On 8/21/2021 12:08 AM, Michael Kelley wrote:

quoted

quoted

quoted

   	}

The whole approach here is to do dma remapping on each individual page
of the I/O buffer.  But wouldn't it be possible to use dma_map_sg() to map
each scatterlist entry as a unit?  Each scatterlist entry describes a range of
physically contiguous memory.  After dma_map_sg(), the resulting dma
address must also refer to a physically contiguous range in the swiotlb
bounce buffer memory.   So at the top of the "for" loop over the scatterlist
entries, do dma_map_sg() if we're in an isolated VM.  Then compute the
hvpfn value based on the dma address instead of sg_page().  But everything
else is the same, and the inner loop for populating the pfn_arry is unmodified.
Furthermore, the dma_range array that you've added is not needed, since
scatterlist entries already have a dma_address field for saving the mapped
address, and dma_unmap_sg() uses that field.

I don't use dma_map_sg() here in order to avoid introducing one more
loop(e,g dma_map_sg()). We already have a loop to populate
cmd_request->dma_range[] and so do the dma map in the same loop.

I'm not seeing where the additional loop comes from.  Storvsc
already has a loop through the sgl entries.  Retain that loop and call
dma_map_sg() with nents set to 1.  Then the sequence is
dma_map_sg() --> dma_map_sg_attrs() --> dma_direct_map_sg() ->
dma_direct_map_page().  The latter function will call swiotlb_map()
to map all pages of the sgl entry as a single operation.

After dma_map_sg(), we still need to go through scatter list again to 
populate payload->rrange.pfn_array. We may just go through the scatter 
list just once if dma_map_sg() accepts a callback and run it during go
through scatter list.