>This could certainly be done, it would be very `pretty' and very very
>inefficient. It would basically end up looking like something like UDI
>(*shiver*).
Yep, you are right. All the offset problems should be calculated once at
fixup time to avoid a function call for accesses, I was probably too
sleepy when I wrote this.
As far as endian is concerned, I beleive it's right to asume little
endian, but I agree with the fact that "non-swap" routines should be
provided. This is especially useful not only for boards with HW byte
swapping, but, as Paul noticed, when moving byte streams (usually between
memory and a FIFO) 16 or 32 bits at a time, where, even on BE machines,
swap must not be done. Drivers doing that could use the non-swap versions
without care of the machine's endian. They can, of course, be re-defined
per-driver, but I beleive there are enough cases of using those to
justify implementing them in the generic io interface.
--
Perso. e-mail: <mailto:bh40@calva.net>
Work e-mail: <mailto:benh@mipsys.com>
BenH. Web : <http://calvaweb.calvacom.fr/bh40/>
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/