mgmt/brick-mux: Avoid sending two response when attach is failed.

We were sending two response back to glusterd when an attach is
failed. One from the handler function glusterfs_handle_attach and
another from rpcsvc_check_and_reply_error. It was causing problems
like ref leaks, transport disconnect etc.

Change-Id: I3bb5b59959530760b568d52becb519499b3dcd2b
updates: bz#1785143
Signed-off-by: Mohammed Rafi KC <rkavunga@redhat.com>

1 files changed, 9 insertions, 1 deletions

diff --git a/glusterfsd/src/glusterfsd-mgmt.c b/glusterfsd/src/glusterfsd-mgmt.c
index dc487e8afcc..2b97dc91165 100644
--- a/glusterfsd/src/glusterfsd-mgmt.c
+++ b/glusterfsd/src/glusterfsd-mgmt.c
@@ -992,7 +992,15 @@ glusterfs_handle_attach(rpcsvc_request_t *req)
         if (ret) {
             ret = -1;
         }
-        glusterfs_translator_info_response_send(req, ret, NULL, NULL);
+        ret = glusterfs_translator_info_response_send(req, ret, NULL, NULL);
+        if (ret) {
+            /* Response sent back to glusterd, req is already destroyed. So
+             * resetting the ret to 0. Otherwise another response will be
+             * send from rpcsvc_check_and_reply_error. Which will lead to
+             * double resource leak.
+             */
+            ret = 0;
+        }
     unlock:
         UNLOCK(&ctx->volfile_lock);
     }