Re: [RFC] dma: passing "attributes" to dma_map_* routines

[Stefan Richter <stefanr@s5r6.in-berlin.de>]

akepner@sgi.com wrote:
> Reorderings are possible on reads and
> writes. Things get synced up by either an interrupt or a write to
> a memory region with a "barrier attribute". Memory allocated with
> dma_alloc_coherent() gets the barrier attribute. The idea here is
> to allow memory allocated with plain old malloc() or whatever to
> get the same attribute.

>From its purpose it sounds like you need this only for few special
memory regions which would typically be mapped by dma_map_single() and
furthermore that drivers who need this behavior will be changed to
explicitly demand it.  If so, a nonintrusive API extension could simply
be to add an

dma_addr_t dma_map_single_write_last(struct device *dev, void *ptr,
size_t size, enum dma_data_direction direction);

or however you'd like to call it.  (DMA-maps a memory area for which it
is guaranteed that of all DMA writes pending at any time, a DMA
reordering interconnect --- if such an interconnect is present --- will
perform DMAs to other areas first and to this area last.  On machines
which don't reorder DMAs, this function is the same as dma_map_single().)

However, your older patch series looks like you want this behavior also
in areas which are mapped by dma_map_sg(), do you?.  Still, adding two
functions of the kind like above, if necessary, might still be
preferable to changing the call parameters of existing functions or to
overloading enum dma_data_direction.

So that would be option 3) of yours, though without your attrs
parameter.  Do you expect the need for even more flags for other kinds
of special behavior?
-- 
Stefan Richter
-=====-=-=== ==-- =--=-
http://arcgraph.de/sr/