On 19.06.14 22:54, Paolo Bonzini wrote:
Il 04/06/2014 14:28, Alexander Graf ha scritto:
But do we need that level of complexity for normal devices usually? In a
normal platform world (SoCs, PV machines) we have a flat memory
layout we
can plug our device memory into. We also have a flat IRQ model where we
can plug our device IRQs into.
So the idea for platform devices arose. A platform device is really
just a
device that exposes its qemu_irq slots and memory regions to the world.
That allows us to write machine specific code that maps a platform
device
wherever the machine thinks fits nicely. It also allows that same
machine
to generate a device tree entry for platform devices easily, as it is
fully
aware of the interrupt lines and places it was mapped to.
A device (read: user) may or may not explictly request to be mapped at a
specific IRQ and/or memory address. If no explicit mapping is requested,
platform devices can get mapped at convenient places by the machine.
The actual pressing issue this solves is that today it's impossible
to spawn
serial ports from the command line. With this patch set, it's
possible to
do so. But it lays the groundwork for much more...
I have more than a suspect that you and Peter Crosthwaite are trying
to do the same. I hope that memory region QOMification can get into
2.1, that can be a starting point that we can work from.
The major difference between the two of us is that Peter wants to have
the user specify everything on the command line so he can build a VM
from the device tree description of a system.
I want to make sure the user does not have to specify anything that
could go wrong on the command line, similar to how he can just use
-device pci-dev today. The more we automatically allocate for him, the
less can go wrong.
Alex
