chammer@HRZ.Uni-Bielefeld.DE wrote:
: To waste my time with one of the more important
: points in the netbsd-amiga port i wonder if there
: is anybody going to beautify some of the device
: driver sources:) 
: I really dont want to offend anybody but looking
: thru it i find here and there locations where i
: would write it in one or another way different.
: 
: Another question i have is(of at least same importance:):
: Would it hurt anything to use the same major numbers
: for devices on the different ports? Of course it
: may hurt backwart compatibility.. 
: 
: Is it possible to attach several drivers to the same
: ZorroII board? As i understand it with the normal way
: by checking man and prod ids only the first device is
: configured, not? What should i do to solve this?
: 
: Is there any kind person who explains to me how xservers normally
: access the gfxcard ram? I thought they map the device with
: a pmap call into ram and have to sync it explicit when write

no, that happens automatically... 

: accesses occured to the ram but i cant find a read,write function

you dont need any.

: in the grf stuff. Does the pmap call handle devices something
: different? I thought it accesses the device with open(),read() calls
: as a normal file. I have the problem that i want to write a device
: for vga cards via crosslink but isa cards accessed via crosslink
: are mapped in a special way(bytes word aligned, words longword aligned)

Hm, you would have to tell that to the X server, by modifying the cfb code
a bit. I imagine this should be possible (tell it to access 16 bit pixels,
but that only the upper half is significant, or s.th. similar). 

Of course, it would be nice for your application, if the pixblt ioctl of
your card would be implemented in grf... then the X server wouldn't have
to think about how to access those splitted registers...

: I imagine to hide this behaviour to the xserver by a special read function
: in the device driver but it seems that there is no such function in
: the grf devices. No way?

You would have to implement that... and what ppl. told me when I wanted
to implement the mapping of the A2410 into the virtual memory space, is,
that thinks like this should better be left out of the kernel,
expecially for the heavy workloads X servers produce.

Regards,
	Ignatios