[lttng-dev] [RFC PATCH lttng-tools] Fix: metadata push -EPIPE should be recoverable

[Jérémie Galarneau]

Merged in master, stable-2.6 and stable-2.5. Thanks!

Jérémie

On Mon, Jul 6, 2015 at 12:21 PM, Mathieu Desnoyers <
mathieu.desnoyers at efficios.com> wrote:

> This return value can be caused by application terminating concurrently
> (when using per-PID buffers), so it should not make the consumer
> management thread exit.
>
> CC: Aravind HT <aravind.ht at gmail.com>
> Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers at efficios.com>
> ---
>  src/bin/lttng-sessiond/ust-app.c      | 13 ++++++++++---
>  src/bin/lttng-sessiond/ust-consumer.c |  7 +++++--
>  2 files changed, 15 insertions(+), 5 deletions(-)
>
> diff --git a/src/bin/lttng-sessiond/ust-app.c
> b/src/bin/lttng-sessiond/ust-app.c
> index c30792c..6f032da 100644
> --- a/src/bin/lttng-sessiond/ust-app.c
> +++ b/src/bin/lttng-sessiond/ust-app.c
> @@ -431,6 +431,9 @@ void delete_ust_app_channel(int sock, struct
> ust_app_channel *ua_chan,
>   * Must be called with the registry lock held.
>   *
>   * On success, return the len of metadata pushed or else a negative value.
> + * Returning a -EPIPE return value means we could not send the metadata,
> + * but it can be caused by recoverable errors (e.g. the application has
> + * terminated concurrently).
>   */
>  ssize_t ust_app_push_metadata(struct ust_registry_session *registry,
>                 struct consumer_socket *socket, int send_zero_data)
> @@ -454,9 +457,10 @@ ssize_t ust_app_push_metadata(struct
> ust_registry_session *registry,
>         /*
>          * On a push metadata error either the consumer is dead or the
>          * metadata channel has been destroyed because its endpoint
> -        * might have died (e.g: relayd). If so, the metadata closed
> -        * flag is set to 1 so we deny pushing metadata again which is
> -        * not valid anymore on the consumer side.
> +        * might have died (e.g: relayd), or because the application has
> +        * exited. If so, the metadata closed flag is set to 1 so we
> +        * deny pushing metadata again which is not valid anymore on the
> +        * consumer side.
>          */
>         if (registry->metadata_closed) {
>                 return -EPIPE;
> @@ -547,6 +551,9 @@ error_push:
>   * of socket throughout this function.
>   *
>   * Return 0 on success else a negative error.
> + * Returning a -EPIPE return value means we could not send the metadata,
> + * but it can be caused by recoverable errors (e.g. the application has
> + * terminated concurrently).
>   */
>  static int push_metadata(struct ust_registry_session *registry,
>                 struct consumer_output *consumer)
> diff --git a/src/bin/lttng-sessiond/ust-consumer.c
> b/src/bin/lttng-sessiond/ust-consumer.c
> index 78e50df..ad076e3 100644
> --- a/src/bin/lttng-sessiond/ust-consumer.c
> +++ b/src/bin/lttng-sessiond/ust-consumer.c
> @@ -511,12 +511,15 @@ int ust_consumer_metadata_request(struct
> consumer_socket *socket)
>         pthread_mutex_lock(&ust_reg->lock);
>         ret_push = ust_app_push_metadata(ust_reg, socket, 1);
>         pthread_mutex_unlock(&ust_reg->lock);
> -       if (ret_push < 0) {
> +       if (ret_push == -EPIPE) {
> +               DBG("Application or relay closed while pushing metadata");
> +       } else if (ret_push < 0) {
>                 ERR("Pushing metadata");
>                 ret = -1;
>                 goto end;
> +       } else {
> +               DBG("UST Consumer metadata pushed successfully");
>         }
> -       DBG("UST Consumer metadata pushed successfully");
>         ret = 0;
>
>  end:
> --
> 2.1.4
>
>


-- 
Jérémie Galarneau
EfficiOS Inc.
http://www.efficios.com
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lttng.org/pipermail/lttng-dev/attachments/20150710/5b181ad6/attachment.html>