Fix: Sync issue when deleting a data stream
authorDavid Goulet <dgoulet@efficios.com>
Thu, 1 Nov 2012 20:27:08 +0000 (16:27 -0400)
committerDavid Goulet <dgoulet@efficios.com>
Thu, 1 Nov 2012 20:34:06 +0000 (16:34 -0400)
A data stream was rescheduled for deletion after a flush on hang up.
Once in a normal read code path, on error, the stream is deleted and
then processed for ERR|HUP error which could also delete again the
stream causing an assert() and a failed trace.

We fix that by setting the local array to NULL for that stream once
deleted and ignoring the stream is subsequent loop if NULL.

Fixes #390

Acked-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: David Goulet <dgoulet@efficios.com>
src/common/consumer.c

index 41f8ae53fe67adde7552b54dfb31aa89d4abedcb..ca6aeba3ffd7a8748256bd9cae356eeff3e977b6 100644 (file)
@@ -288,6 +288,8 @@ static void cleanup_relayd(struct consumer_relayd_sock_pair *relayd,
 
        assert(relayd);
 
+       DBG("Cleaning up relayd sockets");
+
        /* Save the net sequence index before destroying the object */
        netidx = relayd->net_seq_idx;
 
@@ -1941,7 +1943,7 @@ static void validate_endpoint_status_data_stream(void)
        rcu_read_lock();
        cds_lfht_for_each_entry(data_ht->ht, &iter.iter, stream, node.node) {
                /* Validate delete flag of the stream */
-               if (!stream->endpoint_status) {
+               if (stream->endpoint_status != CONSUMER_ENDPOINT_INACTIVE) {
                        continue;
                }
                /* Delete it right now */
@@ -2307,6 +2309,9 @@ void *consumer_thread_data_poll(void *data)
 
                /* Take care of high priority channels first. */
                for (i = 0; i < nb_fd; i++) {
+                       if (local_stream[i] == NULL) {
+                               continue;
+                       }
                        if (pollfd[i].revents & POLLPRI) {
                                DBG("Urgent read on fd %d", pollfd[i].fd);
                                high_prio = 1;
@@ -2315,6 +2320,7 @@ void *consumer_thread_data_poll(void *data)
                                if (len < 0 && len != -EAGAIN && len != -ENODATA) {
                                        /* Clean the stream and free it. */
                                        consumer_del_stream(local_stream[i], data_ht);
+                                       local_stream[i] = NULL;
                                } else if (len > 0) {
                                        local_stream[i]->data_read = 1;
                                }
@@ -2331,6 +2337,9 @@ void *consumer_thread_data_poll(void *data)
 
                /* Take care of low priority channels. */
                for (i = 0; i < nb_fd; i++) {
+                       if (local_stream[i] == NULL) {
+                               continue;
+                       }
                        if ((pollfd[i].revents & POLLIN) ||
                                        local_stream[i]->hangup_flush_done) {
                                DBG("Normal read on fd %d", pollfd[i].fd);
@@ -2339,6 +2348,7 @@ void *consumer_thread_data_poll(void *data)
                                if (len < 0 && len != -EAGAIN && len != -ENODATA) {
                                        /* Clean the stream and free it. */
                                        consumer_del_stream(local_stream[i], data_ht);
+                                       local_stream[i] = NULL;
                                } else if (len > 0) {
                                        local_stream[i]->data_read = 1;
                                }
@@ -2347,12 +2357,15 @@ void *consumer_thread_data_poll(void *data)
 
                /* Handle hangup and errors */
                for (i = 0; i < nb_fd; i++) {
+                       if (local_stream[i] == NULL) {
+                               continue;
+                       }
                        if (!local_stream[i]->hangup_flush_done
                                        && (pollfd[i].revents & (POLLHUP | POLLERR | POLLNVAL))
                                        && (consumer_data.type == LTTNG_CONSUMER32_UST
                                                || consumer_data.type == LTTNG_CONSUMER64_UST)) {
                                DBG("fd %d is hup|err|nval. Attempting flush and read.",
-                                       pollfd[i].fd);
+                                               pollfd[i].fd);
                                lttng_ustconsumer_on_stream_hangup(local_stream[i]);
                                /* Attempt read again, for the data we just flushed. */
                                local_stream[i]->data_read = 1;
@@ -2366,22 +2379,27 @@ void *consumer_thread_data_poll(void *data)
                                DBG("Polling fd %d tells it has hung up.", pollfd[i].fd);
                                if (!local_stream[i]->data_read) {
                                        consumer_del_stream(local_stream[i], data_ht);
+                                       local_stream[i] = NULL;
                                        num_hup++;
                                }
                        } else if (pollfd[i].revents & POLLERR) {
                                ERR("Error returned in polling fd %d.", pollfd[i].fd);
                                if (!local_stream[i]->data_read) {
                                        consumer_del_stream(local_stream[i], data_ht);
+                                       local_stream[i] = NULL;
                                        num_hup++;
                                }
                        } else if (pollfd[i].revents & POLLNVAL) {
                                ERR("Polling fd %d tells fd is not open.", pollfd[i].fd);
                                if (!local_stream[i]->data_read) {
                                        consumer_del_stream(local_stream[i], data_ht);
+                                       local_stream[i] = NULL;
                                        num_hup++;
                                }
                        }
-                       local_stream[i]->data_read = 0;
+                       if (local_stream[i] != NULL) {
+                               local_stream[i]->data_read = 0;
+                       }
                }
        }
 end:
This page took 0.038605 seconds and 4 git commands to generate.