From: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Date: Thu, 25 Apr 2013 18:45:17 +0000 (-0400)
Subject: Fix: consumer metadata switch timer error handling
X-Git-Tag: v2.2.0-rc2~38
X-Git-Url: https://git.lttng.org./?a=commitdiff_plain;h=4419b4fb187cb896a8e64bb97c5471240abca122;p=lttng-tools.git

Fix: consumer metadata switch timer error handling

The thread handling the metadata switch cannot call
consumer_timer_switch_stop(), because it would become stucked waiting
for itself to execute the teardown signal handler.

Moreover, only one thread is allowed to call start/stop timer (by
design). Therefore, it's a race to have the thread handling the timer
handler calling "stop".

Fix this by adding a "switch_timer_error" flag, which just inhibits the
timer if an error is encountered, but does not modify its state
otherwise.

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@efficios.com>
Signed-off-by: David Goulet <dgoulet@efficios.com>
---

diff --git a/src/common/consumer-timer.c b/src/common/consumer-timer.c
index ef056d118..641478697 100644
--- a/src/common/consumer-timer.c
+++ b/src/common/consumer-timer.c
@@ -61,17 +61,17 @@ static void metadata_switch_timer(struct lttng_consumer_local_data *ctx,
 	channel = si->si_value.sival_ptr;
 	assert(channel);
 
+	if (channel->switch_timer_error) {
+		return;
+	}
+
 	DBG("Switch timer for channel %" PRIu64, channel->key);
 	switch (ctx->type) {
 	case LTTNG_CONSUMER32_UST:
 	case LTTNG_CONSUMER64_UST:
 		ret = lttng_ustconsumer_request_metadata(ctx, channel);
 		if (ret < 0) {
-			/*
-			 * An error means that we were unable to request the metadata to
-			 * the session daemon so stop the timer for that channel.
-			 */
-			consumer_timer_switch_stop(channel);
+			channel->switch_timer_error = 1;
 		}
 		break;
 	case LTTNG_CONSUMER_KERNEL:
@@ -83,6 +83,8 @@ static void metadata_switch_timer(struct lttng_consumer_local_data *ctx,
 
 /*
  * Set the timer for periodical metadata flush.
+ * Should be called only from the recv cmd thread (single thread ensures
+ * mutual exclusion).
  */
 void consumer_timer_switch_start(struct lttng_consumer_channel *channel,
 		unsigned int switch_timer_interval)
@@ -120,6 +122,8 @@ void consumer_timer_switch_start(struct lttng_consumer_channel *channel,
 
 /*
  * Stop and delete timer.
+ * Should be called only from the recv cmd thread (single thread ensures
+ * mutual exclusion).
  */
 void consumer_timer_switch_stop(struct lttng_consumer_channel *channel)
 {
diff --git a/src/common/consumer.h b/src/common/consumer.h
index 19a590e80..844310681 100644
--- a/src/common/consumer.h
+++ b/src/common/consumer.h
@@ -146,6 +146,8 @@ struct lttng_consumer_channel {
 	/* For metadata periodical flush */
 	int switch_timer_enabled;
 	timer_t switch_timer;
+	int switch_timer_error;
+
 	/* On-disk circular buffer */
 	uint64_t tracefile_size;
 	uint64_t tracefile_count;