[Xorg] The big multiconsole nasty

[Keith Packard]

Around 13 o'clock on Jul 6, Egbert Eich wrote:

> If you consider fbdev as something that lives solely in the kernel
> the only solution would be to place most - if not all - of the
> functionality mentioned above into the kernel.

I don't think this is a requirement.  The kernel now provides mechanisms to
push functionality required by the kernel up to user mode.  For the
graphics device, we can pend initialization until user mode is accessible
and initialize the graphics device at that point.

> VRAM coordination for printk's is also tricky: on some chipsets you 
> need to know if the chipset is idle. If the kernel can still communicate
> with the daemon process it can let this handle the output.

That's actually very difficult -- printk can occur at interrupt time at 
which point no communication with user mode is possible.  For cards with 
this limitation (which seems to be fewer these days than a few years ago), 
it may well be necessary to pend some printk output to synchronize with 
user mode device access.  Alternatively, you could push even more code 
down into the kernel and build a full kernel-based graphics driver (like 
DRI). For some systems, this may well be the right trade off.

-keith


-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 228 bytes
Desc: not available
URL: <http://lists.x.org/archives/xorg/attachments/20040706/e7583410/attachment.pgp>