X-Git-Url: https://git.liburcu.org/?p=urcu.git;a=blobdiff_plain;f=urcu-call-rcu-impl.h;h=ae93468d8beb4e9fb9167a4dcc5b462892f3d9a2;hp=9beb58cbf59cfba7d8f82aae5e85fb8882db9e70;hb=53a55535da7db6ac6a2f37107baed9130c049333;hpb=739877210827591639f5448815e5f9186f4b31d2

diff --git a/urcu-call-rcu-impl.h b/urcu-call-rcu-impl.h
index 9beb58c..ae93468 100644
--- a/urcu-call-rcu-impl.h
+++ b/urcu-call-rcu-impl.h
@@ -26,11 +26,11 @@
 #include <signal.h>
 #include <assert.h>
 #include <stdlib.h>
+#include <stdint.h>
 #include <string.h>
 #include <errno.h>
 #include <poll.h>
 #include <sys/time.h>
-#include <syscall.h>
 #include <unistd.h>
 #include <sched.h>
 
@@ -39,15 +39,14 @@
 #include "urcu-call-rcu.h"
 #include "urcu-pointer.h"
 #include "urcu/list.h"
-#include "urcu/urcu-futex.h"
+#include "urcu/futex.h"
 
 /* Data structure that identifies a call_rcu thread. */
 
 struct call_rcu_data {
 	struct cds_wfq_queue cbs;
 	unsigned long flags;
-	pthread_mutex_t mtx;
-	int futex;
+	int32_t futex;
 	unsigned long qlen; /* maintained for debugging. */
 	pthread_t tid;
 	int cpu_affinity;
@@ -89,26 +88,6 @@ static struct call_rcu_data *default_call_rcu_data;
 static struct call_rcu_data **per_cpu_call_rcu_data;
 static long maxcpus;
 
-static void call_rcu_wait(struct call_rcu_data *crdp)
-{
-	/* Read call_rcu list before read futex */
-	cmm_smp_mb();
-	if (uatomic_read(&crdp->futex) == -1)
-		futex_async(&crdp->futex, FUTEX_WAIT, -1,
-		      NULL, NULL, 0);
-}
-
-static void call_rcu_wake_up(struct call_rcu_data *crdp)
-{
-	/* Write to call_rcu list before reading/writing futex */
-	cmm_smp_mb();
-	if (unlikely(uatomic_read(&crdp->futex) == -1)) {
-		uatomic_set(&crdp->futex, 0);
-		futex_async(&crdp->futex, FUTEX_WAKE, 1,
-		      NULL, NULL, 0);
-	}
-}
-
 /* Allocate the array if it has not already been allocated. */
 
 static void alloc_cpu_call_rcu_data(void)
@@ -136,7 +115,12 @@ static void alloc_cpu_call_rcu_data(void)
 
 #else /* #if defined(HAVE_SCHED_GETCPU) && defined(HAVE_SYSCONF) */
 
-static const struct call_rcu_data **per_cpu_call_rcu_data = NULL;
+/*
+ * per_cpu_call_rcu_data should be constant, but some functions below, used both
+ * for cases where cpu number is available and not available, assume it it not
+ * constant.
+ */
+static struct call_rcu_data **per_cpu_call_rcu_data = NULL;
 static const long maxcpus = -1;
 
 static void alloc_cpu_call_rcu_data(void)
@@ -195,6 +179,26 @@ int set_thread_cpu_affinity(struct call_rcu_data *crdp)
 }
 #endif
 
+static void call_rcu_wait(struct call_rcu_data *crdp)
+{
+	/* Read call_rcu list before read futex */
+	cmm_smp_mb();
+	if (uatomic_read(&crdp->futex) == -1)
+		futex_async(&crdp->futex, FUTEX_WAIT, -1,
+		      NULL, NULL, 0);
+}
+
+static void call_rcu_wake_up(struct call_rcu_data *crdp)
+{
+	/* Write to call_rcu list before reading/writing futex */
+	cmm_smp_mb();
+	if (unlikely(uatomic_read(&crdp->futex) == -1)) {
+		uatomic_set(&crdp->futex, 0);
+		futex_async(&crdp->futex, FUTEX_WAKE, 1,
+		      NULL, NULL, 0);
+	}
+}
+
 /* This is the code run by each call_rcu thread. */
 
 static void *call_rcu_thread(void *arg)
@@ -204,19 +208,25 @@ static void *call_rcu_thread(void *arg)
 	struct cds_wfq_node **cbs_tail;
 	struct call_rcu_data *crdp = (struct call_rcu_data *)arg;
 	struct rcu_head *rhp;
+	int rt = !!(uatomic_read(&crdp->flags) & URCU_CALL_RCU_RT);
 
 	if (set_thread_cpu_affinity(crdp) != 0) {
 		perror("pthread_setaffinity_np");
 		exit(-1);
 	}
 
