This patch series at first moves get_speed_cap_mask from DRM to PCI, fixes all radeon references (only driver that uses it) and then implements a architecture specific implementation for ppc64 for it.
This is good because the drm_pcie_get_speed_cap_mask function is more architecture goo than something drm-specific, and also because this way architectures that have different means of obtaining these values can do it their way.
