All of lore.kernel.org
 help / color / mirror / Atom feed
* [PATCH] drm/mm: Adjust start/end for coloring first
@ 2014-05-19  6:52 Chris Wilson
  2014-05-19  8:14 ` Daniel Vetter
  2014-05-19 10:24 ` Daniel Vetter
  0 siblings, 2 replies; 5+ messages in thread
From: Chris Wilson @ 2014-05-19  6:52 UTC (permalink / raw)
  To: dri-devel

The current user of the coloring will adjust the end points of the node
to leave a hole between disjoint memory types. This adjustment must be
performed first or else the derived size will conflict with the
adjustment and trigger the BUG_ON sanity checks that the node is within
bounds.

Fixes regression from
commit 62347f9e0f81d50e9b0923ec8a192f60ab7a1801
Author: Lauri Kasanen <cand@gmx.com>
Date:   Wed Apr 2 20:03:57 2014 +0300

    drm: Add support for two-ended allocation, v3

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
---
 drivers/gpu/drm/drm_mm.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
index 04a209e2b66d..911863bed9f3 100644
--- a/drivers/gpu/drm/drm_mm.c
+++ b/drivers/gpu/drm/drm_mm.c
@@ -264,12 +264,12 @@ static void drm_mm_insert_helper_range(struct drm_mm_node *hole_node,
 	if (adj_end > end)
 		adj_end = end;
 
-	if (flags & DRM_MM_CREATE_TOP)
-		adj_start = adj_end - size;
-
 	if (mm->color_adjust)
 		mm->color_adjust(hole_node, color, &adj_start, &adj_end);
 
+	if (flags & DRM_MM_CREATE_TOP)
+		adj_start = adj_end - size;
+
 	if (alignment) {
 		unsigned tmp = adj_start % alignment;
 		if (tmp) {
-- 
2.0.0.rc2

^ permalink raw reply related	[flat|nested] 5+ messages in thread
* Re: [PATCH] drm/mm: Adjust start/end for coloring first
  2014-05-19  6:52 [PATCH] drm/mm: Adjust start/end for coloring first Chris Wilson
@ 2014-05-19  8:14 ` Daniel Vetter
  2014-05-19  8:21   ` Chris Wilson
  2014-05-19 10:24 ` Daniel Vetter
  1 sibling, 1 reply; 5+ messages in thread
From: Daniel Vetter @ 2014-05-19  8:14 UTC (permalink / raw)
  To: Chris Wilson; +Cc: dri-devel

On Mon, May 19, 2014 at 07:52:37AM +0100, Chris Wilson wrote:
> The current user of the coloring will adjust the end points of the node
> to leave a hole between disjoint memory types. This adjustment must be
> performed first or else the derived size will conflict with the
> adjustment and trigger the BUG_ON sanity checks that the node is within
> bounds.
> 
> Fixes regression from
> commit 62347f9e0f81d50e9b0923ec8a192f60ab7a1801
> Author: Lauri Kasanen <cand@gmx.com>
> Date:   Wed Apr 2 20:03:57 2014 +0300
> 
>     drm: Add support for two-ended allocation, v3
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

Do we have a bugzilla for this, or why did igt not scream about this
failure?
-Daniel

> ---
>  drivers/gpu/drm/drm_mm.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
> index 04a209e2b66d..911863bed9f3 100644
> --- a/drivers/gpu/drm/drm_mm.c
> +++ b/drivers/gpu/drm/drm_mm.c
> @@ -264,12 +264,12 @@ static void drm_mm_insert_helper_range(struct drm_mm_node *hole_node,
>  	if (adj_end > end)
>  		adj_end = end;
>  
> -	if (flags & DRM_MM_CREATE_TOP)
> -		adj_start = adj_end - size;
> -
>  	if (mm->color_adjust)
>  		mm->color_adjust(hole_node, color, &adj_start, &adj_end);
>  
> +	if (flags & DRM_MM_CREATE_TOP)
> +		adj_start = adj_end - size;
> +
>  	if (alignment) {
>  		unsigned tmp = adj_start % alignment;
>  		if (tmp) {
> -- 
> 2.0.0.rc2
> 
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/dri-devel

-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

^ permalink raw reply	[flat|nested] 5+ messages in thread
* Re: [PATCH] drm/mm: Adjust start/end for coloring first
  2014-05-19  8:14 ` Daniel Vetter
@ 2014-05-19  8:21   ` Chris Wilson
  2014-05-19 10:22     ` Daniel Vetter
  0 siblings, 1 reply; 5+ messages in thread
From: Chris Wilson @ 2014-05-19  8:21 UTC (permalink / raw)
  To: Daniel Vetter; +Cc: dri-devel

On Mon, May 19, 2014 at 10:14:27AM +0200, Daniel Vetter wrote:
> On Mon, May 19, 2014 at 07:52:37AM +0100, Chris Wilson wrote:
> > The current user of the coloring will adjust the end points of the node
> > to leave a hole between disjoint memory types. This adjustment must be
> > performed first or else the derived size will conflict with the
> > adjustment and trigger the BUG_ON sanity checks that the node is within
> > bounds.
> > 
> > Fixes regression from
> > commit 62347f9e0f81d50e9b0923ec8a192f60ab7a1801
> > Author: Lauri Kasanen <cand@gmx.com>
> > Date:   Wed Apr 2 20:03:57 2014 +0300
> > 
> >     drm: Add support for two-ended allocation, v3
> > 
> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> 
> Do we have a bugzilla for this, or why did igt not scream about this
> failure?

