2012. 11. 10. 오전 2:10 Rob Clark <robdclark@xxxxxxxxx> 작성:
> On Fri, Nov 9, 2012 at 1:39 AM, Inki Dae <inki.dae@xxxxxxxxxxx> wrote:
>> diff --git a/include/drm/drm_crtc.h b/include/drm/drm_crtc.h
>> index 3fa18b7..92889be 100644
>> --- a/include/drm/drm_crtc.h
>> +++ b/include/drm/drm_crtc.h
>> @@ -256,6 +256,7 @@ struct drm_framebuffer {
>> struct kref refcount;
>> struct list_head head;
>> struct drm_mode_object base;
>> + struct drm_crtc *crtc;
>
> note that one fb can be attached to multiple crtc's (and/or planes)..
> so this will never work.
>
I know that. drm_framebuffer'crtc means that crtc'dma is accessing this fb. So when pageflip and setcrtc are called, this crtc will be updated again.
> Anyways, I think the right answer is somehow reference-counting. You
> should *somewhere* be holding a ref to bo's until there is no more dma
> access, either directly or indirectly by holding a ref to the fb
> (which holds a ref to the bo(s))
>
Understood. Yes, we could resolve this issue using reference-counting also. I'd just like to make this issue to be solved more generically. In fact, I couldn't understand why my patch has some problem yet. :(
Thanks for comments,
Inki Dae
> BR,
> -R
>
>> const struct drm_framebuffer_funcs *funcs;
>> unsigned int pitches[4];
>> unsigned int offsets[4];
>> --
>> 1.7.4.1
>>
>> _______________________________________________
>> dri-devel mailing list
>> dri-devel@xxxxxxxxxxxxxxxxxxxxx
>> http://lists.freedesktop.org/mailman/listinfo/dri-devel
> _______________________________________________
> dri-devel mailing list
> dri-devel@xxxxxxxxxxxxxxxxxxxxx
> http://lists.freedesktop.org/mailman/listinfo/dri-devel
_______________________________________________
dri-devel mailing list
dri-devel@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/dri-devel
