Re: [PATCH 03/10] drm/i915: kill waiting_seqno

From: Chris Wilson <chris@chris-wilson.co.uk>
To: intel-gfx@lists.freedesktop.org
Cc: Ben Widawsky <ben@bwidawsk.net>,
	Ben Widawsky <benjamin.widawsky@intel.com>
Subject: Re: [PATCH 03/10] drm/i915: kill waiting_seqno
Date: Sun, 22 Apr 2012 14:46:05 +0100	[thread overview]
Message-ID: <1335102395_20078@CP5-2952> (raw)
In-Reply-To: <1334971412-4826-4-git-send-email-ben@bwidawsk.net>

On Fri, 20 Apr 2012 18:23:25 -0700, Ben Widawsky <ben@bwidawsk.net> wrote:
> diff --git a/drivers/gpu/drm/i915/i915_irq.c b/drivers/gpu/drm/i915/i915_irq.c
> index cf0e9f0..52eb9ed 100644
> --- a/drivers/gpu/drm/i915/i915_irq.c
> +++ b/drivers/gpu/drm/i915/i915_irq.c
> @@ -1846,11 +1846,9 @@ static bool i915_hangcheck_ring_idle(struct intel_ring_buffer *ring, bool *err)

So ring->waiting_seqno was also in effect checking that the ring was
initialised here. Inappropriately, I might add.

Anyway we need:

  if (ring->obj == NULL)
	return true;

>  	if (list_empty(&ring->request_list) ||
>  	    i915_seqno_passed(ring->get_seqno(ring), ring_last_seqno(ring))) {
>  		/* Issue a wake-up to catch stuck h/w. */
> -		if (ring->waiting_seqno && waitqueue_active(&ring->irq_queue)) {
> -			DRM_ERROR("Hangcheck timer elapsed... %s idle [waiting on %d, at %d], missed IRQ?\n",
> -				  ring->name,
> -				  ring->waiting_seqno,
> -				  ring->get_seqno(ring));
> +		if (waitqueue_active(&ring->irq_queue)) {
> +			DRM_ERROR("Hangcheck timer elapsed... %s idle\n",
> +				  ring->name);
>  			wake_up_all(&ring->irq_queue);
>  			*err = true;
>  		}

-- 
Chris Wilson, Intel Open Source Technology Centre