On Wed, Aug 31, 2016 at 10:45 PM, Haixia Shi <hshi@xxxxxxxxxxxx> wrote: > For details see https://bugs.chromium.org/p/chromium/issues/detail?id=468050 > > So drm_mode_config_cleanup() is called from udl_driver_unload() in > which we found there's still a framebuffer left, hence the WARN in > drm_crtc.c:5495. This also forcefully releases all the buffers. > > A bit later the actual drm_buf_release comes in which attempts to > release the buffer again. Leaving a drm_framebuffer behind on unload is definitely a bug, but not fixed with this patch here I think. The dma-buf lifetime issue is far worse, since we simply don't handling those leaking past the lifetime of the exporting drm_device at all. The dma-buf has references to a lot more than just the vma manager. What we probably need is that every exported dma-buf holds a ref on the underlying drm_device, but that means untangling the refcounting&cleanup of that vs unplugging it. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch _______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/dri-devel
