Events: support for EPOLLEXCLUSIVE.

This flag appeared in Linux 4.5 and is useful for avoiding thundering herd
problem.

The current Linux kernel implementation walks the list of exclusive waiters,
and queues an event to each epfd, until it finds the first waiter that has
threads blocked on it via epoll_wait().

5 files changed, 58 insertions, 3 deletions

diff --git a/auto/modules b/auto/modules
index 829079842..614037cb1 100644
--- a/auto/modules
+++ b/auto/modules
@@ -43,6 +43,7 @@ fi
 if [ $NGX_TEST_BUILD_EPOLL = YES ]; then
     have=NGX_HAVE_EPOLL . auto/have
     have=NGX_HAVE_EPOLLRDHUP . auto/have
+    have=NGX_HAVE_EPOLLEXCLUSIVE . auto/have
     have=NGX_HAVE_EVENTFD . auto/have
     have=NGX_TEST_BUILD_EPOLL . auto/have
     EVENT_MODULES="$EVENT_MODULES $EPOLL_MODULE"
diff --git a/auto/os/linux b/auto/os/linux
index 132ce3b42..fae8842c6 100644
--- a/auto/os/linux
+++ b/auto/os/linux
@@ -70,6 +70,22 @@ if [ $ngx_found = yes ]; then
                       ee.data.ptr = NULL;
                       epoll_ctl(efd, EPOLL_CTL_ADD, fd, &ee)"
     . auto/feature
+
+
+    # EPOLLEXCLUSIVE appeared in Linux 4.5, glibc 2.24
+
+    ngx_feature="EPOLLEXCLUSIVE"
+    ngx_feature_name="NGX_HAVE_EPOLLEXCLUSIVE"
+    ngx_feature_run=no
+    ngx_feature_incs="#include <sys/epoll.h>"
+    ngx_feature_path=
+    ngx_feature_libs=
+    ngx_feature_test="int efd = 0, fd = 0;
+                      struct epoll_event ee;
+                      ee.events = EPOLLIN|EPOLLEXCLUSIVE;
+                      ee.data.ptr = NULL;
+                      epoll_ctl(efd, EPOLL_CTL_ADD, fd, &ee)"
+    . auto/feature
 fi
 
 
diff --git a/src/event/modules/ngx_epoll_module.c b/src/event/modules/ngx_epoll_module.c
index dff778831..c267fd6c7 100644
--- a/src/event/modules/ngx_epoll_module.c
+++ b/src/event/modules/ngx_epoll_module.c
@@ -27,6 +27,7 @@
 
 #define EPOLLRDHUP     0x2000
 
+#define EPOLLEXCLUSIVE 0x10000000
 #define EPOLLONESHOT   0x40000000
 #define EPOLLET        0x80000000
 
@@ -610,6 +611,12 @@ ngx_epoll_add_event(ngx_event_t *ev, ngx_int_t event, ngx_uint_t flags)
         op = EPOLL_CTL_ADD;
     }
 
+#if (NGX_HAVE_EPOLLEXCLUSIVE && NGX_HAVE_EPOLLRDHUP)
+    if (flags & NGX_EXCLUSIVE_EVENT) {
+        events &= ~EPOLLRDHUP;
+    }
+#endif
+
     ee.events = events | (uint32_t) flags;
     ee.data.ptr = (void *) ((uintptr_t) c | ev->instance);
 
diff --git a/src/event/ngx_event.c b/src/event/ngx_event.c
index 2f3a098de..9d6c4c91b 100644
--- a/src/event/ngx_event.c
+++ b/src/event/ngx_event.c
@@ -822,15 +822,38 @@ ngx_event_process_init(ngx_cycle_t *cycle)
         rev->handler = (c->type == SOCK_STREAM) ? ngx_event_accept
                                                 : ngx_event_recvmsg;
 
-        if (ngx_use_accept_mutex
 #if (NGX_HAVE_REUSEPORT)
-            && !ls[i].reuseport
+
+        if (ls[i].reuseport) {
+            if (ngx_add_event(rev, NGX_READ_EVENT, 0) == NGX_ERROR) {
+                return NGX_ERROR;
+            }
+
+            continue;
+        }
+
 #endif
-           )
+
+        if (ngx_use_accept_mutex) {
+            continue;
+        }
+
+#if (NGX_HAVE_EPOLLEXCLUSIVE)
+
+        if ((ngx_event_flags & NGX_USE_EPOLL_EVENT)
+            && ccf->worker_processes > 1)
         {
+            if (ngx_add_event(rev, NGX_READ_EVENT, NGX_EXCLUSIVE_EVENT)
+                == NGX_ERROR)
+            {
+                return NGX_ERROR;
+            }
+
             continue;
         }
 
+#endif
+
         if (ngx_add_event(rev, NGX_READ_EVENT, 0) == NGX_ERROR) {
             return NGX_ERROR;
         }
diff --git a/src/event/ngx_event.h b/src/event/ngx_event.h
index 394a0e8e5..27139ee8c 100644
--- a/src/event/ngx_event.h
+++ b/src/event/ngx_event.h
@@ -367,6 +367,9 @@ extern ngx_uint_t            ngx_use_epoll_rdhup;
 #define NGX_ONESHOT_EVENT  EPOLLONESHOT
 #endif
 
+#if (NGX_HAVE_EPOLLEXCLUSIVE)
+#define NGX_EXCLUSIVE_EVENT  EPOLLEXCLUSIVE
+#endif
 
 #elif (NGX_HAVE_POLL)
 
@@ -395,6 +398,11 @@ extern ngx_uint_t            ngx_use_epoll_rdhup;
 #endif
 
 
+#if (NGX_TEST_BUILD_EPOLL)
+#define NGX_EXCLUSIVE_EVENT  0
+#endif
+
+
 #ifndef NGX_CLEAR_EVENT
 #define NGX_CLEAR_EVENT    0    /* dummy declaration */
 #endif