[Intel-gfx] [PATCH 3/6] drm/i915: Use request retirement as context progress

[Chris Wilson]

On Tue, Nov 15, 2016 at 04:36:33PM +0200, Mika Kuoppala wrote:
> As hangcheck score was removed, the active decay of score
> was removed also. This removed feature for hangcheck to detect
> if the gpu client was accidentally or maliciously causing intermittent
> hangs. Reinstate the scoring as a per context property, so that if
> one context starts to act unfavourably, ban it.
> 
> v2: ban_period_secs as a gate to score check (Chris)
> 
> Cc: Chris Wilson <chris at chris-wilson.co.uk>
> Signed-off-by: Mika Kuoppala <mika.kuoppala at intel.com>

> -	elapsed = get_seconds() - ctx->hang_stats.guilty_ts;
> -	if (ctx->hang_stats.ban_period_seconds &&
> -	    elapsed <= ctx->hang_stats.ban_period_seconds) {
> +	if (!hs->ban_period_seconds)
> +		return false;
> +
> +	elapsed = get_seconds() - hs->guilty_ts;
> +	if (elapsed <= hs->ban_period_seconds) {
>  		DRM_DEBUG("context hanging too fast, banning!\n");
>  		return true;
>  	}
>  
> +	if (hs->ban_score >= 40) {
> +		DRM_DEBUG("context hanging too often, banning!\n");
> +		return true;
> +	}
> +
>  	return false;
>  }

> +	hs->ban_score += 10;

This pair should be tunables (i.e. a macro somewhere sensible).

> diff --git a/drivers/gpu/drm/i915/i915_gem_request.c b/drivers/gpu/drm/i915/i915_gem_request.c
> index b9b5253..095c809 100644
> --- a/drivers/gpu/drm/i915/i915_gem_request.c
> +++ b/drivers/gpu/drm/i915/i915_gem_request.c
> @@ -204,6 +204,10 @@ static void i915_gem_request_retire(struct drm_i915_gem_request *request)
>  
>  	trace_i915_gem_request_retire(request);
>  
> +	/* Retirement decays the ban score as it is a sign of ctx progress */
> +	if (request->ctx->hang_stats.ban_score > 0)
> +		request->ctx->hang_stats.ban_score--;

Please put this along with the other request->ctx updates (i.e. after
request->previos_context and before the context_put).

Otherwise lgtm.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre