[ofw][patch][MLX4] improve debug in error flow of command interface

[Leonid Keller]

Index: V:/svn/winib/trunk/hw/mlx4/kernel/bus/net/cmd.c
===================================================================
--- V:/svn/winib/trunk/hw/mlx4/kernel/bus/net/cmd.c (revision 5038)
+++ V:/svn/winib/trunk/hw/mlx4/kernel/bus/net/cmd.c (revision 5039)
@@ -102,6 +102,7 @@
  int   next;
  u64   out_param;
  u16   token;
+ u8   status;
 };
 
 static int mlx4_status_to_errno(u8 status) {
@@ -137,13 +138,13 @@
  return trans_table[status];
 }
 
-static int cmd_pending(struct mlx4_dev *dev)
+static int cmd_pending(struct mlx4_dev *dev, u32 *p_status)
 {
- u32 status = readl(mlx4_priv(dev)->cmd.hcr + HCR_STATUS_OFFSET);
+ *p_status = readl(mlx4_priv(dev)->cmd.hcr + HCR_STATUS_OFFSET);
 
- return (status & swab32(1 << HCR_GO_BIT)) ||
+ return (*p_status & swab32(1 << HCR_GO_BIT)) ||
   (mlx4_priv(dev)->cmd.toggle ==
-   !!(status & swab32(1 << HCR_T_BIT)));
+   !!(*p_status & swab32(1 << HCR_T_BIT)));
 }
 
 static int mlx4_cmd_post(struct mlx4_dev *dev, u64 in_param, u64
out_param,
@@ -154,6 +155,7 @@
  u32 __iomem *hcr = (u32 __iomem *)cmd->hcr;
  int ret = -EAGAIN;
  u64 end;
+ u32 hcr_status;
 
  mutex_lock(&cmd->hcr_mutex);
 
@@ -161,9 +163,12 @@
  if (event)
   end += msecs_to_jiffies(GO_BIT_TIMEOUT_MSECS);
 
- while (cmd_pending(dev)) {
-  if (time_after_eq(jiffies, end))
+ while (cmd_pending(dev, &hcr_status)) {
+  if (time_after_eq(jiffies, end)) {
+   mlx4_err(dev, "Failed to post command %02x during %d msecs,
hcr_status %#x, toggle %#x\n",
+    op, GO_BIT_TIMEOUT_MSECS, hcr_status, mlx4_priv(dev)->cmd.toggle );
    goto out;
+  }
   cond_resched();
  }
 
@@ -212,8 +217,9 @@
  u8 __iomem *hcr = priv->cmd.hcr;
  int err = 0;
  u64 end;
- u8 status;
+ u8 status = 0xff; /* means "unknown" */
  long do_reset;
+ u32 hcr_status;
 
  down(&priv->cmd.poll_sem);
 
@@ -223,13 +229,13 @@
   goto out;
 
  end = msecs_to_jiffies(timeout) + jiffies;
- while (cmd_pending(dev) && time_before(jiffies, end))
+ while (cmd_pending(dev, &hcr_status) && time_before(jiffies, end))
   cond_resched();
 
- if (cmd_pending(dev)) {
+ if (cmd_pending(dev, &hcr_status)) {
   err = -ETIMEDOUT;
-  mlx4_err(dev, "mlx4_cmd_poll: Command %02x completed with timeout
after %d msecs \n",
-     op, timeout);
+  mlx4_err(dev, "mlx4_cmd_poll: Command %02x: timeout after %d msecs,
hcr_status %#x, toggle %#x \n",
+     op, timeout, hcr_status, mlx4_priv(dev)->cmd.toggle);
 
   do_reset = InterlockedCompareExchange(&dev->reset_pending, 1, 0);
   if (!do_reset) {
@@ -258,11 +264,11 @@
 
  status = (u8)(be32_to_cpu((__force __be32)__raw_readl(hcr +
HCR_STATUS_OFFSET)) >> 24);
  err = mlx4_status_to_errno(status);
+
+out:
  if (status && status != 0x50)
-  mlx4_err(dev, "Command failed: op %#hx, status %#02x, errno %d.\n", 
+  mlx4_err(dev, "mlx4_cmd_poll: Command failed: op %#hx, status %#02x,
errno %d.\n", 
    op, status, err);
-
-out:
  up(&priv->cmd.poll_sem);
  return err;
 }
@@ -274,14 +280,14 @@
   &priv->cmd.context[token & priv->cmd.token_mask];
 
  /* previously timed out command completing at long last */
- if (token != context->token)
+ if (token != context->token) {
+  mlx4_err(dev, "mlx4_cmd_event: Command skipped: token %#hx, ctx_token
%#hx\n", 
+   token, context->token);
   return;
+ }
 
  context->result    = mlx4_status_to_errno(status);
- 
- if (status)
-  mlx4_err(dev, "Command failed: token %#hx, status %#02x, errno
%d.\n", 
-   token, status, context->result);
+ context->status    = status;
  context->out_param = out_param;
 
  complete(&context->done);
@@ -296,6 +302,7 @@
  int err = 0;
  u64 out_prm = out_param ? *out_param : 0;
  long do_reset;
+ u8 status = 0xff; /* means "unknown" */
 
  down(&cmd->event_sem);
  if ( dev->flags & MLX4_FLAG_RESET_DRIVER ) {
@@ -312,8 +319,10 @@
 
  init_completion(&context->done);
 
- mlx4_cmd_post(dev, in_param, out_prm,
+ err = mlx4_cmd_post(dev, in_param, out_prm,
   in_modifier, op_modifier, op, context->token, 1);
+ if (err)
+  goto out;
 
  if (wait_for_completion_timeout(&context->done,
msecs_to_jiffies(timeout))) {
   if (!context->done.done) {
@@ -348,12 +357,14 @@
   }
   else {
    err = -EFAULT;
-   mlx4_err(dev, "mlx4_cmd_wait: Unexpected end of waiting for a comand
\n");
+   mlx4_err(dev, "mlx4_cmd_wait: Unexpected end of waiting for a
command \n");
    ASSERT(0);
   }
  }
- else
+ else {
   err = context->result;
+  status = context->status;
+ }
  
  if (err)
   goto out;
@@ -366,6 +377,9 @@
  context->next = cmd->free_head;
  cmd->free_head = (int)(context - cmd->context);
  spin_unlock(&cmd->context_lock);
+ if (status && status != 0x50)
+  mlx4_err(dev, "mlx4_cmd_wait: Command failed: op %#hx, status %#02x,
errno %d, token %#hx.\n", 
+   op, status, err, context->token);
 
 exit: 
  up(&cmd->event_sem);

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/ofw/attachments/20091116/ba0e7d7b/attachment.html>