Re: [Qemu-devel] Re: [PATCH 1/5] Add target memory mapping API

[Anthony Liguori]

Ian Jackson wrote:
> Anthony Liguori writes ("Re: [Qemu-devel] Re: [PATCH 1/5] Add target memory mapping 
> API"):
>   
>
> > > The second criticism is a matter of taste, perhaps, and the third can
> > > be addressed later:
> > >
> > >  * The mixed call/callback API: [...]
> > >
> > >    Callbacks are increasingly dominating the innards of qemu for good
> > >    reasons.  I think I have explained how a callback style can provide
> > >    the necessary functionality.
> > >       
> > I understand the point you make here and I think we'll end up having a 
> > callback API.  But for reasons I've already mentioned, I don't think it 
> > makes sense for it to be the core API since it introduces very difficult 
> > to eliminate deadlocks.
> >     
>
> I think my most recently sketched-out proposal (where the caller
> passes a DMA sg list to the mapping request function) doesn't have
> these deadlocks ?
>   

But what do you do with a DMA request that cannot possibly be bounced 
but that can be split?  In other words, a DMA request to 1G of MMIO 
memory that is disk access.  We can absolutely split that into multiple 
requests.

> It seems to me that the way to avoid the deadlock is to allow the
> caller to make their whole request at once rather than having two
> callers both racing to grab resources.
>   

In practice, you can't rely on being able to allocate enough memory in 
the host to complete a full request.  This is only true for very special 
devices.  You definitely need to support two distinct modes.

Regards,

Anthony Liguori

> Regards,
> Ian.
>
>
>