New DRM model

[Benjamin Herrenschmidt]

On Tue, 2004-02-10 at 15:22, Jon Smirl wrote:
> --- Benjamin Herrenschmidt <benh@kernel.crashing.org> wrote:
> > we synchronously stop the CCE, wait for everything to be idle, blast
> > the text using a very simple low level text engine (like the one I use
> > for early boot debug on ppc), and restart the CCE engine.
> > 
> > That's just an idea at this point... if we go that way, we may want
> > to still use the console loglevel to avoid having things like ethernet
> > link change messages override the Xserver on normal setups :)
> 
> If xerver is composting off from vertical retrace the blast is going to
> disappear 1/60 of second after it is written.

That's true... 

> An oops can generate a 'blue screen of death" without much problem. Just leave
> the monitor in whatever mode it's in (hope that we weren't halfway into a mode
> change), erase the screen, print the oops, and wait for reboot. Force all mode
> changes through the driver will allow this to be implemented.
> 
> printks have got to buffer if xserver is running and then go into a console
> window and get compositied with everything else. If a full screen console window
> is up it should be possible to draw them at interrupt time.

Well... I still like an emergency printk that blasts on top of the screen,
may still be handy for debugging. We could indeed have the screen locked
into some "penguin screen of frozen icecream", but it would be nice to
still be able to switch to normal display with sysrq for example. A lot
of oops will leave the box up, and in some case, you really want it still
up to gather more data about the problem...

Ben.