which changes the
-        * application behavior too much compared to using barriers when
-        * synchronize_rcu is used repeatedly (without using call_rcu).
-        * Don't use membarrier for now, unless its use has been
-        * explicitly forced when building liburcu.
-        */
+       if (!available)
+               return;
+       urcu_bp_has_sys_membarrier = 1;
 }
 #endif
 
+static
+void rcu_sys_membarrier_init(void)
+{
+       bool available = false;
+       int mask;
+
+       mask = membarrier(MEMBARRIER_CMD_QUERY, 0);
+       if (mask >= 0) {
+               if (mask & MEMBARRIER_CMD_PRIVATE_EXPEDITED) {
+                       if (membarrier(MEMBARRIER_CMD_REGISTER_PRIVATE_EXPEDITED, 0))
+                               urcu_die(errno);
+                       available = true;
+               }
+       }
+       rcu_sys_membarrier_status(available);
+}
+
 static
 void rcu_bp_init(void)
 {
@@ -614,9 +634,7 @@ void rcu_bp_init(void)
                                urcu_bp_thread_exit_notifier);
                if (ret)
                        abort();
-               ret = membarrier(MEMBARRIER_CMD_QUERY, 0);
-               rcu_sys_membarrier_status(ret >= 0
-                               && (ret & MEMBARRIER_CMD_SHARED));
+               rcu_sys_membarrier_init();
                initialized = 1;
        }
        mutex_unlock(&init_lock);