src: fix race condition in ~NodeTraceBuffer

Libuv does not guarantee that handles have their close
callbacks called in the order in which they were added
(and in fact, currently calls them in reverse order).

This patch ensures that the `flush_signal_` handle
is no longer in use (i.e. its close callback has already
been run) when we signal to the main thread that
`~NodeTraceBuffer` may be destroyed.

Credit for debugging goes to Gireesh Punathil.

Fixes: #25512

src/tracing/node_trace_buffer.cc

@@ -1,4 +1,5 @@
 #include "tracing/node_trace_buffer.h"
+#include "util-inl.h"
 
 namespace node {
 namespace tracing {
@@ -170,15 +171,25 @@ void NodeTraceBuffer::NonBlockingFlushSignalCb(uv_async_t* signal) {
 
 // static
 void NodeTraceBuffer::ExitSignalCb(uv_async_t* signal) {
-  NodeTraceBuffer* buffer = reinterpret_cast<NodeTraceBuffer*>(signal->data);
-  uv_close(reinterpret_cast<uv_handle_t*>(&buffer->flush_signal_), nullptr);
-  uv_close(reinterpret_cast<uv_handle_t*>(&buffer->exit_signal_),
+  NodeTraceBuffer* buffer =
+      ContainerOf(&NodeTraceBuffer::exit_signal_, signal);
+
+  // Close both flush_signal_ and exit_signal_.
+  uv_close(reinterpret_cast<uv_handle_t*>(&buffer->flush_signal_),
            [](uv_handle_t* signal) {
+    NodeTraceBuffer* buffer =
+        ContainerOf(&NodeTraceBuffer::flush_signal_,
+                    reinterpret_cast<uv_async_t*>(signal));
+
+    uv_close(reinterpret_cast<uv_handle_t*>(&buffer->exit_signal_),
+             [](uv_handle_t* signal) {
       NodeTraceBuffer* buffer =
-          reinterpret_cast<NodeTraceBuffer*>(signal->data);
-      Mutex::ScopedLock scoped_lock(buffer->exit_mutex_);
-      buffer->exited_ = true;
-      buffer->exit_cond_.Signal(scoped_lock);
+          ContainerOf(&NodeTraceBuffer::exit_signal_,
+                      reinterpret_cast<uv_async_t*>(signal));
+        Mutex::ScopedLock scoped_lock(buffer->exit_mutex_);
+        buffer->exited_ = true;
+        buffer->exit_cond_.Signal(scoped_lock);
+    });
   });
 }