Re: [PATCH 3/6] drm/i915: Wrap the preallocated BIOS framebuffer and preserve for KMS fbcon v7

[Jesse Barnes <jbarnes@virtuousgeek.org>]

On Thu, 12 Dec 2013 23:54:37 +0100
Daniel Vetter <daniel@ffwll.ch> wrote:

> > @@ -258,8 +357,102 @@ static void intel_fbdev_destroy(struct drm_device *dev,
> >  
> >  	drm_fb_helper_fini(&ifbdev->helper);
> >  
> > -	drm_framebuffer_unregister_private(&ifbdev->ifb.base);
> > -	intel_framebuffer_fini(&ifbdev->ifb);
> > +	drm_framebuffer_unregister_private(&ifbdev->fb->base);
> > +	intel_framebuffer_fini(ifbdev->fb);
> > +	kfree(ifbdev->fb);  
> 
> No need to go the private fb route here anymore since now the fb is
> free-standing. Normal refcounting should work. But a separate prep/cleanup
> patch (prep since switching ifbdev->fb from struct to point would look
> neat as a separate patch).

Oh and can you explain this?  I wouldn't be surprised if I got the
refcounting wrong, but given how tricky it can be, can you explain
where we'll take the ref here, and show that the right thing will
happen if/when we mode set away from this buffer?

I haven't actually seen a bug here with or without this patch (no
crashes or warns), but I thought I needed this to make sure the obj
didn't get a negative count after a mode set...

-- 
Jesse Barnes, Intel Open Source Technology Center