All of lore.kernel.org
 help / color / mirror / Atom feed
* Using arch/x86_64/pci-dma in x86_64 Xen
@ 2006-05-04 14:57 Langsdorf, Mark
  2006-05-04 16:18 ` Jon Mason
  2006-05-07 12:19 ` Muli Ben-Yehuda
  0 siblings, 2 replies; 5+ messages in thread
From: Langsdorf, Mark @ 2006-05-04 14:57 UTC (permalink / raw)
  To: xen-devel

The current implementation of pci-dma in x86_64 Xen
Linux is to use the i386 implementation.  I don't
believe this is sustainable in the long term, as
x86_64 Xen is going to need more DMA solutions than
just SWIOTLB and plain hardware as various IOMMU 
implementations arrive.  

The mainstream kernel has a clean abstraction layer
with the dma_ops that should be portable to Xen dom0
and domU.  I've started implementing this in a
pci-dma-xen for x86_64 Xen Linux.  Am I stepping on 
anyone's toes or is there anyone I should be 
co-ordinating this with?

-Mark Langsdorf
AMD, Inc.

^ permalink raw reply	[flat|nested] 5+ messages in thread
* RE: Using arch/x86_64/pci-dma in x86_64 Xen
@ 2006-05-08 13:31 Langsdorf, Mark
  0 siblings, 0 replies; 5+ messages in thread
From: Langsdorf, Mark @ 2006-05-08 13:31 UTC (permalink / raw)
  To: Christian.Limpach, Muli Ben-Yehuda; +Cc: xen-devel

> > > The mainstream kernel has a clean abstraction layer
> > > with the dma_ops that should be portable to Xen dom0
> > > and domU.  I've started implementing this in a
> > > pci-dma-xen for x86_64 Xen Linux.  Am I stepping on 
> > > anyone's toes or 
> > > is there anyone I should be co-ordinating this with?
> 
> I think that using the dma_ops abstraction layer on x86_64 is 
> definitely the way to go.  It's just not clear how the Xen 
> specific dma ops should be implemented in the end, whether 
> having a common implementation is worthwhile considering that 
> it will require glue code to make it fit with the different 
> interfaces.

I don't think there's enough differences between the bare
metal dma ops and the Xen dma ops to justify a common
implementation.  Also, the common implementation implies
that each IOMMU interface needs to have some knowledge of
the other interfaces, and I don't think that's sustainable
long term.

-Mark Langsdorf
AMD, Inc.

^ permalink raw reply	[flat|nested] 5+ messages in thread

end of thread, other threads:[~2006-05-08 13:31 UTC | newest]

Thread overview: 5+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2006-05-04 14:57 Using arch/x86_64/pci-dma in x86_64 Xen Langsdorf, Mark
2006-05-04 16:18 ` Jon Mason
2006-05-07 12:19 ` Muli Ben-Yehuda
2006-05-07 19:45   ` Christian Limpach
  -- strict thread matches above, loose matches on Subject: below --
2006-05-08 13:31 Langsdorf, Mark

This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.