urcu-defer: make call_rcu() energy efficient using futex()

[urcu.git] / urcu-defer.c

diff --git a/urcu-defer.c b/urcu-defer.c
index cd939c7ac85b4bd03bb3d24f769cbb8ef8f70604..5b904f8321a23c2ef3b84752c64c84cc9f9c865a 100644 (file)
--- a/urcu-defer.c
+++ b/urcu-defer.c
@@ -28,11 +28,17 @@
 #include <string.h>
 #include <errno.h>
 #include <poll.h>
+#include <linux/futex.h>
+#include <sys/time.h>
+#include <syscall.h>
+#include <unistd.h>
 
 #include "urcu-defer-static.h"
 /* Do not #define _LGPL_SOURCE to ensure we can emit the wrapper symbols */
 #include "urcu-defer.h"
 
+#define futex(...)     syscall(__NR_futex, __VA_ARGS__)
+
 void __attribute__((destructor)) urcu_defer_exit(void);
 
 extern void synchronize_rcu(void);
@@ -43,6 +49,8 @@ extern void synchronize_rcu(void);
 static pthread_mutex_t urcu_defer_mutex = PTHREAD_MUTEX_INITIALIZER;
 static pthread_mutex_t defer_thread_mutex = PTHREAD_MUTEX_INITIALIZER;
 
+int defer_thread_futex;
+
 /*
  * Written to only by each individual deferer. Read by both the deferer and
  * the reclamation tread.
@@ -62,7 +70,28 @@ static struct deferer_registry *registry;
 static int num_deferers, alloc_deferers;
 
 static pthread_t tid_defer;
-static int exit_defer;
+
+/*
+ * Wake-up any waiting defer thread. Called from many concurrent threads.
+ */
+void wake_up_defer(void)
+{
+       if (unlikely(atomic_read(&defer_thread_futex) == -1))
+               atomic_set(&defer_thread_futex, 0);
+               futex(&defer_thread_futex, FUTEX_WAKE,
+                     0, NULL, NULL, 0);
+}
+
+/*
+ * Defer thread waiting. Single thread.
+ */
+static void wait_defer(void)
+{
+       atomic_dec(&defer_thread_futex);
+       if (atomic_read(&defer_thread_futex) == -1)
+               futex(&defer_thread_futex, FUTEX_WAIT, -1,
+                     NULL, NULL, 0);
+}
 
 static void internal_urcu_lock(pthread_mutex_t *mutex)
 {
@@ -81,6 +110,7 @@ static void internal_urcu_lock(pthread_mutex_t *mutex)
                        perror("Error in pthread mutex lock");
                        exit(-1);
                }
+               pthread_testcancel();
                poll(NULL,0,10);
        }
 #endif /* #else #ifndef DISTRUST_SIGNALS_EXTREME */
@@ -133,9 +163,12 @@ static void rcu_defer_barrier_queue(struct defer_queue *queue,
 
 static void _rcu_defer_barrier_thread(void)
 {
-       unsigned long head;
+       unsigned long head, num_items;
 
        head = defer_queue.head;
+       num_items = head - defer_queue.tail;
+       if (unlikely(!num_items))
+               return;
        synchronize_rcu();
        rcu_defer_barrier_queue(&defer_queue, head);
 }
@@ -148,30 +181,66 @@ void rcu_defer_barrier_thread(void)
        internal_urcu_unlock(&urcu_defer_mutex);
 }
 
+/*
+ * rcu_defer_barrier - Execute all queued rcu callbacks.
+ *
+ * Execute all RCU callbacks queued before rcu_defer_barrier() execution.
+ * All callbacks queued on the local thread prior to a rcu_defer_barrier() call
+ * are guaranteed to be executed.
+ * Callbacks queued by other threads concurrently with rcu_defer_barrier()
+ * execution are not guaranteed to be executed in the current batch (could
+ * be left for the next batch). These callbacks queued by other threads are only
+ * guaranteed to be executed if there is explicit synchronization between
+ * the thread adding to the queue and the thread issuing the defer_barrier call.
+ */
+
 void rcu_defer_barrier(void)
 {
        struct deferer_registry *index;
+       unsigned long num_items = 0;
 
        if (!registry)
                return;
 
        internal_urcu_lock(&urcu_defer_mutex);
-       for (index = registry; index < registry + num_deferers; index++)
+       for (index = registry; index < registry + num_deferers; index++) {
                index->last_head = LOAD_SHARED(index->defer_queue->head);
+               num_items += index->last_head - index->defer_queue->tail;
+       }
+       if (likely(!num_items)) {
+               /*
+                * We skip the grace period because there are no queued
+                * callbacks to execute.
+                */
+               goto end;
+       }
        synchronize_rcu();
        for (index = registry; index < registry + num_deferers; index++)
                rcu_defer_barrier_queue(index->defer_queue,
                                          index->last_head);
+end:
        internal_urcu_unlock(&urcu_defer_mutex);
 }
 
 void *thr_defer(void *args)
 {
        for (;;) {
-               if (LOAD_SHARED(exit_defer))
-                       break;
-               poll(NULL,0,100);       /* wait for 100ms */
+               pthread_testcancel();
+               printf("a\n");
+               /*
+                * "Be green". Don't wake up the CPU if there is no RCU work
+                * to perform whatsoever. Aims at saving laptop battery life by
+                * leaving the processor in sleep state when idle.
+                */
+               printf("b\n");
+               wait_defer();
+               printf("e\n");
+               /* Sleeping after wait_defer to let many callbacks enqueue */
+               //TEST poll(NULL,0,100);        /* wait for 100ms */
+               printf("f\n");
                rcu_defer_barrier();
+               printf("perform deferred call_rcu() from worker thread %lu.\n",
+                       time(NULL));
        }
 
        return NULL;
@@ -250,7 +319,8 @@ static void stop_defer_thread(void)
        int ret;
        void *tret;
 
-       STORE_SHARED(exit_defer, 1);
+       pthread_cancel(tid_defer);
+       wake_up_defer();
        ret = pthread_join(tid_defer, &tret);
        assert(!ret);
 }