On 09/28/2010 09:23 AM, Michel D?nzer wrote: > On Mon, 2010-09-27 at 15:20 +0200, Thomas Hellstrom wrote: > >> Avoids an oops during drm_irq_uninstall. >> > Why not fix the oops instead of providing dummy hooks to work around it? > > > This is because the vblank system in DRM appears unconditional. You can't choose not to enable it, because then you'd hit uninitialized values during various ioctls and drm_irq_uninstall.
Simply testing for presence of the get_vblank_counter function before calling it would mean providing the dummy hook in core drm instead of the driver... A correct fix would require making the vblank system optional, and isn't a trivial task, so I figured better to put that on the to-do list and fix the oops now. /Thomas