+	/*
+	 * If callbacks take a read-side lock, we need to be registered.
+	 */
+	rcu_register_thread();
+
 	thread_call_rcu_data = crdp;
+	if (!rt) {
+		uatomic_dec(&crdp->futex);
+		/* Decrement futex before reading call_rcu list */
+		cmm_smp_mb();
+	}
 	for (;;) {
-		if (!(crdp->flags & URCU_CALL_RCU_RT)) {
-			uatomic_dec(&crdp->futex);
-			/* Decrement futex before reading call_rcu list */
-			cmm_smp_mb();
-		}
 		if (&crdp->cbs.head != _CMM_LOAD_SHARED(crdp->cbs.tail)) {
 			while ((cbs = _CMM_LOAD_SHARED(crdp->cbs.head)) == NULL)
 				poll(NULL, 0, 1);
@@ -240,25 +250,37 @@ static void *call_rcu_thread(void *arg)
 			} while (cbs != NULL);
 			uatomic_sub(&crdp->qlen, cbcount);
 		}
-		if (crdp->flags & URCU_CALL_RCU_STOP) {
-			if (!(crdp->flags & URCU_CALL_RCU_RT)) {
+		if (uatomic_read(&crdp->flags) & URCU_CALL_RCU_STOP)
+			break;
+		rcu_thread_offline();
+		if (!rt) {
+			if (&crdp->cbs.head
+			    == _CMM_LOAD_SHARED(crdp->cbs.tail)) {
+				call_rcu_wait(crdp);
+				poll(NULL, 0, 10);
+				uatomic_dec(&crdp->futex);
 				/*
-				 * Read call_rcu list before write futex.
+				 * Decrement futex before reading
+				 * call_rcu list.
 				 */
 				cmm_smp_mb();
-				uatomic_set(&crdp->futex, 0);
+			} else {
+				poll(NULL, 0, 10);
 			}
-			break;
+		} else {
+			poll(NULL, 0, 10);
 		}
-		if (!(crdp->flags & URCU_CALL_RCU_RT)) {
-			if (&crdp->cbs.head == _CMM_LOAD_SHARED(crdp->cbs.tail))
-				call_rcu_wait(crdp);
-		}
-		poll(NULL, 0, 10);
+		rcu_thread_online();
 	}
-	call_rcu_lock(&crdp->mtx);
-	crdp->flags |= URCU_CALL_RCU_STOPPED;
-	call_rcu_unlock(&crdp->mtx);
+	if (!rt) {
+		/*
+		 * Read call_rcu list before write futex.
+		 */
+		cmm_smp_mb();
+		uatomic_set(&crdp->futex, 0);
+	}
+	uatomic_or(&crdp->flags, URCU_CALL_RCU_STOPPED);
+	rcu_unregister_thread();
 	return NULL;
 }
 
@@ -282,10 +304,6 @@ static void call_rcu_data_init(struct call_rcu_data **crdpp,
 	memset(crdp, '\0', sizeof(*crdp));
 	cds_wfq_init(&crdp->cbs);
 	crdp->qlen = 0;
-	if (pthread_mutex_init(&crdp->mtx, NULL) != 0) {
-		perror("pthread_mutex_init");
-		exit(-1);
-	}
 	crdp->futex = 0;
 	crdp->flags = flags;
 	cds_list_add(&crdp->list, &call_rcu_data_list);
@@ -365,9 +383,10 @@ struct call_rcu_data *create_call_rcu_data(unsigned long flags,
 
 int set_cpu_call_rcu_data(int cpu, struct call_rcu_data *crdp)
 {
-	int warned = 0;
+	static int warned = 0;
 
 	call_rcu_lock(&call_rcu_mutex);
+	alloc_cpu_call_rcu_data();
 	if (cpu < 0 || maxcpus <= cpu) {
 		if (!warned) {
 			fprintf(stderr, "[error] liburcu: set CPU # out of range\n");
@@ -377,13 +396,21 @@ int set_cpu_call_rcu_data(int cpu, struct call_rcu_data *crdp)
 		errno = EINVAL;
 		return -EINVAL;
 	}
-	alloc_cpu_call_rcu_data();
-	call_rcu_unlock(&call_rcu_mutex);
+
 	if (per_cpu_call_rcu_data == NULL) {
+		call_rcu_unlock(&call_rcu_mutex);
 		errno = ENOMEM;
 		return -ENOMEM;
 	}
+
+	if (per_cpu_call_rcu_data[cpu] != NULL && crdp != NULL) {
+		call_rcu_unlock(&call_rcu_mutex);
+		errno = EEXIST;
+		return -EEXIST;
+	}
+
 	per_cpu_call_rcu_data[cpu] = crdp;
+	call_rcu_unlock(&call_rcu_mutex);
 	return 0;
 }
 
