diff --git a/src/utils.c b/src/utils.c
index c086633b..da863813 100644
--- a/src/utils.c
+++ b/src/utils.c
@@ -26,6 +26,19 @@
 
 #define MAX_PATH_LEN 1024
 
+/* Calculate "x * n / d" without unnecessary overflow or loss of precision.
+ *
+ * Reference:
+ * https://elixir.bootlin.com/linux/v6.10.7/source/include/linux/math.h#L121
+ */
+static inline uint64_t mult_frac(uint64_t x, uint64_t n, uint64_t d)
+{
+    const uint64_t q = x / d;
+    const uint64_t r = x % d;
+
+    return q * n + r * n / d;
+}
+
 static void get_time_info(int32_t *tv_sec, int32_t *tv_nsec)
 {
 #if defined(HAVE_POSIX_TIMER)
@@ -40,14 +53,13 @@ static void get_time_info(int32_t *tv_sec, int32_t *tv_nsec)
      */
     if (info.denom == 0)
         (void) mach_timebase_info(&info);
-    /* Hope that the multiplication doesn't overflow. */
-    uint64_t nsecs = mach_absolute_time() * info.numer / info.denom;
+    uint64_t nsecs = mult_frac(mach_absolute_time(), info.numer, info.denom);
     *tv_sec = nsecs / 1e9;
     *tv_nsec = nsecs - (*tv_sec * 1e9);
 #else /* low resolution timer */
     clock_t t = clock();
     *tv_sec = t / CLOCKS_PER_SEC;
-    *tv_nsec = (t % CLOCKS_PER_SEC) * (1e9 / CLOCKS_PER_SEC);
+    *tv_nsec = mult_frac(t % CLOCKS_PER_SEC, 1e9, CLOCKS_PER_SEC);
 #endif
 }