[PATCH 2/2] DMAENGINE: driver for the ARM PL080/PL081 PrimeCells

[Dan Williams]

On Tue, Jun 29, 2010 at 4:36 AM, Linus Walleij
<linus.walleij at stericsson.com> wrote:
> This creates a DMAengine driver for the ARM PL080/PL081 PrimeCells
> based on the implementation earlier submitted by Peter Pearse.
> This is working like a charm for memcpy and slave DMA to the PL011
> PrimeCell on the PB11MPCore.
>
> This DMA controller is used in mostly unmodified form in the ARM
> RealView and Versatile platforms, in the ST-Ericsson Nomadik, and
> in the ST SPEAr platform.
>
> It has been converted to use the header from the Samsung PL080
> derivate instead of its own defintions. The Samsungs have a custom
> driver in their mach-* folders though, atleast we can share the
> register definitions.
>
> Cc: Peter Pearse <peter.pearse at arm.com>
> Cc: Ben Dooks <ben-linux at fluff.org>
> Cc: Kukjin Kim <kgene.kim at samsung.com>
> Cc: Alessandro Rubini <rubini at unipv.it>
> Cc: Viresh Kumar <viresh.kumar at st.com>
> Signed-off-by: Linus Walleij <linus.walleij at stericsson.com>
> ---
> Changes since last revision: this has now been successfully tested
> with PL011 UART on the PB11MPCore.
>
> It now uses the generic runtime slave control and not the PrimeCell
> extensions, and thus depends only on that patch set. With the previous
> patch to the header file this is cleanly separated from the ARM tree
> and any PrimeCell drivers and can go into the async_tx tree as it is
> without further dependencies.
> ---

Russell I am assuming this implementation and testing is sufficient to
allay your prior nak [1]?  Linus is oversubscribing virtual channels
to physical channels to handle muxing.

> +#ifdef MODULE
> +
> +# error "AMBA PL08X DMA CANNOT BE COMPILED AS A LOADABLE MODULE AT PRESENT"
> +
> +/*
> +       a) Some devices might make use of DMA during boot
> +          (esp true for DMAENGINE implementation)
> +       b) Memory allocation will need much more attention
> +          before load/unload can be supported
> + */
> +#endif

This is taken care of by being "config bool", but I assume you want
this for documentation purposes?

> +       txd = kzalloc(sizeof(struct pl08x_txd), GFP_KERNEL);

...appears in the prep routines.  Will this driver be used by any
storage controllers on the platfom?  Might we deadlock waiting on i/o
that needs to allocate a descriptor to complete?

--
Dan

[1]: http://marc.info/?l=linux-arm-kernel&m=127350477914098&w=2