Here's another doublescan fix for atyfb.
Previosly var<->crtc conversion would do different things depending on=20
which way the conversion was made.=20
var->crtc: vdisp =3D ((yres-1)<<1)=20
crtc->var: yres =3D ((vdisp+1)>>1)
That's obviously wrong. And thus fbset would report incorrect yres value.=
The patch changes var->crtc conversion to do vdisp =3D ((yres<<1)-1).
I'm not entirely sure which way is the right way but at least now things=20
work as expected.