Re: [RFC] move dma_head/dma_maps out of skb_shared_info and into sk_buff

[Alexander Duyck <alexander.duyck@gmail.com>]

On Thu, 2009-11-05 at 17:40 -0800, David Miller wrote:
> From: Alexander Duyck <alexander.h.duyck@intel.com>
> Date: Thu, 05 Nov 2009 16:54:33 -0800
> 
> > During testing we found issues with the use of skb_dma_map/unmap on
> > systems that had iommu enabled and were configured to use a bridge.
> > The
> > issue is that if two ports are members of the same bridge, and a
> > broadcast packet is sent out on the bridge skb_clone will be used to
> > send a copy to all ports, but the clones run into issues because the
> > dma mappings for the cloned skbs all share the shared_info structure
> > where the dma mappings are stored.
> > 
> > To resolve that this patch moves those dma mappings out of the
> > shared_info structure and into the sk_buff itself.  This allows cloned
> > skbs to be mapped separately without causing dma unmapping errors.
> > 
> > Signed-off-by: Alexander Duyck <alexander.h.duyck@intel.com>
> 
> DMA mappings are a property of the data not the SKB.
> 
> And the DMA mappings are in the shared area exactly so that
> we don't eat the space for every clone of the data.
> 
> We need another way around this problem and I'll just as
> soon remove the DMA mapping bits entirely from SKBs before
> I'll put them into struct sk_buff

We may need to pull the dma mapping bits entirely then since we end up
effectively needing to store a copy of the dma mapping every time
dma_map_single/page is called w/ iommu enabled, otherwise we can only
map the data portion of the skb once without leaking dma mappings.

Would there be any complaint if I were to submit a patch that removed
skb_dma_map/unmap?

Thanks,

Alex