@@ -417,22 +444,17 @@ struct call_rcu_data *get_default_call_rcu_data(void)
  */
 struct call_rcu_data *get_call_rcu_data(void)
 {
-	int curcpu;
-	static int warned = 0;
+	struct call_rcu_data *crd;
 
 	if (thread_call_rcu_data != NULL)
 		return thread_call_rcu_data;
-	if (maxcpus <= 0)
-		return get_default_call_rcu_data();
-	curcpu = sched_getcpu();
-	if (!warned && (curcpu < 0 || maxcpus <= curcpu)) {
-		fprintf(stderr, "[error] liburcu: gcrd CPU # out of range\n");
-		warned = 1;
+
+	if (maxcpus > 0) {
+		crd = get_cpu_call_rcu_data(sched_getcpu());
+		if (crd)
+			return crd;
 	}
-	if (curcpu >= 0 && maxcpus > curcpu &&
-	    per_cpu_call_rcu_data != NULL &&
-	    per_cpu_call_rcu_data[curcpu] != NULL)
-	    	return per_cpu_call_rcu_data[curcpu];
+
 	return get_default_call_rcu_data();
 }
 
@@ -464,7 +486,9 @@ void set_thread_call_rcu_data(struct call_rcu_data *crdp)
 /*
  * Create a separate call_rcu thread for each CPU.  This does not
  * replace a pre-existing call_rcu thread -- use the set_cpu_call_rcu_data()
- * function if you want that behavior.
+ * function if you want that behavior. Should be paired with
+ * free_all_cpu_call_rcu_data() to teardown these call_rcu worker
+ * threads.
  */
 
 int create_all_cpu_call_rcu_data(unsigned long flags)
@@ -568,12 +592,10 @@ void call_rcu_data_free(struct call_rcu_data *crdp)
 	if (crdp == NULL || crdp == default_call_rcu_data) {
 		return;
 	}
-	if ((crdp->flags & URCU_CALL_RCU_STOPPED) == 0) {
-		call_rcu_lock(&crdp->mtx);
-		crdp->flags |= URCU_CALL_RCU_STOP;
-		call_rcu_unlock(&crdp->mtx);
+	if ((uatomic_read(&crdp->flags) & URCU_CALL_RCU_STOPPED) == 0) {
+		uatomic_or(&crdp->flags, URCU_CALL_RCU_STOP);
 		wake_call_rcu_thread(crdp);
-		while ((crdp->flags & URCU_CALL_RCU_STOPPED) == 0)
+		while ((uatomic_read(&crdp->flags) & URCU_CALL_RCU_STOPPED) == 0)
 			poll(NULL, 0, 1);
 	}
 	if (&crdp->cbs.head != _CMM_LOAD_SHARED(crdp->cbs.tail)) {
@@ -587,9 +609,11 @@ void call_rcu_data_free(struct call_rcu_data *crdp)
 		*cbs_endprev = cbs;
 		uatomic_add(&default_call_rcu_data->qlen,
 			    uatomic_read(&crdp->qlen));
-		cds_list_del(&crdp->list);
-		free(crdp);
+		wake_call_rcu_thread(default_call_rcu_data);
 	}
+
+	cds_list_del(&crdp->list);
+	free(crdp);
 }
 
 /*
@@ -638,7 +662,7 @@ void call_rcu_after_fork_parent(void)
  */
 void call_rcu_after_fork_child(void)
 {
-	struct call_rcu_data *crdp;
+	struct call_rcu_data *crdp, *next;
 
 	/* Release the mutex. */
 	call_rcu_unlock(&call_rcu_mutex);
@@ -651,13 +675,10 @@ void call_rcu_after_fork_child(void)
 	(void)get_default_call_rcu_data();
 
 	/* Dispose of all of the rest of the call_rcu_data structures. */
-	while (call_rcu_data_list.next != call_rcu_data_list.prev) {
-		crdp = cds_list_entry(call_rcu_data_list.prev,
-				      struct call_rcu_data, list);
+	cds_list_for_each_entry_safe(crdp, next, &call_rcu_data_list, list) {
 		if (crdp == default_call_rcu_data)
-			crdp = cds_list_entry(crdp->list.prev,
-					      struct call_rcu_data, list);
-		crdp->flags = URCU_CALL_RCU_STOPPED;
+			continue;
+		uatomic_set(&crdp->flags, URCU_CALL_RCU_STOPPED);
 		call_rcu_data_free(crdp);
 	}
 }