Re: fbcon

["Adam C. Powell, IV" <adam powell nist gov>]

Bibek Sahu wrote:

>         The current kernel (generic) framebuffer code won't work on
> alphas.  It has some memory access ... methods ... that aren't very
> portable.  (note that the TGA framebuffer works because it bypasses all of
> these generic functions)
>
>         There is currently a matroxfb in development which provides
> partial acceleration for the console framebuffer.  It can be found in the
> cvs right now, and will probably be a part of 2.1.129.  It doesn't work on
> Alphas at the moment, but I'm working with the author to remedy that
> situation; Petr has been very helpful and cooperative so far.

Cool!  Any progress in the last two months?  I tried building 2.2.0-final with
matroxfb and booting, but no dice, even with video=vc.  Same problem as others
on this list: I get the first line of kernel output then it dies, regardless of
the video variable I give it.

>         In the distant past, GGI patched the kernel in order to work.
> With the inclusion of the framebuffer stuff, patches are no longer
> necessary.  It builds kgicon, a module which connects to the fbcon
> interface to provide its functionality but also provides the functionality
> of kgi -- the ggi Kernel Graphics Interface.  Unlike the kernel
> framebuffer setup, KGI has the ability to accelerate all parts of 2D video
> display (i.e., when complete it could give X a run for its money).  3D
> support is also in the works.  There is currently an XGGI which uses the
> ggi access routines to provide accelerated video for all ggi-supported
> cards.  Matrox support is pretty good.
>
>         However, ggi has io memory (mmio) access problems similar to that
> of the kernel framebuffer console drivers.  I am currently working with
> the GGI folks to remedy this situation as well.  However, this is a
> slightly larger undertaking than fixing matroxfb and is moving along more
> slowly.
>
>         Most GGI programs (including XGGI) use libggi for graphics
> routines.  Libggi has support for various targets.  One is direct console
> drawing via. kgi, another is the normal kernel framebuffer (fbcon), and
> the third is an X window.  The client program doesn't know or care which
> of these it's talking to.  An svgalib wrapper for GGI exists, and I
> believe it to work on Alphas (though I haven't tried).  Therefore, you can
> run svgalib programs in an X window.  If I ever get kgi working, you could
> also run those from the console.  If you have a TGA card, libggi has been
> reported to work with TGAfb.

Cool (though I don't have a TGA card).  Where is this documented?

>         For the record, I have a Matrox Millenium II with 4MB of vram, I
> normally use the standard linux text console and/or XF86_SVGA.  If I ever
> get GGI working right, that may change.  I have an sx164, 533MHz.

Me too, except LX.

>         If you're willing to test / code / anything, lemme know.  To the
> best of my knowledge, I'm the only one working on either of these for
> Alphas, and I'm in a rut with the code (for both systems) right now.  I
> could use all the help I can get.

Definitely willing to test, and spend some time coding, but can't claim great
video hacking skills.

-Adam `Cold Fusion' Powell, IV http://www.ctcms.nist.gov/~powell/ ____
USDoC, National Institute of Standards & Technology (NIST)  |\ ||<  |
Center for Theoretical and Computational Materials Science  | \||_> |