Re: [Qemu-devel] [RFC] QOMification of AXI streams

[Anthony Liguori <anthony@codemonkey.ws>]

On 06/12/2012 07:37 PM, Benjamin Herrenschmidt wrote:
> Not really no, we don't have proper DMA APIs to shoot from devices.
>
> What the DMAContext patches provide is a generic dma_* API but if we are
> going to get rid of DMAContext in favor of a (modified ?) MemoryRegion
> I'd rather not expose that to devices.
>
> Since I need something _now_ for 1.2 (this has been going on for way too
> long), I'm going to go for a quick kill for PCI&  PAPR VIO only using a
> slightly modified version of the existing iommu patches that provides
> pci_* wrappers to the DMA ops that take the PCIDevice as an argument.
>
> That way we can replace the infrastructure and remove DMAContext without
> affecting devices in a second stage (unless you think you can come up
> with a new scheme in the next few days :-) I really am not familiar
> enough with those parts of qemu to aim for the full schebang for 1.2 but
> maybe you guys can :-)

I think pci_* wrappers is the right thing to do in the short term (and probably 
long term too).

Regards,

Anthony Liguori

>
> Cheers,
> Ben.
>
>> --
>> error compiling committee.c: too many arguments to function
>>
>>
>>
>