Re: [PATCH][1/2] usb: dma bounce buffer support V3

[Alan Stern <stern@xxxxxxxxxxxxxxxxxxx>]

On Tue, 22 Jan 2008, David Brownell wrote:

> > +static int map_urb_for_dma(struct usb_hcd *hcd, struct urb *urb,
> > +			   gfp_t mem_flags)
> > +{
...
> Right about here, I think, is where a comment is needed to explain
> what's really going on here.  Key points:
> 
> 	- We need "local" memory, canonical example being
> 	  a small SRAM on a discrete controller being the
> 	  only memory that the controller can read ...
> 	  (a) "normal" kernel memory is no good, and
> 	  (b) there's not enough to share
> 
> 	- The only *portable* hook for such stuff in the
> 	  DMA framework is dma_declare_coherent_memory()
> 
> 	- So we use that, even though the primary requirement
> 	  is that the memory be "local" (hence addressible
> 	  by that device), not "coherent".
> 
> Someday DMA memory management may be smart enough to handle this
> stuff without such workarounds.  Having a clear description of
> this particular sub-problem should help.  Having that knowledge
> living outside the source tree ... won't.
> 
> 
> > +		else if (hcd->driver->flags & HCD_LOCAL_MEM)
> > +			ret = hcd_alloc_coherent(
> > +					urb->dev->bus, mem_flags,
> > +					&urb->setup_dma,
> > +					(void **)&urb->setup_packet,
> > +					sizeof (struct usb_ctrlrequest),
> > +					DMA_TO_DEVICE);
> > +	}

Isn't the logical location for these comments at the beginning of 
hcd_alloc_coherent()'s definition, rather than the place where it is 
called?

Alan Stern


-------------------------------------------------------------------------
This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
_______________________________________________
linux-usb-devel@xxxxxxxxxxxxxxxxxxxxx
To unsubscribe, use the last form field at:
https://lists.sourceforge.net/lists/listinfo/linux-usb-devel