Re: vesafb mtrr setup question

[Arvind Sankar <arvinds@MIT.EDU>]

On Tue, Aug 26, 2003 at 04:42:23PM +0100, Alan Cox wrote:
> On Llu, 2003-08-25 at 20:43, Arvind Sankar wrote:
> > In the first place, the power of two computation computes the largest
> > power of 2 that is _smaller_ than video_size, so it looks like an
> > off-by-1 bug.
> 
> Not a bug - we don't know what lives above it so we can't extend the
> mtrr safely
> 
Ah. On a side not, could you drop a quick hint as to how
screen_info.lfb_size is obtained?

In older (or just different?) versions of vesafb, the video_size was
actually computed by multiplying xres, yres, and the bpp.

> > >         /* Try and find a power of two to add */
> > >         while (temp_size && mtrr_add(video_base, temp_size, MTRR_TYPE_WRCOMB, 1)==-EINVAL) {
> > >                 temp_size >>= 1;
> > >         }
> > > }
> > 
> > Secondly, what's the point of requesting a smaller write-combining
> > segment that won't cover all the video memory being used?
> 
> Generally we don't use all the videoram. Its a heuristic rather than
> perfection. You might want to play with improvements
> 

I thought the yres_virtual was computed based on how much video_ram was
being used, so all of it _is_ being used, for scrollback?

-- arvind