Skip to contentSkip to conversation

drm/i915: Fix the offset issue for the stolen GEM objects

PATCH In review [email protected] 3 participants 4 messages Jan 9, 2014

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 1/7 drm/i915: Fix the offset issue for the stolen GEM objects

From: Akash Goel <[email protected]>

The 'offset' field of the 'scatterlist' structure was wrongly programmed with the offset value from the base of stolen area, whereas this field indicates the offset from where the interested data starts within the PAGE pointed to by the 'page-link' field. As a result when a new GEM object allocated from the stolen area is mapped to GTT, it could lead to an overwrite of GTT entries as the page count calculation will go wrong, refer the function 'sg_page_count'.

---
 drivers/gpu/drm/i915/i915_gem_stolen.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
drivers/gpu/drm/i915/i915_gem_stolen.c +1 -1
index fed87ec..1a24e84 100644
@@ -250,7 +250,7 @@ i915_pages_create_for_stolen(struct drm_device *dev,
 	}
 
 	sg = st->sgl;
-	sg->offset = offset;
+	sg->offset = 0;
 	sg->length = size;
 
 	sg_dma_address(sg) = (dma_addr_t)dev_priv->mm.stolen_base + offset;
Signature · 1 line
1.8.5.2
#1 lore raw
Quoted 10 lines · On Thu, Jan 09, 2014 at 11:00:17AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

The 'offset' field of the 'scatterlist' structure was wrongly programmed with the offset value from the base of stolen area, whereas this field indicates the offset from where the interested data starts within the PAGE pointed to by the 'page-link' field. As a result when a new GEM object allocated from the stolen area is mapped to GTT, it could lead to an overwrite of GTT entries as the page count calculation will go wrong, refer the function 'sg_page_count'.

This statement is incorrect since my use of sg here predates sg_page_iter.

The stolen sg has no page_link, the meaning of offset/length here are relative to the base of the stolen area.

However, if you wish to rephrase the above...
-Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre
#2 lore raw
Quoted 12 lines · Chris Wilson
Quoted 10 lines · On Thu, Jan 09, 2014 at 11:00:17AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

The 'offset' field of the 'scatterlist' structure was wrongly programmed with the offset value from the base of stolen area, whereas this field indicates the offset from where the interested data starts within the PAGE pointed to by the 'page-link' field. As a result when a new GEM object allocated from the stolen area is mapped to GTT, it could lead to an overwrite of GTT entries as the page count calculation will go wrong, refer the function 'sg_page_count'.

Quoted 7 lines

This statement is incorrect since my use of sg here predates sg_page_iter.

The stolen sg has no page_link, the meaning of offset/length here are relative to the base of the stolen area.

However, if you wish to rephrase the above...

Actually we add offset both to sg->offset and adjust the dma_bus_addr since this has been introduced in

commit 0104fdbb84d7adb0e377ed05bf75eba97b007544

Author: Chris Wilson <[email protected]>
Date:   Thu Nov 15 11:32:26 2012 +0000
    drm/i915: Introduce i915_gem_object_create_stolen()

But only Imre's conversion to the sg_page_iter started to pay any attention to sg->offset in

commit 6e995e231a90ce7c5ce2a9eae23c8e22f4388db1

Author: Imre Deak <[email protected]>
Date:   Mon Feb 18 19:28:04 2013 +0200

drm/i915: use for_each_sg_page for setting up the gtt ptes

So with a bit of commit message rewording and these references and cc:
stable this looks good.

Cheers, Daniel

Signature · 3 lines
Daniel Vetter
Software Engineer, Intel Corporation
+41 (0) 79 365 57 48 - http://blog.ffwll.ch
Tested-by
#3 lore raw
Quoted 33 lines · Daniel Vetter
Quoted 12 lines · Chris Wilson
Quoted 10 lines · On Thu, Jan 09, 2014 at 11:00:17AM +0530, [email protected] wrote:

From: Akash Goel <[email protected]>

The 'offset' field of the 'scatterlist' structure was wrongly programmed with the offset value from the base of stolen area, whereas this field indicates the offset from where the interested data starts within the PAGE pointed to by the 'page-link' field. As a result when a new GEM object allocated from the stolen area is mapped to GTT, it could lead to an overwrite of GTT entries as the page count calculation will go wrong, refer the function 'sg_page_count'.

Quoted 7 lines

This statement is incorrect since my use of sg here predates sg_page_iter.

The stolen sg has no page_link, the meaning of offset/length here are relative to the base of the stolen area.

However, if you wish to rephrase the above...

Actually we add offset both to sg->offset and adjust the dma_bus_addr since this has been introduced in

commit 0104fdbb84d7adb0e377ed05bf75eba97b007544

Author: Chris Wilson <[email protected]>
Date:   Thu Nov 15 11:32:26 2012 +0000
    drm/i915: Introduce i915_gem_object_create_stolen()

But only Imre's conversion to the sg_page_iter started to pay any attention to sg->offset in

Quoted 8 lines

commit 6e995e231a90ce7c5ce2a9eae23c8e22f4388db1

Author: Imre Deak <[email protected]>
Date:   Mon Feb 18 19:28:04 2013 +0200

drm/i915: use for_each_sg_page for setting up the gtt ptes

So with a bit of commit message rewording and these references and cc:
stable this looks good.

And

Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=69104

Maybe a reviewed-by when I see a good comment/changelog. -Chris

Signature · 1 line
Chris Wilson, Intel Open Source Technology Centre