[Intel-gfx] [PATCH 55/59] drm/i915: Remove fallback poll for ring buffer space

[Daniel Vetter]

On Thu, Mar 19, 2015 at 04:33:12PM +0000, John Harrison wrote:
> On 19/03/2015 15:16, Jani Nikula wrote:
> >On Thu, 19 Mar 2015, "Daniel, Thomas" <thomas.daniel at intel.com> wrote:
> >>>-	if (&request->list == &ring->request_list)
> >>>+	/* It should always be possible to find a suitable request! */
> >>>+	if (&request->list == &ring->request_list) {
> >>>+		WARN_ON(true);
> >>>  		return -ENOSPC;
> >>>+	}
> >>Don’t we normally say
> >>	if (WARN_ON(&request->list == &ring->request_list))
> >>		return -ENOSPC;
> >Yes, particularly since we've amended WARN_ON within i915 to print out
> >the condition that failed. "true" is not very useful with that. ;)
> >
> >BR,
> >Jani.
> >
> The issue I have with 'if(WARN_ON(x))' is that it looks like something that
> would disappear in a non debugging build. Whereas, this is a check that
> wants to exist regardless of build options.

Yeah kernel isn't like that, WARN_ON is always executed. We rely on that
all over the place actually by sometimes wrapping full function-calls with
side-effects with a WARN_ON for cases we don't expect anything to ever
fail.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch