>My belief here is that there is too much hard coding of magic numbers in the
>Linux kernel, and this is something that as the kernel starts running on more
>and more architectures, it ought to be got away from.
Do you want to modify all the ISA drivers? Like it or not this assumption
exists in a lot of places. For new code yes it's clearly wrong to hardwire
IRQ numbers based on a specific architecture, but you can make quite a strong
case for an ISA bus having to have specific IRQ numbering in much the same way
that we arrange for the ISA I/O ports to map to the 0...0xfff region. (ISA
shared memory is obviously another can of worms entirely, but in fact it ought
to pretty much come out in the wash so long as the ioremap fudge factor of
0xe0000000 stays.)
The IRQ numbers are currently completely arbitrary and I don't see any
disadvantage to rearranging them like this. The fact that it will cause some
machines to have an IRQ "hole" at 0..15 is scarcely a big deal, and you
could always put something else there on such architectures.
p.
unsubscribe: body of `unsubscribe linux-arm' to [EMAIL PROTECTED]
