[openib-general] How to support IOMMUs for ipath driver

[Ralph Campbell]

On Sun, 2006-09-17 at 12:57 +0300, Or Gerlitz wrote:
> Ralph Campbell wrote:
> > Here is my thinking so far:
> > 
> > The driver is passed an LKEY/RKEY plus an address.
> > For ib_get_dma_mr(), the address is currently from
> > dma_map_single(), dma_map_page(), or dma_map_sg().
> > With the ib_dma_*() routines, I can intercept these calls
> > and return something instead of a bus or IOMMU address.
> > I would like to return a kernel virtual address since that
> > is the simplest and is what I ultimately need. This is
> > trivial for dma_map_single() and trivial for low memory
> > pages for dma_map_page().
> > 
> > I think I can safely just return error for architectures
> > with high memory pages since the driver really only works
> > on 64-bit systems (for a variety of reasons which I won't
> > go into) and those systems don't have high memory.
> 
> Again (and please go and check me), pages you need to DMA (ie move over 
> IB) need **not** be mapped into the kernel virtual address space and 
> this happens **not** only under ia32 high-memory scheme, please see my 
> other email for two examples (direct I/O etc)
> 
> 
> > ib_sg_dma_address would return the page_address() of sg->page
> > but wouldn't be able to rely on other fields which might be in
> > the struct scatterlist.
> 
> your design seems to reply on three fields: page, offset and length, so
> 
> ib_sg_map_sg(scat) is kmap-ping whatever pages which are not mapped now 
> into kvirt
> 
> ib_dma_unmap_sg(scat) is kunmap-ping those pages you were mapping before 
> (you might need an aux data structure to keep which need kunmap)
> 
> ib_sg_dma_address(scat) is page_address(scat->page) + scat->offset
> 
> ib_sg_dma_len(scat) is scat->length
> 
> Or.

Correct.