Andreas Busse bubbles:
> Neil made the assumption that it will take a bit more than half a man-year
> to port the kernel and drivers. This depends on how many filesystems we will
> support. For me, ext2fs and the ISO fs would be enough.
The filesystems are entirely in the machine-independant code.
I don't think many changes at all will be needed for porting all of
the filesystems. The MMU, DMA, interrupt handling and simplest device
drivers (console, serial, naif SCSI/ether) will probably be the longest to
get going because the machine is useless until they do.
> Everything else *should* compile out of the box, except the X11-server stuff.
> That sounds not too bad, isn't it ?
With a dumb frame buffer there will be very little work getting X11 going.
There's already lots of support in the MIT code for 8 bit deep dumb
frame buffers. It'll just be a matter of reading the keyboard and mouse
properly (I think - I could be completely wrong).
Libraries could be a challenge to port. Not because its very difficult
or machine dependant, but all kinds of correct decisions need to be made.
For example, if I were to port linux to an 4000, should off_t etc be
64 bit? If so, should the library and compile environment have support
for a 32 bit environment and a 64 bit one (by having the 32 bit syscall
lib functions convert their args and returns to 64 bit).
> I can't say how long it takes to make the hardware working, because it
> depends on what we want.
Bearing in mind that 4000 based commercial machines are becoming available
*now* at reasonable or nearly reasonable prices. For example, a brocure
from a company called "Tangent" turned up here selling an R4000 based
machine (can't remember which variant) running at 100MHz, 16MB ram,
16k cache (cannot sure if its I+D combined or 16k each), 200MB disk, hi
res graphics (~1000x1000x8) with 64 bit path to CPU + S3 accellerator,
E?ISA slots, monitor and Windows NT (SVR4 port available) for $4000 US.
That was the base machine with others up to 150MHz, more cache and memory
and so on. A similar machine is $10000 here, but that's predictable.
Whether or not thay can actually deliver is another question. Its also
possible that they'd sell motherboards alone.
I'm not cash starved and I'm keen to port linux to a Mips machine
(I've started already, mainly by looking at the i386 kernel for machine
dependancies and isolating them), so this kind of thing is an option
for me. It also means there's some chance of a Mips kernel waiting for
any board that gets built.