arch/risc-v/src/mpfs/mpfs_ddr.c: Re-try DDR training on any error

Also errors which currently return other than -EAGAIN, are typically recoverable with retraining. So just re-try trainining 20 times on any error, resetting the controller in between. Only reset if it is not recovered in this time. Signed-off-by: Jukka Laitinen <[email protected]>
tiiuae · Jan 29, 2025 · 44558d5 · 44558d5
1 parent 4e8326d
commit 44558d5
Showing 1 changed file with 17 additions and 6 deletions.
diff --git a/arch/risc-v/src/mpfs/mpfs_ddr.c b/arch/risc-v/src/mpfs/mpfs_ddr.c
@@ -3941,19 +3941,30 @@ static int mpfs_ddr_setup(struct mpfs_ddr_priv_s *priv)
 int mpfs_ddr_init(void)
 {
   struct mpfs_ddr_priv_s *priv = &g_mpfs_ddr_priv;
-  int ddr_status;
-
-  /* On -EAGAIN, the whole training is restarted from the very beginning */
+  int ddr_status = -EAGAIN;
+  int retry_count = 20;
+
+  /* Restart training until it passes or retry_count reaches 0.
+   * Errors which return -EAGAIN are known to be always
+   * recoverable by controller reset; don't count these errors.
+   * Only count errors which return -EIO; they are typically recoverable.
+   * If they persist, however, eventually return failure, leading to
+   * re-trying after full reset.
+   */
 
-  do
+  while (ddr_status != OK && retry_count > 0)
     {
       ddr_status = mpfs_ddr_setup(priv);
-      if (ddr_status == -EAGAIN)
+      if (ddr_status != OK)
         {
           mpfs_ddr_fail(priv);
         }
+
+      if (ddr_status != -EAGAIN)
+        {
+          retry_count--;
+        }
     }
-  while (ddr_status == -EAGAIN);
 
   if (ddr_status == 0)
     {