Allow get_next_notification to return when interrupted

Applications (and scripts) which consume a given set of notifications
indefinitely may fail to exit if a SIGTERM handler is registered.

lttng_notification_channel_get_next_notification() blocks indefinitely
on recvmsg() until a new notification is available. The wrapper that
is used to do so automatically restarts the recvmsg() if it is
interrupted, thus not allowing clients a change to cleanly exit.

This change causes the notification channel to wait for a message to
be available using select() before starting the actual reception of
the data and return LTTNG_NOTIFICATION_CHANNEL_STATUS_INTERRUTPED if a
signal occurs during the wait.

If data is available, it is assumed that the message is well-formed
and can promptly be received in its entirety. The goal of this change
is to give a monitoring application a chance to leave the
get_next_notification() function and check if it should exit.

Signed-off-by: Jérémie Galarneau <jeremie.galarneau@efficios.com>

diff --git a/include/lttng/notification/channel.h b/include/lttng/notification/channel.h
index 7708cfd56e915a293007d481f0f36c1dd8980454..878d37ee1a3f742bfa5fb5c5ac67ade036f26e19 100644 (file)
--- a/include/lttng/notification/channel.h
+++ b/include/lttng/notification/channel.h
@@ -31,6 +31,7 @@ struct lttng_notification_channel;
 
 enum lttng_notification_channel_status {
        LTTNG_NOTIFICATION_CHANNEL_STATUS_NOTIFICATIONS_DROPPED = 1,
+       LTTNG_NOTIFICATION_CHANNEL_STATUS_INTERRUPTED = 2,
        LTTNG_NOTIFICATION_CHANNEL_STATUS_OK = 0,
        LTTNG_NOTIFICATION_CHANNEL_STATUS_ERROR = -1,
        LTTNG_NOTIFICATION_CHANNEL_STATUS_CLOSED = -2,
@@ -81,10 +82,14 @@ extern struct lttng_notification_channel *lttng_notification_channel_create(
  * Notifications can be dropped if a client consumes the notifications sent
  * through the notification channel too slowly.
  *
- * Returns LTTNG_NOTIFICATION_CHANNEL_STATUS_OK and a notification on success,
- * LTTNG_NOTIFICATION_CHANNEL_STATUS_INVALID if an invalid parameter was
- * provided, or LTTNG_NOTIFICATION_CHANNEL_STATUS_NOTIFICATIONS_DROPPED if
- * notifications were dropped.
+ * Returns
+ *   - LTTNG_NOTIFICATION_CHANNEL_STATUS_OK and a notification on success,
+ *   - LTTNG_NOTIFICATION_CHANNEL_STATUS_INVALID if an invalid parameter was
+ *     provided,
+ *   - LTTNG_NOTIFICATION_CHANNEL_STATUS_NOTIFICATIONS_DROPPED if notifications
+ *     were dropped,
+ *   - LTTNG_NOTIFICATION_CHANNEL_STATUS_INTERRUPTED if a signal was received
+ *     that caused the reception to fail.
  */
 extern enum lttng_notification_channel_status
 lttng_notification_channel_get_next_notification(
@@ -102,9 +107,10 @@ lttng_notification_channel_get_next_notification(
  * lttng_notification_channel_get_next_notification() can be called and
  * is guaranteed to not block.
  *
- * Returns LTTNG_NOTIFICATION_CHANNEL_STATUS_OK on success or
- * LTTNG_NOTIFICATION_CHANNEL_STATUS_INVALID if an invalid parameter was
- * provided.
+ * Returns
+ *   - LTTNG_NOTIFICATION_CHANNEL_STATUS_OK on success,
+ *   - LTTNG_NOTIFICATION_CHANNEL_STATUS_INVALID if an invalid parameter was
+ *     provided.
  */
 extern enum lttng_notification_channel_status
 lttng_notification_channel_has_pending_notification(
@@ -120,10 +126,11 @@ lttng_notification_channel_has_pending_notification(
  * An error will be reported if the client tries to subscribe to the same
  * condition multiple times without unsubscribing.
  *
- * Returns LTTNG_NOTIFICATION_CHANNEL_STATUS_OK on success,
- * LTTNG_NOTIFICATION_CHANNEL_STATUS_INVALID if an invalid parameter was
- * provided, or LTTNG_NOTIFICATION_CHANNEL_STATUS_ALREADY_SUBSCRIBED if the
- * client was already subscribed to the condition through this channel.
+ * Returns
+ *   - LTTNG_NOTIFICATION_CHANNEL_STATUS_OK on success,
+ *   - LTTNG_NOTIFICATION_CHANNEL_STATUS_INVALID if an invalid parameter was
+ *     provided, or LTTNG_NOTIFICATION_CHANNEL_STATUS_ALREADY_SUBSCRIBED if the
+ *     client was already subscribed to the condition through this channel.
  */
 extern enum lttng_notification_channel_status
 lttng_notification_channel_subscribe(
@@ -139,10 +146,11 @@ lttng_notification_channel_subscribe(
  * An error will be reported if the client tries to unsubscribe to from a
  * conditions' notifications to which it was not previously subscribed.
  *
- * Returns LTTNG_NOTIFICATION_CHANNEL_STATUS_OK on success,
- * LTTNG_NOTIFICATION_CHANNEL_STATUS_INVALID if an invalid parameter was
- * provided, or LTTNG_NOTIFICATION_CHANNEL_STATUS_UNKNOWN_CONDITION if the
- * client was not already subscribed to the condition through this channel.
+ * Returns
+ *   - LTTNG_NOTIFICATION_CHANNEL_STATUS_OK on success,
+ *   - LTTNG_NOTIFICATION_CHANNEL_STATUS_INVALID if an invalid parameter was
+ *     provided, or LTTNG_NOTIFICATION_CHANNEL_STATUS_UNKNOWN_CONDITION if the
+ *     client was not already subscribed to the condition through this channel.
  */
 extern enum lttng_notification_channel_status
 lttng_notification_channel_unsubscribe(

