On Mon, 2019-02-04 at 23:44 +0100, David Kozub wrote:

On Mon, 4 Feb 2019, Christoph Hellwig wrote:

quoted

quoted

+	/* first do a discovery0 */
+	error = opal_discovery0_step(dev);

+	for (state = 0; !error && state < n_steps; state++)
+		error = execute_step(dev, &steps[state], state);
+
+	/*
+	 * For each OPAL command the first step in steps starts some sort of
+	 * session. If an error occurred in the initial discovery0 or if an
+	 * error occurred in the first step (and thus stopping the loop with
+	 * state == 1) then there was an error before or during the attempt to
+	 * start a session. Therefore we shouldn't attempt to terminate a
+	 * session, as one has not yet been created.
+	 */
+	if (error && state > 1)
+		end_opal_session_error(dev);

 	return error;

The flow here is a little too condensed for my taste.  Why not the
plain obvoious, if a little longer:

	error = error = opal_discovery0_step(dev);
	if (error)
		return error;

	for (state = 0; state < n_steps; state++) {
		error = execute_step(dev, &steps[state], state);
		if (error)
			goto out_error;
	}

	return 0;

out_error:
	if (state > 1)
		end_opal_session_error(dev);
	return error;

No problem, I can use this version. But I think there is a minor issue - 
the same one I hit in my original change, just from the other direction:

If the loop succeds for the 0-th element of steps, and then fails for the 
1st element, then state equals 1 yet the session has been started, so we 
should close it.

I think the condition in out_error should be if (state > 0).

Best regards,
David

Looks good with Christoph's suggestion (for 14/16) and your state check fix


Reviewed-by: Jon Derrick <redacted>