Re: [Linux-fbdev-devel] Re: [PATCH 1/1: 2.6.15-rc5-git3] Fixed and updated CyblaFB

[Knut P. <Knu...@t-...>]

Andrew Morton wrote:

>Well yes, but effects on cyblafb are possible!  We'd be putting new and
>un-tested-in-linus's-tree code into the tree at the last minute.
>
>It would hardly be a calamity if we were to merge this into 2.6.15 of
>course, but the principle which you describe ain't right.
>  
>
I write and maintain cyblafb because vesafb is really too slow on my  
532 Mhz
system.

All changes to cyblafb fall into the following categories:

    -  code to take advantage of all available video memory and thus to 
allow
        ywrap scrolling.
    - code to allow vxres > xres and thus to allow xwrap scrolling while
       rotation is set to 1 or 3
    - code that helps in testing the upper framebuffer layers. As far as 
I know,
       cyblafb is the only driver that can be configured at boot or 
module load
       time to request image data with 8, 16 or 32 bit image data alignment.
    - bug fixes

I use the cyblafb framebuffer console for about 80% of my work. Any bug 
in the first
two categories would be immediately recognized by either a locked PC or 
display
corruptions, the code of the third category has been seriously tested, 
and there is also
no real possibility of remaining bugs as these also would be either 
visible or would
lock the system immediately.

I don´t know of any cyblafb user that is running the -mm kernels, so 
testing the
changes there does conform to principles but does not help much in reality.

>Masking pointers to 32 bits.  Is this driver supposed to be run on 64-bit
>machines?
>  
>
Never can it be run on anything else but 32 bit machines because this 
graphics core is
integrated in the VIA VT8601A northbridge. So the cpu is either a VIA C3,
a VIA Eden, a Pentium III or a Celeron. It is planned to add support for 
closely related
graphics cores, but also those can be run with nothing but 32 bit cpus, 
starting from the
first pentium. Even if this driver would be compiled into a 64 bit 
kernel, the
imageblit function never would be executed as the init code never will 
find the
cyberblade/i1 graphics core in such a system.

Sorry for the whitespace damage.

cu,
 knut