How would igt scream? Look at the patch and think of how many possible
ways the current kernel would explode. Then think about how they are
exposed to userspace.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre

^ permalink raw reply	[flat|nested] 5+ messages in thread
* Re: [PATCH] drm/mm: Adjust start/end for coloring first
  2014-05-19  8:21   ` Chris Wilson
@ 2014-05-19 10:22     ` Daniel Vetter
  0 siblings, 0 replies; 5+ messages in thread
From: Daniel Vetter @ 2014-05-19 10:22 UTC (permalink / raw)
  To: Chris Wilson, Daniel Vetter, dri-devel

On Mon, May 19, 2014 at 09:21:23AM +0100, Chris Wilson wrote:
> On Mon, May 19, 2014 at 10:14:27AM +0200, Daniel Vetter wrote:
> > On Mon, May 19, 2014 at 07:52:37AM +0100, Chris Wilson wrote:
> > > The current user of the coloring will adjust the end points of the node
> > > to leave a hole between disjoint memory types. This adjustment must be
> > > performed first or else the derived size will conflict with the
> > > adjustment and trigger the BUG_ON sanity checks that the node is within
> > > bounds.
> > > 
> > > Fixes regression from
> > > commit 62347f9e0f81d50e9b0923ec8a192f60ab7a1801
> > > Author: Lauri Kasanen <cand@gmx.com>
> > > Date:   Wed Apr 2 20:03:57 2014 +0300
> > > 
> > >     drm: Add support for two-ended allocation, v3
> > > 
> > > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> > 
> > Do we have a bugzilla for this, or why did igt not scream about this
> > failure?
> 
> How would igt scream? Look at the patch and think of how many possible
> ways the current kernel would explode. Then think about how they are
> exposed to userspace.

Yeah, reading the patch helps ;-) So this only collides with the top-down
vs. bottum-up separation patch for cache levels you've posted somewhere.
-Daniel
-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

^ permalink raw reply	[flat|nested] 5+ messages in thread
* Re: [PATCH] drm/mm: Adjust start/end for coloring first
  2014-05-19  6:52 [PATCH] drm/mm: Adjust start/end for coloring first Chris Wilson
  2014-05-19  8:14 ` Daniel Vetter
@ 2014-05-19 10:24 ` Daniel Vetter
  1 sibling, 0 replies; 5+ messages in thread
From: Daniel Vetter @ 2014-05-19 10:24 UTC (permalink / raw)
  To: Chris Wilson; +Cc: dri-devel

On Mon, May 19, 2014 at 07:52:37AM +0100, Chris Wilson wrote:
> The current user of the coloring will adjust the end points of the node
> to leave a hole between disjoint memory types. This adjustment must be
> performed first or else the derived size will conflict with the
> adjustment and trigger the BUG_ON sanity checks that the node is within
> bounds.
> 
> Fixes regression from
> commit 62347f9e0f81d50e9b0923ec8a192f60ab7a1801
> Author: Lauri Kasanen <cand@gmx.com>
> Date:   Wed Apr 2 20:03:57 2014 +0300
> 
>     drm: Add support for two-ended allocation, v3
> 
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> ---
>  drivers/gpu/drm/drm_mm.c | 6 +++---
>  1 file changed, 3 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/gpu/drm/drm_mm.c b/drivers/gpu/drm/drm_mm.c
> index 04a209e2b66d..911863bed9f3 100644
> --- a/drivers/gpu/drm/drm_mm.c
> +++ b/drivers/gpu/drm/drm_mm.c
> @@ -264,12 +264,12 @@ static void drm_mm_insert_helper_range(struct drm_mm_node *hole_node,
>  	if (adj_end > end)
>  		adj_end = end;
>  
> -	if (flags & DRM_MM_CREATE_TOP)
> -		adj_start = adj_end - size;
> -
>  	if (mm->color_adjust)
>  		mm->color_adjust(hole_node, color, &adj_start, &adj_end);
>  
> +	if (flags & DRM_MM_CREATE_TOP)
> +		adj_start = adj_end - size;

With this we still don't handle interactions with alignment correctly. I
guess while we're at this we might as well fix that, too.
-Daniel

> +
>  	if (alignment) {
>  		unsigned tmp = adj_start % alignment;
>  		if (tmp) {
> -- 
> 2.0.0.rc2
> 
> _______________________________________________
> dri-devel mailing list
> dri-devel@lists.freedesktop.org
> http://lists.freedesktop.org/mailman/listinfo/dri-devel

-- 
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch

^ permalink raw reply	[flat|nested] 5+ messages in thread
end of thread, other threads:[~2014-05-19 10:24 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2014-05-19  6:52 [PATCH] drm/mm: Adjust start/end for coloring first Chris Wilson
2014-05-19  8:14 ` Daniel Vetter
2014-05-19  8:21   ` Chris Wilson
2014-05-19 10:22     ` Daniel Vetter
2014-05-19 10:24 ` Daniel Vetter

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.