From: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Date: Wed, 19 Dec 2012 00:31:21 +0000 (-0500)
Subject: Fix call_rcu fork handling
X-Git-Tag: v0.6.9~5
X-Git-Url: https://git.lttng.org./?a=commitdiff_plain;h=483c853e4227f9d0b3ad03f7a4bd8038fe26da89;p=userspace-rcu.git

Fix call_rcu fork handling

Fix call_rcu fork handling by putting all call_rcu threads in a
quiescent state before fork (paused state), and unpausing them when the
parent returns from fork.

On the child, everything will run fine as long as we don't issue fork()
from a call_rcu callback.

Side-note: pthread_atfork is not appropriate when using with multithread
and malloc/free. The glibc malloc implementation sadly expects that all
malloc/free are executed from the context of a single thread while
pthread atfork handlers are running, which leads to interesting hang in
glibc.

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

diff --git a/urcu-call-rcu-impl.h b/urcu-call-rcu-impl.h
index e38123b..6ea7b9a 100644
--- a/urcu-call-rcu-impl.h
+++ b/urcu-call-rcu-impl.h
@@ -64,8 +64,9 @@ CDS_LIST_HEAD(call_rcu_data_list);
 
 static __thread struct call_rcu_data *thread_call_rcu_data;
 
-/* Guard call_rcu thread creation. */
-
+/*
+ * Guard call_rcu thread creation and atfork handlers.
+ */
 static pthread_mutex_t call_rcu_mutex = PTHREAD_MUTEX_INITIALIZER;
 
 /* If a given thread does not have its own call_rcu thread, this is default. */
@@ -239,6 +240,21 @@ static void *call_rcu_thread(void *arg)
 		cmm_smp_mb();
 	}
 	for (;;) {
+		if (uatomic_read(&crdp->flags) & URCU_CALL_RCU_PAUSE) {
+			/*
+			 * Pause requested. Become quiescent: remove
+			 * ourself from all global lists, and don't
+			 * process any callback. The callback lists may
+			 * still be non-empty though.
+			 */
+			rcu_unregister_thread();
+			cmm_smp_mb__before_uatomic_or();
+			uatomic_or(&crdp->flags, URCU_CALL_RCU_PAUSED);
+			while ((uatomic_read(&crdp->flags) & URCU_CALL_RCU_PAUSE) != 0)
+				poll(NULL, 0, 1);
+			rcu_register_thread();
+		}
+
 		if (&crdp->cbs.head != _CMM_LOAD_SHARED(crdp->cbs.tail)) {
 			while ((cbs = _CMM_LOAD_SHARED(crdp->cbs.head)) == NULL)
 				poll(NULL, 0, 1);
@@ -702,12 +718,25 @@ void free_all_cpu_call_rcu_data(void)
 
 /*
  * Acquire the call_rcu_mutex in order to ensure that the child sees
- * all of the call_rcu() data structures in a consistent state.
+ * all of the call_rcu() data structures in a consistent state. Ensure
+ * that all call_rcu threads are in a quiescent state across fork.
  * Suitable for pthread_atfork() and friends.
  */
 void call_rcu_before_fork(void)
 {
+	struct call_rcu_data *crdp;
+
 	call_rcu_lock(&call_rcu_mutex);
+
+	cds_list_for_each_entry(crdp, &call_rcu_data_list, list) {
+		uatomic_or(&crdp->flags, URCU_CALL_RCU_PAUSE);
+		cmm_smp_mb__after_uatomic_or();
+		wake_call_rcu_thread(crdp);
+	}
+	cds_list_for_each_entry(crdp, &call_rcu_data_list, list) {
+		while ((uatomic_read(&crdp->flags) & URCU_CALL_RCU_PAUSED) == 0)
+			poll(NULL, 0, 1);
+	}
 }
 
 /*
@@ -717,6 +746,10 @@ void call_rcu_before_fork(void)
  */
 void call_rcu_after_fork_parent(void)
 {
+	struct call_rcu_data *crdp;
+
+	cds_list_for_each_entry(crdp, &call_rcu_data_list, list)
+		uatomic_and(&crdp->flags, ~URCU_CALL_RCU_PAUSE);
 	call_rcu_unlock(&call_rcu_mutex);
 }
 
@@ -749,7 +782,11 @@ void call_rcu_after_fork_child(void)
 	rcu_set_pointer(&per_cpu_call_rcu_data, NULL);
 	thread_call_rcu_data = NULL;
 
-	/* Dispose of all of the rest of the call_rcu_data structures. */
+	/*
+	 * Dispose of all of the rest of the call_rcu_data structures.
+	 * Leftover call_rcu callbacks will be merged into the new
+	 * default call_rcu thread queue.
+	 */
 	cds_list_for_each_entry_safe(crdp, next, &call_rcu_data_list, list) {
 		if (crdp == default_call_rcu_data)
 			continue;
diff --git a/urcu-call-rcu.h b/urcu-call-rcu.h
index b109f4e..35b31ab 100644
--- a/urcu-call-rcu.h
+++ b/urcu-call-rcu.h
@@ -44,10 +44,12 @@ struct call_rcu_data;
 
 /* Flag values. */
 
-#define URCU_CALL_RCU_RT	0x1
-#define URCU_CALL_RCU_RUNNING	0x2
-#define URCU_CALL_RCU_STOP	0x4
-#define URCU_CALL_RCU_STOPPED	0x8
+#define URCU_CALL_RCU_RT	(1U << 0)
+#define URCU_CALL_RCU_RUNNING	(1U << 1)
+#define URCU_CALL_RCU_STOP	(1U << 2)
+#define URCU_CALL_RCU_STOPPED	(1U << 3)
+#define URCU_CALL_RCU_PAUSE	(1U << 4)
+#define URCU_CALL_RCU_PAUSED	(1U << 5)
 
 /*
  * The rcu_head data structure is placed in the structure to be freed