X-Git-Url: https://git.lttng.org/?a=blobdiff_plain;f=src%2Fbin%2Flttng-sessiond%2Fust-thread.c;h=080afa53a692487d2e7bc12f29ae0bee9c932f72;hb=3a89d11a0ad0ca327e2df99eaf13d3a0c63e3af3;hp=76d6ef99da84d31c77f89ef02de72d18931940f6;hpb=d0b96690836f4b876096f3dc14801f8e25281a77;p=lttng-tools.git

diff --git a/src/bin/lttng-sessiond/ust-thread.c b/src/bin/lttng-sessiond/ust-thread.c
index 76d6ef99d..080afa53a 100644
--- a/src/bin/lttng-sessiond/ust-thread.c
+++ b/src/bin/lttng-sessiond/ust-thread.c
@@ -23,13 +23,16 @@
 #include "fd-limit.h"
 #include "lttng-sessiond.h"
 #include "ust-thread.h"
+#include "health-sessiond.h"
+#include "testpoint.h"
 
 /*
  * This thread manage application notify communication.
  */
 void *ust_thread_manage_notify(void *data)
 {
-	int i, ret, pollfd;
+	int i, ret, pollfd, err = -1;
+	ssize_t size_ret;
 	uint32_t revents, nb_fd;
 	struct lttng_poll_event events;
 
@@ -38,6 +41,15 @@ void *ust_thread_manage_notify(void *data)
 	rcu_register_thread();
 	rcu_thread_online();
 
+	health_register(health_sessiond,
+		HEALTH_SESSIOND_TYPE_APP_MANAGE_NOTIFY);
+
+	if (testpoint(sessiond_thread_app_manage_notify)) {
+		goto error_testpoint;
+	}
+
+	health_code_update();
+
 	ret = sessiond_set_thread_pollset(&events, 2);
 	if (ret < 0) {
 		goto error_poll_create;
@@ -49,13 +61,17 @@ void *ust_thread_manage_notify(void *data)
 		goto error;
 	}
 
+	health_code_update();
+
 	while (1) {
 		DBG3("[ust-thread] Manage notify polling on %d fds",
 				LTTNG_POLL_GETNB(&events));
 
 		/* Inifinite blocking call, waiting for transmission */
 restart:
+		health_poll_entry();
 		ret = lttng_poll_wait(&events, -1);
+		health_poll_exit();
 		if (ret < 0) {
 			/*
 			 * Restart interrupted system call.
@@ -69,6 +85,8 @@ restart:
 		nb_fd = ret;
 
 		for (i = 0; i < nb_fd; i++) {
+			health_code_update();
+
 			/* Fetch once the poll data */
 			revents = LTTNG_POLL_GETEV(&events, i);
 			pollfd = LTTNG_POLL_GETFD(&events, i);
@@ -76,6 +94,7 @@ restart:
 			/* Thread quit pipe has been closed. Killing thread. */
 			ret = sessiond_check_thread_quit_pipe(pollfd, revents);
 			if (ret) {
+				err = 0;
 				goto exit;
 			}
 
@@ -92,14 +111,14 @@ restart:
 					goto error;
 				}
 
-				do {
-					/* Get socket from dispatch thread. */
-					ret = read(apps_cmd_notify_pipe[0], &sock, sizeof(sock));
-				} while (ret < 0 && errno == EINTR);
-				if (ret < 0 || ret < sizeof(sock)) {
+				/* Get socket from dispatch thread. */
+				size_ret = lttng_read(apps_cmd_notify_pipe[0],
+						&sock, sizeof(sock));
+				if (size_ret < sizeof(sock)) {
 					PERROR("read apps notify pipe");
 					goto error;
 				}
+				health_code_update();
 
 				ret = lttng_poll_add(&events, sock,
 						LPOLLIN | LPOLLERR | LPOLLHUP | LPOLLRDHUP);
@@ -128,29 +147,26 @@ restart:
 						goto error;
 					}
 
-					ret = close(pollfd);
-					if (ret < 0) {
-						PERROR("close sock %d", pollfd);
-					}
-					lttng_fd_put(LTTNG_FD_APPS, 1);
+					/* The socket is closed after a grace period here. */
+					ust_app_notify_sock_unregister(pollfd);
 				} else if (revents & (LPOLLIN | LPOLLPRI)) {
 					ret = ust_app_recv_notify(pollfd);
 					if (ret < 0) {
-						ret = lttng_poll_del(&events, pollfd);
-						if (ret < 0) {
-							goto error;
-						}
-
-						ret = close(pollfd);
-						if (ret < 0) {
-							PERROR("close sock %d", pollfd);
-						}
-						lttng_fd_put(LTTNG_FD_APPS, 1);
+						/*
+						 * If the notification failed either the application is
+						 * dead or an internal error happened. In both cases,
+						 * we can only continue here. If the application is
+						 * dead, an unregistration will follow or else the
+						 * application will notice that we are not responding
+						 * on that socket and will close it.
+						 */
+						continue;
 					}
 				} else {
 					ERR("Unknown poll events %u for sock %d", revents, pollfd);
 					continue;
 				}
+				health_code_update();
 			}
 		}
 	}
@@ -159,9 +175,15 @@ exit:
 error:
 	lttng_poll_clean(&events);
 error_poll_create:
+error_testpoint:
 	utils_close_pipe(apps_cmd_notify_pipe);
 	apps_cmd_notify_pipe[0] = apps_cmd_notify_pipe[1] = -1;
 	DBG("Application notify communication apps thread cleanup complete");
+	if (err) {
+		health_error();
+		ERR("Health error occurred in %s", __func__);
+	}
+	health_unregister(health_sessiond);
 	rcu_thread_offline();
 	rcu_unregister_thread();
 	return NULL;