Fix for https://issues.redhat.com/browse/MODCLUSTER-774

Author: jfclere

native/mod_proxy_cluster/mod_proxy_cluster.c

@@ -201,6 +201,11 @@ static void add_hcheck(server_rec *s, const proxy_server_conf *conf, proxy_worke
 
 static int (*ap_proxy_retry_worker_fn)(const char *proxy_function, proxy_worker *worker, server_rec *s) = NULL;
 
+static int is_worker_empty(proxy_worker *worker)
+{
+    return worker && (worker->s->port == 0 || worker->s->scheme[0] == '\0' || worker->s->hostname[0] == '\0');
+}
+
 static void check_workers(const proxy_server_conf *conf, const server_rec *s)
 {
     int i;
@@ -211,13 +216,12 @@ static void check_workers(const proxy_server_conf *conf, const server_rec *s)
         proxy_worker **workers;
         workers = (proxy_worker **)balancer->workers->elts;
         for (j = 0; j < balancer->workers->nelts; j++, workers++) {
-            volatile proxy_worker *worker = *workers;
+            proxy_worker *worker = *workers;
             proxy_cluster_helper *helper;
             int stop_worker = 0;
             helper = (proxy_cluster_helper *)worker->context;
             ap_assert(helper); /* we are in trouble ... */
-            if (worker->s->port == 0 && worker->s->scheme[0] == '\0' && worker->s->hostname[0] == '\0' &&
-                worker->s->route[0] == '\0') {
+            if (is_worker_empty(worker)) {
                 /* this happens when a new child process is created and it "cleaned" some old slotmem */
                 /* it is like the remove_workers_node we try to restore the non shared memory allocated in
                  * create_worker() */
@@ -796,7 +800,11 @@ static proxy_worker *get_worker_from_id_stat(const proxy_server_conf *conf, int
             proxy_worker **worker = (proxy_worker **)ptrw;
             proxy_cluster_helper *helper = (proxy_cluster_helper *)(*worker)->context;
             if ((*worker)->s == stat && helper->index == id) {
-                return *worker;
+                if (is_worker_empty(*worker)) {
+                    return NULL;
+                } else {
+                    return *worker;
+                }
             }
             if (helper->index == id) {
                 unpair_worker_node((*worker)->s, node);
@@ -2048,7 +2056,7 @@ static int proxy_node_get_free_id(request_rec *r, int node_table_size)
             proxy_worker **workers;
             workers = (proxy_worker **)balancer->workers->elts;
             for (j = 0; j < balancer->workers->nelts; j++, workers++) {
-                volatile proxy_worker *worker = *workers;
+                proxy_worker *worker = *workers;
                 proxy_cluster_helper *helper;
                 if (worker->s->index >= node_table_size) {
                     ap_log_error(APLOG_MARK, APLOG_WARNING, 0, r->server,