I am missing a subtle detail from the bus_space_map man page:

If I have a PCI graphics card, and use 

  bus_space_map(..., BUS_SPACE_MAP_LINEAR, ..)
  bus_space_vaddr()

to create a linear mapping for the framebuffer and get a pointer to it's start -
is there any endianess guarantee? Like most bus_space_* functions make sure
PCI access is little endian, is this true for the above mapping too?

On sparc64 it currently is not, but it used to be. The change was more or less
accidental, and I'm wondering if it the current behaviour is better or worse
than before.

Martin