Re: [Qemu-devel] [PATCH 1/3] sh4: mmio based CF support on r2d board.

[Blue Swirl]

On 10/29/08, andrzej zaborowski <address@hidden> wrote:
> 2008/10/27 Blue Swirl <address@hidden>:
>
> > On 10/27/08, andrzej zaborowski <address@hidden> wrote:
>  >> 2008/10/26 Blue Swirl <address@hidden>:
>  >>  > It would be better to make the init function take instead of mmio
>  >>  > pointer, two target_phys_addr_t parameters and do the physical memory
>  >>  > registration there.
>  >>
>  >>
>  >> Why would it be better?  This implementation seems more flexible.
>  >
>  > How is this more flexible? Now r2d.c has to know the number of
>  > registers in the IDE controller which should be internal to ide.c.
>
>
> I think the size of the mapping is not related to the number of the
>  registers - it should be the same as the mapping on the real machine.

Yes¸ but I think more exotic setups should be done by some kind of
glue logic, that way the interface to the device stays simple.