Re: [RFC PATCH V1 01/12] hvm/ioreq: Make x86's IOREQ feature common

[Julien Grall <julien@xxxxxxx>]

On 13/08/2020 00:08, Stefano Stabellini wrote:
> > > It is very similar to your second patch with a small change on calling
> > > try_handle_mmio from handle_mmio and setting the register there. Do you
> > > think that would work?
> > If I understood correctly what you had suggested and properly implemented then
> > it works, at least I didn't face any issues during testing.
> > I think this variant adds some extra operations comparing to the previous one
> > (for example an attempt to find a mmio handler at try_handle_mmio). But, if
> > you think new variant is cleaner and better represents how the state machine
> > should look like, I would be absolutely OK to take this variant for non-RFC
> > series. Please note, there was a request to move try_fwd_ioserv() to
> > arm/ioreq.c (I am going to move new handle_ioserv() as well).
>   
> Yes, I think this version better represents the state machine, thanks
> for looking into it. I think it is good.
>
> In terms of number of operations, it doesn't look very concerning (in
> the sense that it doesn't seem to add that many ops.) However we could
> maybe improve it by passing a reference to the right mmio handler from
> handle_mmio to try_handle_mmio if we have it. Or maybe we could save a
> reference to the mmio handler as part of v->arch.hvm.hvm_io.io_req.
There is no MMIO handler for the IOREQ handling. So I am not entirely 
sure what you are suggesting.
Cheers,

--
Julien Grall