On Sat, Aug 25, 2018 at 05:38:12PM +0200, Mark Kettenis wrote: > > Here is a diff to speed up the initial framebuffer console on amd64. > It remaps the framebuffer write-combining early on, which speeds up > writing to the framebuffer considerably. > > Remapping is done using _bus_space_map(). Otherwise inteldrm(4) and > radeondrm(4) will fail to attach (or even panic) because the graphics > aperture can't be mapped. This mapping stays around even if efifb(4) > doesn't attach. There is no good place to unmap it without risking > some random kernel printf causing a fault. > > The rasops(4) bit is a bugfix that probably should go in regardless. > It makes sure we don't clear the character mapping if the RI_CLEAR > flag isn't set. > > Can somebody with a UEFI machine that has radeondrm(4) test this for > me?
Works for me. Tested with: $ dmesg | grep radeondrm radeondrm0 at pci1 dev 0 function 0 vendor "ATI", unknown product 0x665f rev 0x81 drm0 at radeondrm0 radeondrm0: msi radeondrm0: 1024x768, 32bpp wsdisplay0 at radeondrm0 mux 1: console (std, vt100 emulation), using wskbd0