Re: [Qemu-devel] [PATCH v4 3/7] Implement fw_cfg DMA interface

[Laszlo Ersek <lersek@redhat.com>]

On 10/01/15 19:18, Peter Maydell wrote:
> On 1 October 2015 at 15:36, Laszlo Ersek <lersek@redhat.com> wrote:
>> I think I finally understand this difference now. It is all rooted in
>> the difference between the internal APIs sysbus_add_io() and
>> sysbus_init_mmio(). Both of these are called from the device realize
>> functions, but the first (sysbus_add_io()) wants the IO port address at
>> once, whereas the second (sysbus_init_mmio()) doesn't want the address
>> -- the actual mapping (sysbus_mmio_map()) is delayed to board code; the
>> device code doesn't want to be aware of it.
> 
> Yes. The sysbus_add_io() API is firmly wedded to the x86 I/O port
> concept and to the idea that devices are at fixed I/O port addresses
> which don't depend on what board they're in (because the few
> non-x86 systems that set up some kind of "IO port" abstraction
> are generally doing it to look more x86-like).
> 
> That said, sysbus_add_io() is one of those odd functions which
> we use half a dozen times in the whole codebase and which leaves
> me wondering if it ought to be refactored to work differently
> (eg split into "declare IO ports" and "map IO ports into IO space"
> like the mmio functions)...

I had the same idea looming in the back of my mind, but I wouldn't know
how to attack the refactoring (plus I wouldn't want to delay Marc's work
with it -- after all the function is not being introduced by this
series), so I didn't bring it up. :P

Thanks!
Laszlo