rfnoc: Fix potential double-overrun handling in rx stream terminator

It's feasible that error messages are sent out during overrun handling,
which would cause a nested invocation of handle_overrun(). This adds a
lock to prevent that.

1 files changed, 6 insertions, 0 deletions

diff --git a/host/lib/rfnoc/rx_stream_terminator.cpp b/host/lib/rfnoc/rx_stream_terminator.cpp
index e05d9cd49..7d38e2101 100644
--- a/host/lib/rfnoc/rx_stream_terminator.cpp
+++ b/host/lib/rfnoc/rx_stream_terminator.cpp
@@ -63,6 +63,12 @@ void rx_stream_terminator::set_rx_streamer(bool active, const size_t)
 
 void rx_stream_terminator::handle_overrun(boost::weak_ptr<uhd::rx_streamer> streamer, const size_t)
 {
+    std::unique_lock<std::mutex> l(_overrun_handler_mutex, std::defer_lock);
+    if (!l.try_lock()) {
+        // We're already handling overruns, so just stop right there
+        return;
+    }
+
     std::vector<boost::shared_ptr<uhd::rfnoc::radio_ctrl_impl> > upstream_radio_nodes =
         find_upstream_node<uhd::rfnoc::radio_ctrl_impl>();
     const size_t n_radios = upstream_radio_nodes.size();