diff --git a/cpp/include/tensorrt_llm/common/stringUtils.h b/cpp/include/tensorrt_llm/common/stringUtils.h
index e806c714527..a4803cba372 100644
--- a/cpp/include/tensorrt_llm/common/stringUtils.h
+++ b/cpp/include/tensorrt_llm/common/stringUtils.h
@@ -74,6 +74,28 @@ void printElement(std::ostream& os, std::tuple<Args...> const& t)
     printTupleImpl(os, t, std::index_sequence_for<Args...>{});
 }
 
+class va_list_guard
+{
+public:
+    explicit va_list_guard(va_list& args)
+        : mArgs(args)
+    {
+    }
+
+    ~va_list_guard()
+    {
+        va_end(mArgs);
+    }
+
+    va_list_guard(va_list_guard const&) = delete;
+    va_list_guard& operator=(va_list_guard const&) = delete;
+    va_list_guard(va_list_guard&&) = delete;
+    va_list_guard& operator=(va_list_guard&&) = delete;
+
+private:
+    va_list& mArgs;
+};
+
 } // namespace
 
 // Override operator<< for any tuple
@@ -117,6 +139,8 @@ inline std::string fmtstr(char const* format, ...)
 
     va_list args;
     va_start(args, format);
+    va_list_guard args_guard(args);
+
     fmtstr_(
         format,
         [](void* target, size_t count) -> char*
@@ -131,7 +155,6 @@ inline std::string fmtstr(char const* format, ...)
             return str->data();
         },
         &result, args);
-    va_end(args);
 
     return result;
 }
diff --git a/cpp/tensorrt_llm/common/stringUtils.cpp b/cpp/tensorrt_llm/common/stringUtils.cpp
index 75052ad4fad..283dec8842c 100644
--- a/cpp/tensorrt_llm/common/stringUtils.cpp
+++ b/cpp/tensorrt_llm/common/stringUtils.cpp
@@ -34,6 +34,8 @@ void fmtstr_(char const* format, fmtstr_allocator alloc, void* target, va_list a
     size_t constexpr init_size = 2048;
     char fixed_buffer[init_size];
     auto const size = std::vsnprintf(fixed_buffer, init_size, format, args0);
+    va_end(args0);
+
     TLLM_CHECK_WITH_INFO(size >= 0, std::string(std::strerror(errno)));
     if (size == 0)
     {