[Intel-gfx] [PATCH v2] drm/i915: Mark uneven memory banks on gen4 desktop as unknown swizzling

[Daniel Vetter]

On Thu, Nov 19, 2015 at 09:58:05AM +0000, Chris Wilson wrote:
> We have varied reports of swizzling corruption on gen4 desktop, and
> confirmation that one at least is triggered by uneven memory banks
> (L-shaped memory). The implication is that the swizzling varies between
> the paired channels and the remainder of memory on the single channel. As
> the object then has unpredictable swizzling (it will vary depending on
> exact page allocation and may even change during the object's lifetime as
> the pages are replaced), we have to report to userspace that the swizzling
> is unknown.
> 
> However, some existing userspace is buggy when it meets an unknown
> swizzling configuration and so we need to tell another white lie and
> mark the swizzling as NONE but report it as UNKNOWN through the extended
> get-tiling-ioctl. See
> 
> commit 5eb3e5a5e11d14f9deb2a4b83555443b69ab9940
> Author: Chris Wilson <chris at chris-wilson.co.uk>
> Date:   Sun Jun 28 09:19:26 2015 +0100
> 
>     drm/i915: Declare the swizzling unknown for L-shaped configurations
> 
> for the previous example where we found that telling the truth to
> userspace just ends up in a world of hurt.
> 
> Also since we don't truly know what the swizzling is on the pages, we
> need to keep them pinned to prevent swapping as the reports also
> suggest that some gen4 devices have previously undetected bit17
> swizzling.
> 
> v2: Combine unknown + quirk patches to prevent userspace ever seeing
> unknown swizzling through the normal get-tiling-ioctl. Also use the same
> path for the existing uneven bank detection for mobile gen4.
> 
> Reported-by: Matti Hämäläinen <ccr at tnsp.org>
> Tested--by: Matti Hämäläinen <ccr at tnsp.org>
> References: https://bugs.freedesktop.org/show_bug.cgi?id=90725
> Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
> Cc: Matti Hämäläinen <ccr at tnsp.org>
> Cc: Daniel Vetter <daniel.vetter at ffwll.ch>
> Cc: Jani Nikula <jani.nikula at intel.com>
> Cc: stable at vger.kernel.org

Reviewed-by: Daniel Vetter <daniel.vetter at ffwll.ch>

> ---
>  drivers/gpu/drm/i915/i915_gem_fence.c | 36 ++++++++++++++++++++++++++---------
>  1 file changed, 27 insertions(+), 9 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem_fence.c b/drivers/gpu/drm/i915/i915_gem_fence.c
> index 40a10b25956c..f010391b87f5 100644
> --- a/drivers/gpu/drm/i915/i915_gem_fence.c
> +++ b/drivers/gpu/drm/i915/i915_gem_fence.c
> @@ -642,11 +642,10 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev)
>  		}
>  
>  		/* check for L-shaped memory aka modified enhanced addressing */
> -		if (IS_GEN4(dev)) {
> -			uint32_t ddc2 = I915_READ(DCC2);
> -
> -			if (!(ddc2 & DCC2_MODIFIED_ENHANCED_DISABLE))
> -				dev_priv->quirks |= QUIRK_PIN_SWIZZLED_PAGES;
> +		if (IS_GEN4(dev) &&
> +		    !(I915_READ(DCC2) & DCC2_MODIFIED_ENHANCED_DISABLE)) {
> +			swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN;
> +			swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN;
>  		}
>  
>  		if (dcc == 0xffffffff) {
> @@ -675,16 +674,35 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev)
>  		 * matching, which was the case for the swizzling required in
>  		 * the table above, or from the 1-ch value being less than
>  		 * the minimum size of a rank.
> +		 *
> +		 * Reports indicate that the swizzling actually
> +		 * varies depending upon page placement inside the
> +		 * channels, i.e. we see swizzled pages where the
> +		 * banks of memory are paired and unswizzled on the
> +		 * uneven portion, so leave that as unknown.
>  		 */
> -		if (I915_READ16(C0DRB3) != I915_READ16(C1DRB3)) {
> -			swizzle_x = I915_BIT_6_SWIZZLE_NONE;
> -			swizzle_y = I915_BIT_6_SWIZZLE_NONE;
> -		} else {
> +		if (I915_READ16(C0DRB3) == I915_READ16(C1DRB3)) {
>  			swizzle_x = I915_BIT_6_SWIZZLE_9_10;
>  			swizzle_y = I915_BIT_6_SWIZZLE_9;
>  		}
>  	}
>  
> +	if (swizzle_x == I915_BIT_6_SWIZZLE_UNKNOWN ||
> +	    swizzle_y == I915_BIT_6_SWIZZLE_UNKNOWN) {
> +		/* Userspace likes to explode if it sees unknown swizzling,
> +		 * so lie. We will finish the lie when reporting through
> +		 * the get-tiling-ioctl by reporting the physical swizzle
> +		 * mode as unknown instead.
> +		 *
> +		 * As we don't strictly know what the swizzling is, it may be
> +		 * bit17 dependent, and so we need to also prevent the pages
> +		 * from being moved.
> +		 */
> +		dev_priv->quirks |= QUIRK_PIN_SWIZZLED_PAGES;
> +		swizzle_x = I915_BIT_6_SWIZZLE_NONE;
> +		swizzle_y = I915_BIT_6_SWIZZLE_NONE;
> +	}
> +
>  	dev_priv->mm.bit_6_swizzle_x = swizzle_x;
>  	dev_priv->mm.bit_6_swizzle_y = swizzle_y;
>  }
> -- 
> 2.6.2
> 

-- 
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch