Fix: LPOLLHUP and LPOLLERR when there is still data in pipe/socket

[lttng-tools.git] / src / bin / lttng-sessiond / ht-cleanup.c

diff --git a/src/bin/lttng-sessiond/ht-cleanup.c b/src/bin/lttng-sessiond/ht-cleanup.c
index 890c9a8215e26967c5ff6b42b5cf171295cdc306..c477e1a649624fd38d9b82e4ec66ccc970689d28 100644 (file)
--- a/src/bin/lttng-sessiond/ht-cleanup.c
+++ b/src/bin/lttng-sessiond/ht-cleanup.c
@@ -24,10 +24,12 @@
 
 #include "lttng-sessiond.h"
 #include "health-sessiond.h"
+#include "testpoint.h"
 
 void *thread_ht_cleanup(void *data)
 {
        int ret, i, pollfd, err = -1;
+       ssize_t size_ret;
        uint32_t revents, nb_fd;
        struct lttng_poll_event events;
 
@@ -38,6 +40,10 @@ void *thread_ht_cleanup(void *data)
 
        health_register(health_sessiond, HEALTH_SESSIOND_TYPE_HT_CLEANUP);
 
+       if (testpoint(sessiond_thread_ht_cleanup)) {
+               goto error_testpoint;
+       }
+
        health_code_update();
 
        ret = sessiond_set_thread_pollset(&events, 2);
@@ -54,13 +60,14 @@ void *thread_ht_cleanup(void *data)
        health_code_update();
 
        while (1) {
-               DBG3("[ht-thread] Polling on %d fds.",
-                       LTTNG_POLL_GETNB(&events));
+               DBG3("[ht-thread] Polling.");
 
                /* Inifinite blocking call, waiting for transmission */
 restart:
                health_poll_entry();
                ret = lttng_poll_wait(&events, -1);
+               DBG3("[ht-thread] Returning from poll on %d fds.",
+                       LTTNG_POLL_GETNB(&events));
                health_poll_exit();
                if (ret < 0) {
                        /*
@@ -83,6 +90,14 @@ restart:
                        revents = LTTNG_POLL_GETEV(&events, i);
                        pollfd = LTTNG_POLL_GETFD(&events, i);
 
+                       if (!revents) {
+                               /*
+                                * No activity for this FD
+                                * (poll implementation).
+                                */
+                               continue;
+                       }
+
                        /* Thread quit pipe has been closed. Killing thread. */
                        ret = sessiond_check_thread_quit_pipe(pollfd, revents);
                        if (ret) {
@@ -91,33 +106,31 @@ restart:
                        }
                        assert(pollfd == ht_cleanup_pipe[0]);
 
-                       if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
+                       if (revents & LPOLLIN) {
+                               /* Get socket from dispatch thread. */
+                               size_ret = lttng_read(ht_cleanup_pipe[0], &ht,
+                                               sizeof(ht));
+                               if (size_ret < sizeof(ht)) {
+                                       PERROR("ht cleanup notify pipe");
+                                       goto error;
+                               }
+                               health_code_update();
+                               /*
+                                * The whole point of this thread is to call
+                                * lttng_ht_destroy from a context that is NOT:
+                                * 1) a read-side RCU lock,
+                                * 2) a call_rcu thread.
+                                */
+                               lttng_ht_destroy(ht);
+
+                               health_code_update();
+                       } else if (revents & (LPOLLERR | LPOLLHUP | LPOLLRDHUP)) {
                                ERR("ht cleanup pipe error");
                                goto error;
-                       } else if (!(revents & LPOLLIN)) {
-                               /* No POLLIN and not a catched error, stop the thread. */
-                               ERR("ht cleanup failed. revent: %u", revents);
-                               goto error;
-                       }
-
-                       do {
-                               /* Get socket from dispatch thread. */
-                               ret = read(ht_cleanup_pipe[0], &ht, sizeof(ht));
-                       } while (ret < 0 && errno == EINTR);
-                       if (ret < 0 || ret < sizeof(ht)) {
-                               PERROR("ht cleanup notify pipe");
+                       } else {
+                               ERR("Unexpected poll events %u for sock %d", revents, pollfd);
                                goto error;
                        }
-                       health_code_update();
-                       /*
-                        * The whole point of this thread is to call
-                        * lttng_ht_destroy from a context that is NOT:
-                        * 1) a read-side RCU lock,
-                        * 2) a call_rcu thread.
-                        */
-                       lttng_ht_destroy(ht);
-
-                       health_code_update();
                }
        }
 
@@ -125,6 +138,7 @@ exit:
 error:
        lttng_poll_clean(&events);
 error_poll_create:
+error_testpoint:
        utils_close_pipe(ht_cleanup_pipe);
        ht_cleanup_pipe[0] = ht_cleanup_pipe[1] = -1;
        DBG("[ust-thread] cleanup complete.");