diff --git a/src/lib/lttng-ctl/channel.c b/src/lib/lttng-ctl/channel.c
index 5214fe73a41eeae6d96985e16fece47d45f1d7f8..5487d051b5013fd60b080491ce865966bc6ffeab 100644 (file)
--- a/src/lib/lttng-ctl/channel.c
+++ b/src/lib/lttng-ctl/channel.c
@@ -211,6 +211,7 @@ lttng_notification_channel_get_next_notification(
        struct lttng_notification *notification = NULL;
        enum lttng_notification_channel_status status =
                        LTTNG_NOTIFICATION_CHANNEL_STATUS_OK;
+       fd_set read_fds;
 
        if (!channel || !_notification) {
                status = LTTNG_NOTIFICATION_CHANNEL_STATUS_INVALID;
@@ -239,6 +240,28 @@ lttng_notification_channel_get_next_notification(
                goto end_unlock;
        }
 
+       /*
+        * Block on select() instead of the message reception itself as the
+        * recvmsg() wrappers always restard on EINTR. We choose to wait
+        * using select() in order to:
+        *   1) Return if a signal occurs,
+        *   2) Not deal with partially received messages.
+        *
+        * The drawback to this approach is that we assume that messages
+        * are complete/well formed. If a message is shorter than its
+        * announced length, receive_message() will block on recvmsg()
+        * and never return (even if a signal is received).
+        */
+       FD_ZERO(&read_fds);
+       FD_SET(channel->socket, &read_fds);
+       ret = select(channel->socket + 1, &read_fds, NULL, NULL, NULL);
+       if (ret == -1) {
+               status = errno == EINTR ?
+                       LTTNG_NOTIFICATION_CHANNEL_STATUS_INTERRUPTED :
+                       LTTNG_NOTIFICATION_CHANNEL_STATUS_ERROR;
+               goto end_unlock;
+       }
+
        ret = receive_message(channel);
        if (ret) {
                status = LTTNG_NOTIFICATION_CHANNEL_STATUS_ERROR;