Bug 103502

Created attachment 135152 [details]
dmesg

some other errors that are present while booting.

Hmm, tiling alarm bells for i915gm:

Active (rcs0) [14]:
    00000000_0050b000    57344 3e 02 [ 36f00 00 00 00 00 ] 00 Y dirty uncache
...
0x00507034:      0x7d8e0001: 3DSTATE_BUFFER_INFO
0x00507038:      0x03600800:    color, tiling = Y, pitch=2048
0x0050703c:      0x0050b000:    address

Unease as we tend to assume these implicit fencing require the "unfenced" alignment. Likely we reused an offset and failed to notice a change in alignment constraints. Something like:

diff --git a/drivers/gpu/drm/i915/i915_gem_execbuffer.c b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
index 3d7190764f10..7f53b4860428 100644
--- a/drivers/gpu/drm/i915/i915_gem_execbuffer.c
+++ b/drivers/gpu/drm/i915/i915_gem_execbuffer.c
@@ -331,6 +331,10 @@ eb_vma_misplaced(const struct drm_i915_gem_exec_object2 *entry,
        if (entry->alignment && !IS_ALIGNED(vma->node.start, entry->alignment))
                return true;
 
+       if (flags & __EXEC_OBJECT_NEEDS_MAP &&
+           !IS_ALIGNED(vma->node.start, vma->fence_alignment))
+               return true;
+
        if (flags & EXEC_OBJECT_PINNED &&
            vma->node.start != entry->offset)
                return true;

commit 1d033beb20d6d5885587a02a393b6598d766a382
Author: Chris Wilson <chris@chris-wilson.co.uk>
Date:   Tue Oct 31 10:36:07 2017 +0000

    drm/i915: Check incoming alignment for unfenced buffers (on i915gm)
    
    In case the object has changed tiling between calls to execbuf, we need
    to check if the existing offset inside the GTT matches the new tiling
    constraint. We even need to do this for "unfenced" tiled objects, where
    the 3D commands use an implied fence and so the object still needs to
    match the physical fence restrictions on alignment (only required for
    gen2 and early gen3).
    
    In commit 2889caa92321 ("drm/i915: Eliminate lots of iterations over
    the execobjects array"), the idea was to remove the second guessing and
    only set the NEEDS_MAP flag when required. However, the entire check
    for an unusable offset for fencing was removed and not just the
    secondary check. I.e.
    
            /* avoid costly ping-pong once a batch bo ended up non-mappable */
            if (entry->flags & __EXEC_OBJECT_NEEDS_MAP &&
                !i915_vma_is_map_and_fenceable(vma))
                    return !only_mappable_for_reloc(entry->flags);
    
    was entirely removed as the ping-pong between execbuf passes was fixed,
    but its primary purpose in forcing unaligned unfenced access to be
    rebound was forgotten.
    
    Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103502
    Fixes: 2889caa92321 ("drm/i915: Eliminate lots of iterations over the execobjects array")
    Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
    Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
    Link: https://patchwork.freedesktop.org/patch/msgid/20171031103607.17836-1-chris@chris-wilson.co.uk
    Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

*** Bug 103723 has been marked as a duplicate of this bug. ***

(In reply to Chris Wilson from comment #4)
> *** Bug 103723 has been marked as a duplicate of this bug. ***

How do I implement this FIX on Ubuntu Artful Aardvark?

(In reply to Chris Wilson from comment #3)
> commit 1d033beb20d6d5885587a02a393b6598d766a382
> Author: Chris Wilson <chris@chris-wilson.co.uk>
> Date:   Tue Oct 31 10:36:07 2017 +0000
> 
>     drm/i915: Check incoming alignment for unfenced buffers (on i915gm)
>     
>     In case the object has changed tiling between calls to execbuf, we need
>     to check if the existing offset inside the GTT matches the new tiling
>     constraint. We even need to do this for "unfenced" tiled objects, where
>     the 3D commands use an implied fence and so the object still needs to
>     match the physical fence restrictions on alignment (only required for
>     gen2 and early gen3).
>     
>     In commit 2889caa92321 ("drm/i915: Eliminate lots of iterations over
>     the execobjects array"), the idea was to remove the second guessing and
>     only set the NEEDS_MAP flag when required. However, the entire check
>     for an unusable offset for fencing was removed and not just the
>     secondary check. I.e.
>     
>             /* avoid costly ping-pong once a batch bo ended up non-mappable
> */
>             if (entry->flags & __EXEC_OBJECT_NEEDS_MAP &&
>                 !i915_vma_is_map_and_fenceable(vma))
>                     return !only_mappable_for_reloc(entry->flags);
>     
>     was entirely removed as the ping-pong between execbuf passes was fixed,
>     but its primary purpose in forcing unaligned unfenced access to be
>     rebound was forgotten.
>     
>     Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=103502
>     Fixes: 2889caa92321 ("drm/i915: Eliminate lots of iterations over the
> execobjects array")
>     Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
>     Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
>     Link:
> https://patchwork.freedesktop.org/patch/msgid/20171031103607.17836-1-
> chris@chris-wilson.co.uk
>     Reviewed-by: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>

This is the FIX I was asking about how to apply :D 
I clicked the wrong reply button.

Latest tip or stable should have it upstream already.
https://cgit.freedesktop.org/drm-tip
https://www.kernel.org

Closing issue. Thanks.