[PATCH] drm/i915: Suppress EIO during set-to-cpu-domain

[PATCH] drm/i915: Suppress EIO during set-to-cpu-domain

[Chris Wilson]

If the hardware is dead, we can simply discard any outstanding writes
from the GPU and presume the buffer is either in the GTT domain or
already in the CPU domain and continue on flushing the CPU caches.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_gem.c |    4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
index 19ca320..80fb67d 100644
--- a/drivers/gpu/drm/i915/i915_gem.c
+++ b/drivers/gpu/drm/i915/i915_gem.c
@@ -3087,12 +3087,12 @@ i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write)
 		return 0;
 
 	ret = i915_gem_object_flush_gpu_write_domain(obj);
-	if (ret)
+	if (ret && ret != -EIO)
 		return ret;
 
 	if (write || obj->pending_gpu_write) {
 		ret = i915_gem_object_wait_rendering(obj);
-		if (ret)
+		if (ret && ret != -EIO)
 			return ret;
 	}
 
-- 
1.7.10

Re: [PATCH] drm/i915: Suppress EIO during set-to-cpu-domain

[Daniel Vetter]

On Sat, Apr 14, 2012 at 10:24:21AM +0100, Chris Wilson wrote:
> If the hardware is dead, we can simply discard any outstanding writes
> from the GPU and presume the buffer is either in the GTT domain or
> already in the CPU domain and continue on flushing the CPU caches.
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Pardon my ignorance, but I fail to see the goal of these -EIO patches ...
Care to mind the dense?
-Daniel

> ---
>  drivers/gpu/drm/i915/i915_gem.c |    4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem.c b/drivers/gpu/drm/i915/i915_gem.c
> index 19ca320..80fb67d 100644
> --- a/drivers/gpu/drm/i915/i915_gem.c
> +++ b/drivers/gpu/drm/i915/i915_gem.c
> @@ -3087,12 +3087,12 @@ i915_gem_object_set_to_cpu_domain(struct drm_i915_gem_object *obj, bool write)
>  		return 0;
>  
>  	ret = i915_gem_object_flush_gpu_write_domain(obj);
> -	if (ret)
> +	if (ret && ret != -EIO)
>  		return ret;
>  
>  	if (write || obj->pending_gpu_write) {
>  		ret = i915_gem_object_wait_rendering(obj);
> -		if (ret)
> +		if (ret && ret != -EIO)
>  			return ret;
>  	}
>  
> -- 
> 1.7.10
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx

-- 
Daniel Vetter
Mail: daniel@ffwll.ch
Mobile: +41 (0)79 365 57 48

Re: [PATCH] drm/i915: Suppress EIO during set-to-cpu-domain

[Chris Wilson]

On Sun, 15 Apr 2012 00:16:15 +0200, Daniel Vetter <daniel@ffwll.ch> wrote:
> On Sat, Apr 14, 2012 at 10:24:21AM +0100, Chris Wilson wrote:
> > If the hardware is dead, we can simply discard any outstanding writes
> > from the GPU and presume the buffer is either in the GTT domain or
> > already in the CPU domain and continue on flushing the CPU caches.
> > 
> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> 
> Pardon my ignorance, but I fail to see the goal of these -EIO patches ...
> Care to mind the dense?

I encountered a page-fault-of-doom, due to i915_gem_fault() spinning on
a flush. i915_gem_reset() should fixup the buffers to prevent that.
However, I thought the EIO checks served as nice documentation along
paths that simply did not care why the GPU was no longer accessing the
buffer only that it wasn't.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre