Comment 75 for bug 219630

Oliver Grawert has provided some analysis of this symlink case and found that the symlink itself does not cause problems; the problems only manifest when the kernel framebuffer module is loaded. I'm persuaded by his and Bryce's investigations that the symlink is not the direct cause of the problems, and that even if it is indirectly related, for compatibility reasons dropping the symlink is not the appropriate choice of action in the SRU due to the implied regressions.

I'm therefore accepting Bryce's subsequent upload of the -geode package into hardy-proposed, which restores the symlink. If you still see problems with LTSP and the amd_drv.so link installed, please first try to blacklist the kernel framebuffer module (Oliver can provide a recipe for this) and report the results.