In some instances it's possible for cpg_dispatch() to return
CS_ERR_LIBRARY if cpg_finalise() is called from the callback routine.
The error isn't a problem but it can cause calling applications to log
an error where there is no actual problem (this has been seen in cmirrord)
The trouble is caused because control_buffer->ipc_closed is (correctly)
set in this situation which causes ipc_sem_wait() to return CS_ERR_LIBRARY.
My solution is to check for control_buffer->ipc_closed in coroipcc_put()
and instead of simply passing CS_ERR_LIBRARY up the stack, change it to
a benign CS_OK as it is quite reasonable behaviour when shutting down.
Signed-Off-By: Christine Caulfield <ccaulfie@xxxxxxxxxx>
diff --git a/lib/coroipcc.c b/lib/coroipcc.c
index 774e835..5aff6e0 100644
--- a/lib/coroipcc.c
+++ b/lib/coroipcc.c
@@ -957,6 +957,13 @@ coroipcc_dispatch_put (hdb_handle_t handle)
retry_ipc_sem_wait:
res = ipc_sem_wait (ipc_instance->control_buffer, SEMAPHORE_DISPATCH, ipc_instance->fd);
if (res != CS_OK) {
+ /*
+ * IPC was closed earlier in dispatch, return OK
+ */
+ if (ipc_instance->control_buffer->ipc_closed) {
+ res = CS_OK;
+ goto error_exit;
+ }
if (res == CS_ERR_TRY_AGAIN) {
priv_change_send (ipc_instance);
goto retry_ipc_sem_wait;
_______________________________________________
discuss mailing list
discuss@xxxxxxxxxxxx
http://lists.corosync.org/mailman/listinfo/discuss
