diff mbox

[1/2] drm/i915: Mark uneven memory banks on gen4 desktop as unknown swizzling

Message ID 1447756852-15183-1-git-send-email-chris@chris-wilson.co.uk
State New, archived
Headers show

Commit Message

Chris Wilson Nov. 17, 2015, 10:40 a.m. UTC
We have varied reports of swizzling corruption on gen4 desktop, and
confirmation that it is triggered by uneven memory banks. The
implication is that the swizzling various 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.

Reported-by: Matti Hämäläinen <ccr@tnsp.org>
References: https://bugs.freedesktop.org/show_bug.cgi?id=90725
Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
Cc: Matti Hämäläinen <ccr@tnsp.org>
Cc: stable@vger.kernel.org
---
 drivers/gpu/drm/i915/i915_gem_fence.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

Comments

Chris Wilson Nov. 17, 2015, 10:51 a.m. UTC | #1
On Tue, Nov 17, 2015 at 10:40:51AM +0000, Chris Wilson wrote:
> We have varied reports of swizzling corruption on gen4 desktop, and
> confirmation that it is triggered by uneven memory banks. The
s/it/one at least/

> implication is that the swizzling various between the paired channels
s/various/varies/

> 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.
Chris Wilson Nov. 18, 2015, 10:05 a.m. UTC | #2
On Tue, Nov 17, 2015 at 10:40:51AM +0000, Chris Wilson wrote:
> We have varied reports of swizzling corruption on gen4 desktop, and
> confirmation that it is triggered by uneven memory banks. The
> implication is that the swizzling various 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.
> 
> Reported-by: Matti Hämäläinen <ccr@tnsp.org>
> References: https://bugs.freedesktop.org/show_bug.cgi?id=90725
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Matti Hämäläinen <ccr@tnsp.org>
> Cc: stable@vger.kernel.org
Tested-by: Matti Hämäläinen <ccr@tnsp.org>
-Chris
Daniel Vetter Nov. 18, 2015, 4:46 p.m. UTC | #3
On Tue, Nov 17, 2015 at 10:40:51AM +0000, Chris Wilson wrote:
> We have varied reports of swizzling corruption on gen4 desktop, and
> confirmation that it is triggered by uneven memory banks. The
> implication is that the swizzling various 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.
> 
> Reported-by: Matti Hämäläinen <ccr@tnsp.org>
> References: https://bugs.freedesktop.org/show_bug.cgi?id=90725
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Matti Hämäläinen <ccr@tnsp.org>
> Cc: stable@vger.kernel.org
> ---
>  drivers/gpu/drm/i915/i915_gem_fence.c | 4 ++--
>  1 file changed, 2 insertions(+), 2 deletions(-)
> 
> diff --git a/drivers/gpu/drm/i915/i915_gem_fence.c b/drivers/gpu/drm/i915/i915_gem_fence.c
> index 40a10b25956c..6ba68083aa95 100644
> --- a/drivers/gpu/drm/i915/i915_gem_fence.c
> +++ b/drivers/gpu/drm/i915/i915_gem_fence.c
> @@ -677,8 +677,8 @@ i915_gem_detect_bit_6_swizzle(struct drm_device *dev)
>  		 * the minimum size of a rank.
>  		 */
>  		if (I915_READ16(C0DRB3) != I915_READ16(C1DRB3)) {
> -			swizzle_x = I915_BIT_6_SWIZZLE_NONE;
> -			swizzle_y = I915_BIT_6_SWIZZLE_NONE;
> +			swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN;
> +			swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN;

Existing userspace goes boom if we hand it _UNKNOWN as the swizzle type.
We need the same lie as with L-shaped on g4x and only set the quirk, so
that phys_swizzle_mode is _UNKOWN. See

commit 5eb3e5a5e11d14f9deb2a4b83555443b69ab9940
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Sun Jun 28 09:19:26 2015 +0100

    drm/i915: Declare the swizzling unknown for L-shaped configurations

So if you keep the swizzle put add the

	dev_priv->quirks |= QUIRK_PIN_SWIZZLED_PAGES;

here instead (and drop patch 2) this looks good and gets my r-b. But
please include a reference to the above commit so we don't forget all
this.
-Daniel

>  		} else {
>  			swizzle_x = I915_BIT_6_SWIZZLE_9_10;
>  			swizzle_y = I915_BIT_6_SWIZZLE_9;
> -- 
> 2.6.2
> 
> _______________________________________________
> Intel-gfx mailing list
> Intel-gfx@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/intel-gfx
diff mbox

Patch

diff --git a/drivers/gpu/drm/i915/i915_gem_fence.c b/drivers/gpu/drm/i915/i915_gem_fence.c
index 40a10b25956c..6ba68083aa95 100644
--- a/drivers/gpu/drm/i915/i915_gem_fence.c
+++ b/drivers/gpu/drm/i915/i915_gem_fence.c
@@ -677,8 +677,8 @@  i915_gem_detect_bit_6_swizzle(struct drm_device *dev)
 		 * the minimum size of a rank.
 		 */
 		if (I915_READ16(C0DRB3) != I915_READ16(C1DRB3)) {
-			swizzle_x = I915_BIT_6_SWIZZLE_NONE;
-			swizzle_y = I915_BIT_6_SWIZZLE_NONE;
+			swizzle_x = I915_BIT_6_SWIZZLE_UNKNOWN;
+			swizzle_y = I915_BIT_6_SWIZZLE_UNKNOWN;
 		} else {
 			swizzle_x = I915_BIT_6_SWIZZLE_9_10;
 			swizzle_y = I915_BIT_6_SWIZZLE_9;